Functional correlates of preserved naming performance in amnestic Mild Cognitive Impairment

Neuropsychologia 76 (2015) 136–152

Contents lists available at ScienceDirect

Neuropsychologia journal homepage: www.elsevier.com/locate/neuropsychologia

Functional correlates of preserved naming performance in amnestic Mild Cognitive Impairment Eleonora Catricalà a,n,1, Pasquale A. Della Rosa b,1, Laura Parisi c, Antonio G. Zippo b, Virginia M. Borsa c, Antonella Iadanza d, Isabella Castiglioni b, Andrea Falini d, Stefano F. Cappa a,c a

Institute for Advanced Study IUSS Pavia, Palazzo del Broletto - Piazza della Vittoria n.15, 27100, Pavia, Italy Institute of Molecular Bioimaging and Physiology, National Research Council (IBFM-CNR), Milan, Italy c Division of Neuroscience, San Raffaele Scientific Institute, Milan, Italy d Department of Neuroradiology and CERMAC, San Raffaele Scientific Institute, Vita-Salute San Raffaele University, Milan, Italy b

art ic l e i nf o

a b s t r a c t

Article history: Received 23 July 2014 Received in revised form 4 January 2015 Accepted 5 January 2015 Available online 8 January 2015

Naming abilities are typically preserved in amnestic Mild Cognitive Impairment (aMCI), a condition associated with increased risk of progression to Alzheimer's disease (AD). We compared the functional correlates of covert picture naming and word reading between a group of aMCI subjects and matched controls. Unimpaired picture naming performance was associated with more extensive activations, in particular involving the parietal lobes, in the aMCI group. In addition, in the condition associated with higher processing demands (blocks of categorically homogeneous items, living items), increased activity was observed in the aMCI group, in particular in the left fusiform gyrus. Graph analysis provided further evidence of increased modularity and reduced integration for the homogenous sets in the aMCI group. The functional modifications associated with preserved performance may reflect, in the case of more demanding tasks, compensatory mechanisms for the subclinical involvement of semantic processing areas by AD pathology. & 2015 Elsevier Ltd. All rights reserved.

Keywords: Amnestic Mild Cognitive Impairment fMRI Functional connectivity Graph analysis Picture naming Semantic context effect Living–nonliving Fusiform gyrus

1. Introduction The neural basis of semantic processing has been addressed by many functional neuroimaging studies in healthy subjects. Recent meta-analyses have underlined the role of a distributed network involving several regions, i.e., the posterior inferior parietal lobule (angular gyrus and portions of supramarginal gyrus), middle temporal gyrus, fusiform and parahippocampal gyri, dorsomedial prefrontal cortex, inferior frontal gyrus, ventromedial prefrontal cortex, and posterior cingulate gyrus (Binder et al., 2009), and the specific contribution of the anterior temporal lobe (ATL) as a crucial hub for semantic processing (Visser et al., 2010). Studies of patients with neurodegenerative conditions provide considerable additional evidence on the neural substrate of semantic memory. In particular, patients suffering from Alzheimer's disease (AD) show semantic memory deficits (Chertkow and Bub, 1990; Hodges, 2006), characterized by a greater impairment of n

Corresponding author. Fax: þ39 0382 375899. E-mail address: [email protected] (E. Catricalà). 1 These authors contributed equally to this work.

http://dx.doi.org/10.1016/j.neuropsychologia.2015.01.009 0028-3932/& 2015 Elsevier Ltd. All rights reserved.

subordinate than superordinate knowledge (Chertkow and Bub, 1990; Alathari et al., 2004; Duarte et al., 2009; Garrard et al., 2005; Giffard et al., 2001, 2002), and of biological more than non-biological entities (Fung et al., 2001; Gonnerman et al., 1997; Whatmough et al., 2003; Catricalà et al., 2014). Naming abilities are impaired from the early stages of the disease, with errors consisting of anomias and semantic paraphasias or superordinate responses. These disorders are generally attributed to an underlying semantic memory deficit, due to loss (Chertkow and Bub, 1990; Hodges et al., 1992; Martin and Fedio, 1983; Lin et al., 2014), or defective access to semantic information (Nebes et al., 1989; Nebes 1992; Ober and Shenaut, 1988). The possible contribution of lexical impairment is also discussed, at least for the early phases of disease (Funnell and Hodges, 1991). Consistent with the hypothesis of a underlying lexical-semantic deficit for the naming impairment in AD, PET and structural MRI studies have found that naming deficits are predominantly associated with abnormalities in the temporal cortex (Teipel et al., 2006; Hirono et al., 2001; Melrose et al., 2009; Domoto-Reilly et al., 2012), including left fusiform gyrus (Teipel et al., 2006; Melrose et al., 2009) and the left ATL (Domoto-Reilly et al., 2012; Frings et al., 2011; Melrose et al., 2009), and extending to frontal and parieto-occipital cortex (Apostolova et al., 2008).

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

Zahn et al. (2004) have shown that metabolism in left anterior temporal, posterior inferior temporal, inferior parietal and medial occipital areas (Brodmann areas: 21/38, 37, 40 and 19) correlated with both verbal and nonverbal semantic performance in AD. Similarly, Libon et al. (2013) found that gray matter atrophy affecting the left posterolateral temporal–parietal regions (including fusiform, parahippocampal, superior temporal and inferior parietal (BA 40) gyri) predicted the AD performance on a semantic categorization task. Changes in brain activation and in functional connectivity may precede structural changes and behavioral manifestations (Wierenga and Bondi, 2007; Bokde et al., 2006). In this view, functional magnetic resonance imaging (fMRI) activation studies may represent a possible preclinical biomarker of AD, as they are able to detect early abnormalities in brain function and to highlight the defective integration of different regions into a neural network for successful completion of a task that could reflect initial stages of AD neuropathology. AD is often preceded by a phase characterized by an isolated episodic memory impairment, defined as amnestic Mild Cognitive Impairment (aMCI) (Petersen, 2004). While the majority of structural imaging has focused on the involvement of the medial temporal lobe structures (Varon et al., 2014), functional PET imaging has shown more widespread changes in aMCI, including an involvement of the distributed semantic network, i.e., in the inferior parietal and posterior cingulate cortex and in the left parahippocampal gyrus (Anchisi et al., 2005). An MR study reported that atrophy of the middle and inferior temporal gyrus and the fusiform gyrus (in addition to anterior medial temporal regions) had predictive value for progression to AD in Mild Cognitive Impairment subjects (Convit et al., 2000). As an episodic memory deficit is the hallmark and the major symptom of AD, the majority of the functional activation studies on patients in the prodromal stage of AD focused on memory tasks (Sugarman et al., 2012; Wierenga and Bondi, 2007). The results, however, are discordant (showing both hypoactivity and hyperactivity), and difficult to interpret, given the associated defective performance (Sperling et al., 2010; Wierenga and Bondi, 2007). An additional problem is that episodic memory decline is also present in normal aging (Nilsson, 2003), and thus does not necessarily reflect AD neuropathology. The findings may thus simply reflect defective performance, rather than disease-specific functional changes (Woodard et al., 2009). The use of tasks that patients can perform accurately may be useful to prevent ambiguities in results interpretation due to defective performance (Price and Friston, 1999; Wierenga and Bondi, 2007), allowing valid inferences on the pathogenesis of cognitive alterations and on the contribution of specific brain regions to cognitive processes. Unlike episodic memory skills, semantic memory abilities remain relatively intact both in normal aging (Nilsson, 2003) and in aMCI, at least when assessed with clinical tests (Joubert et al., 2008). Naming deficits are not prominent in aMCI, who may have mild, subclinical naming disorders. Deficits in naming objects are in fact an uncommon finding (Balthazar et al., 2008; Clague et al., 2011; Adlam et al., 2006; Choi et al., 2013; Gardini et al., 2013; but see Joubert et al., 2010 and Ahmed et al., 2008), while naming pictures of unique entities including famous people, famous buildings and famous public events seems to be more consistently affected (Joubert et al., 2010; Gardini et al., 2013; Ahmed et al., 2008; Estévez-González et al., 2004; Clague et al., 2011). On the other hand, disease progression is associated with semantic memory dysfunction, and there is some evidence that subtle semantic deficits (semantic fluency) are a predictor of progression towards dementia (Gainotti et al., 2014). All the outlined clinical evidence indicates semantic processing as a possible subclinical marker of AD pathology, leading us to

137

assess in the present context if alterations of the neural underpinnings of semantic processing at the neural level are already present in aMCI subjects showing a normal performance on picture naming task. Our prediction is that functional changes can be already observed in these subjects in regions subserving or contributing to semantic processing. An hyperactivation of the specific “semantic” circuits, and/or an involvement of additional regions not normally engaged in healthy individuals (compensatory “recruitment”) may represent the mechanisms supporting the preserved naming performance. As a control condition, we used a task of reading regular words. This is a performance with limited requirements for semantic processing (Woollams et al., 2007), which has been shown, accordingly, to be spared until the late stages of AD (Colombo et al., 2000). To further increase the sensitivity of the task for subtle semantic dysfunction, we used a naming paradigm taking into account the influence of the semantic context. Behavioral studies in healthy subjects documented greater latencies in naming objects blocks belonging to the same semantic category (homogeneous condition) when compared to blocks including different categories (heterogeneous condition) (Damian et al., 2001; Levelt et al., 1999; Kroll and Stewart, 1994). An fMRI study in healthy subjects revealed increased perfusion fMRI signal bilaterally in the hippocampus and in the left middle to posterior superior temporal cortex for the homogeneous condition (Hocking et al., 2009). In addition, we looked for the possible influence of the semantic category, which has been shown to affect naming performance in AD. An inferior performance for living than non-living entities has been repeatedly reported in AD (Fung et al., 2001; Gonnerman et al., 1997; Whatmough et al., 2003). We thus hypothesized that living things may be particularly sensitive to early subclinical dysfunction. In the homogeneous condition, we created different blocks for living and non-living things. The two classes of entities were balanced for a large number of intrinsic variables that could account for dissociations, in order to exclude that differences in brain activity could be ascribed to different stimulus processing demands. For example, the activations in the premotor cortex generally reported for tools appear to be due to the manipulability of the objects, as they are found when tools are compared with animals but not with vegetables (Devlin et al., 2002; Gerlach et al., 2002). In order to avoid spurious activation differences between living and non-living items, the two categories were balanced also for volumetric manipulability (i.e., how an object is associated to gestures used to pick up the object).

2. Methods 2.1. Subjects We screened a large population of subjects with a clinical diagnosis of aMCI (Petersen et al., 2001) and we enrolled patients who fulfilled the following criteria: 1) no evidence of other causes of memory impairment as demonstrated by neuroimaging (standard brain MR scan) and laboratory tests (blood tests for hepatic and kidney functions, electrolytes, glucose, thyroid functions, lipids, syphilis infection, B12 vitamin and folic acid deficits) 2) a selective and isolated deficit of episodic memory as investigated with the Rey Auditory Verbal Learning Test (Rey AVLT) (Table 1); 3) semantic memory integrity, as revealed by a detailed semantic memory battery, including two naming tasks, one with colored pictures and the other in response to an oral description, a word-picture matching task, a picture sorting task and a sentence verification task (Catricalà et al., 2013). Eight patients were thus enrolled in the study (5 males; age range: 59–77, mean¼71.62; education range: 5–18). Sixteen controls (8 males) matched for education (education

138

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

Table 1 Demographical and neuropsychological data of aMCI. P1

P2

P3

P4

P5

P6

P7

P8

M 74 18 24

M 77 13 28

F 75 8 30

F 72 12 26

M 72 11 28

M 69 5 27

M 75 17 29

F 59 18 30

Neuropsychological assessment Token Test 33 Phonological fluency 46 Category fluency 36 Digit Span 6 Digit Span Backward 4 Corsi Test 4 Raven Matrices 26 Attentional Matrices 47 Copy of Rey figure 31

32 27 34 5 na 4 26 Na 26

33 30 40 5 4 4 33 54 35

35 57 40 6 4 4 27 59 34

31 35 24 5 4 6 28 54 36

29 22 34 4 4 4 25 45 33

34 27 46 6 3 4 27 57 36

na 50 41 7 4 5 32 51 34

29 2a

35 0a

22a 3a

19a 2a

34 3a

43 3a

13

7.5

12

17

9

7a

47 45

48 47

47 38

47 38

43 42

48 41

Demographic data Gender Age (years) Education MMSE

Episodic memory Rey AVLT Immediate recall Delayed recall

34 0a

Delayed recall of Rey figure

8

23 4.5a/ 7.5a,b 6a

48 43

46 43

Semantic Memory Picture naming (48) Naming to verbal description (48) Word picture matching (48) Picture sorting task (192) Sentence verification (480)

48 48 183 189 477 466

48 48 48 48 48 48 189 192 192 187 na 191 470 na 461 453a 471 473

a

Pathological performance on the base of normative data; na ¼not available. Episodic memory was assessed trough delayed recall logical memory and words pairs association tests. b

range: 8–18; p ¼0.89) but not for age (age range: 56–76, mean ¼63.44; p o0.05) with the 8 patients were also included. None of the controls had a history of neurological illness or mental decline, and all had an adjusted score on the MMSE (Mini mental state examination, Folstein et al., 1975) Z25.44. All subjects gave informed written consent to the experimental procedure, which was approved by the Ospedale San Raffaele Ethics Committee. 2.2. Stimuli The same 32 items, 16 living and 16 non-living, selected from Catricalà et al. (2013), were used for picture naming and reading tasks. Photographs were obtained from Viggiano et al. (2004). Living items belonged to three distinct semantic categories (i.e., animals, fruits and vegetables), and non-living items belonged to four distinct semantic categories (i.e., vehicles, tools, kitchen utensils, furniture). In each category there were 4 items, with the exception of the animal category in which we selected 8 items. In addition, 4 categories were manipulable (fruits and vegetables for the living items and tools and kitchen tools for the non-living things) and 4 non-manipulable (2 for animals for the living and furniture and vehicles for the non-living). In order to distinguish between manipulable and non-manipulable items, we used the volumetric manipulability values taken from Catricalà et al. (2013). In particular, items with manipulability values between the lower value of the manipulability distribution and the 40th percentile were considered as non-manipulable, whereas items with values ranging from the 60th percentile to the highest value of the distribution were considered as manipulable. The same 32 stimuli were used to build 8 homogeneous and

8 heterogeneous blocks. Each homogeneous block consisted of 4 items belonging to the same category. Based on a normative study on semantic features (Catricalà et al., 2013), in order to maximize the semantic similarity we chose stimuli belonging to the same category with at least five features in common. Four homogeneous blocks were composed of living items (two were also manipulable and two were not manipulable) and the other four blocks were composed of non-living things (of which two were manipulable and two were not manipulable). In the heterogeneous condition, the same items were used, and each block consisted of 4 pictures belonging to different categories: one living manipulable (e.g., pineapple), one living nonmanipulable (e.g., giraffe), one non-living manipulable (e.g., drill) and one non-living non-manipulable item (e.g., helicopter). Thirty two non-pictures were generated in order to create a baseline condition for the picture naming task. The 32 previously selected images were manipulated using Photoshop. A 64  64 pixels matrix mosaic was created for each image. Margins were vanished with a Gaussian filter. We used these stimuli to create 16 non-picture blocks corresponding to the homogeneous and heterogeneous ones. Similarly, for the “non-word” reading task, 32 strings of letters (“non-words”) based on the words were created. The total number of vowels and consonants, the consonant/vowel ratio and the frequency of each letter for block type was computed. The latter variable was used to generate letter strings (“non-words”) representative of both consonants and vowels present in each of the 32 words with pair-wise word-length matching (p ¼ 0.84) and with a consonant/vowel ratio comparable to the corresponding block of words (p at least ¼ 0.61). The letter strings (“non-words”) were generated according to 4 different consonant–vowel (cv) sequences, namely: ccvv, cccvvv, vvcc, vvvccc. Thus, items were organized in the following experimental conditions: 8 homogeneous pictures/words blocks (i.e., 2 living manipulable (fruits and vegetables), 2 living non-manipulable (animals), 2 non-living manipulable (tools, kitchen tools), 2 nonliving non-manipulable (vehicles and furniture); 8 heterogeneous pictures/words blocks; 8 homogeneous non-pictures/non-words blocks, 8 heterogeneous non-pictures/non-words blocks. In order to control for the visual similarity of items across blocks, we administered a rating task to 16 subjects (8 males), who did not participate in the subsequent fMRI experiment. The age of the subjects ranged from 55 and 85 years (age mean ¼64.43; SD ¼ 9.10). The aim was to obtain quantitative judgements about how similar in appearance the four stimuli of each block were. We used the same homogenous and heterogeneous blocks described above. Two different blocks were used as examples, the first block composed of items with high visual similarity but belonging to different semantic categories; the latter was characterized by items with low visual similarity, but sharing the same semantic category. Pictures were displayed on the monitor of the computer; we presented each block at a time with a PowerPoint presentation. The subject had to say how similar were the 4 images. The scale of similarity ranged from 1 to 7, in which 1 indicated not at all similar and 7 indicated very similar. The participants were instructed to ignore semantic similarity (the category to which the 4 pictures belonged) and to judge similarity on the basis of visual appearance (shape, color and size). The instructions for similarity were based largely on those used by Damian et al. (2001), but with a different range. The order of the items in each block was randomized (4 different list versions were created) as the order of the blocks, with a randomization between homogeneous and heterogeneous blocks. Mean rated visual similarity for homogeneous blocks was 4.81 (SD ¼0.78), while for heterogeneous blocks it resulted to be significantly lower (mean ¼1.52; SD ¼ 0.23) (p o0.001). Each homogeneous block was finally associated with a

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

heterogeneous block and matched pairwise for visual familiarity (p ranging from 0.096 to 0.86), visual complexity (p ranging from 0.14 to 0.99) and the total number of distinctive features in each block (p at least ¼0.99). Specifically, with regard to the word stimuli for the reading task, homogeneous and heterogeneous blocks were matched for bigram frequency (p ranging from 0.08 to 0.85), number of letters (p ranging from 0.18 to 0.84, with the exception of a p value ¼0.01), number of syllables (p ranging from 0.18 to 1), total number of letters in each block (p ¼1) and Token scores (p ranging from 0.27 to 0.99). In addition, for each word, the block onset weight was computed as a measure of the position and the total number of shared letters at onset between each word and all the words present in each block (i.e., if two words shared the first letter a value of 1 was assigned, the second letter a value of 0.5, the third a value of 0.25). Homogeneous and heterogeneous blocks were matched pairwise for onset weight (p ¼0.870) A very strict matching between living and non-living was achieved. Values of visual, psycholinguistic and semantic variables were taken from Viggiano et al. (2004), Dell’Acqua et al. (2000), Della Rosa et al. (2010), LEXVAR database (Barca et al., 2002) and from Catricalà et al. (2013). Considering the Homogeneous condition, living vs non-living were matched for bigrams frequency (p ¼0.250), number of letters (p ¼0.116), number of syllables (p ¼0.468), number of letters for block (p ¼0.285), letters frequency (p¼ 0.992), visual familiarity (p ¼0.862), visual complexity (p ¼0.993), name agreement (p ¼0.98), frequency of word use (p ¼0.08, livingo nonliving), concreteness (p ¼0.152), imageability (p ¼0.824), age of acquisition (p ¼0.359), familiarity (p ¼0.623), context availability (p¼ 0.616), emotional valence (p¼ 0.218), volumetric manipulability (p ¼0.42), semantic distance (p ¼0.387), number of distinctive features (p ¼0.609), Garrard's distinctiveness (p ¼0.375; Garrard et al., 2001) and for number of distinctive features for each word over the total number of distinctive features for each block (p ¼1). In addition the four living blocks and the 4 non-living blocks were matched for visual similarity (p ¼ 0.144). However, living and nonliving concepts could not be matched for functional manipulability (p o0.001) and typicality (p ¼0.051), with higher values for nonliving.

139

2.3. Experimental paradigm The experiment consisted of four runs (i.e., two runs for picture naming and two runs for word reading). Each experimental run consisted of: 4 homogeneous blocks (HOM) and 4 heterogeneous (HET) blocks of pictures or words. The subjects were asked to name each picture or read each word covertly. 4 homogeneous blocks of non-pictures or non-words (the nonpictures/non-words were obtained from the pictures/words of the respective homogeneous block) and 4 heterogeneous blocks of non-pictures or non-words (the non-pictures/non-words were obtained from the pictures of the respective heterogeneous block). The subjects were asked to covertly say “ok” after each non-picture or non-word. These blocks were used as high-level baselines in order to maximize the specificity of brain activation in areas sensitive to semantic processing, while canceling out brain activity elicited by complex higher-order visual processing or by speech production (Price et al., 2005). Thus, for each run a total of 16 blocks were presented, leading to a total of 64 trials per run (16 blocks  4 items in each block). Each homogeneous and heterogeneous picture or word block was followed or preceded by its corresponding non-picture or non-word block; items were not repeated more than twice in each run. Moreover, for each run the order of blocks was pseudo-randomized, in order to avoid that blocks of the same experimental condition were presented in a consecutive order (e.g., homogeneous-block of pictures followed by another homogeneous block of pictures). Two different lists were created for each task (i.e., picture naming – PN; word reading – WR), containing the same items, but with a different order between blocks and for items within blocks. All subjects performed both lists for each task (i.e., PN-1 and PN-2; WR-1 and WR-2), however the order of presentation of the two lists for each task as well as the order of the two tasks was counterbalanced between subjects. The block presentation was as follows: (i) At first the instructions “fix the cross” appeared for 2000 ms, then a fixation cross ( þ) was presented at the center of a computer screen for 12000 ms. (ii) Instructions specific for the block appeared and stayed on the screen for 2000 ms. (iii) The fixation cross appeared again and stayed on the screen for 12000 ms.

Fig. 1. Schematic representation of the picture naming task in the homogeneous condition (vehicle category).

140

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

(iv) Four visual stimuli, belonging to each block, were presented on a black background for 2200 ms, replaced immediately by a centrally positioned fixation cross (see Fig. 1). An alternation of stimulus and fixation cross presentation was repeated for a total of 4 times. The ISI (interstimulus interval) between pictures/words or non-picture/non-words lasted for 1750 or 2250 or 3000 ms and different ISIs were randomized within blocks. Each run lasted 11 min, for a total experimental time of 45 min. Subjects were asked to respond covertly as quickly but as accurately as possible and were instructed to name or read as silently as possible to reduce severe movement artifacts in the EPI data potentially due to overt naming or reading. Prior to fMRI scanning (range 2–10 days before), subjects were asked to perform the entire experiment on a laptop (i.e., the 2 PN runs and 2 WR runs) outside the scanner in order to familiarize with both the entire set of stimuli and the tasks. Behavioral voice onset times (VOTs) for each item and averaged across block type for both PN and WR were also collected. Outside the scanner, all the subjects were instructed to overtly name the presented objects as quickly and as accurately as possible. Vocal responses and VOTs were recorded via a microphone in order to assess behavioral performance. The stimuli were presented with the stimulus delivery program, Presentation software V10.3 (NeuroBehavioral Systems Inc., Albany, CA). Inside the scanner, stimuli were presented to participants via a PC placed outside the magnet room. A projector delivered stimuli on a translucent screen placed at the foot of the magnet bore. Participants viewed the screen through a mirror system attached to the top of the head coil. 2.4. fMRI data acquisition An fMRI-blocked technique was used (3 T Intera Philips body scanner, Philips Medical Systems, Best, NL, 8 channels-sense head coil, sense reduction factor¼ 2, TE ¼30 ms, TR ¼3000 ms, FOV ¼240  240, matrix size ¼96  96, 51 contiguous axial slices per volume, 234 volumes for each run, slice thickness¼2.5 mm, gap ¼0.2 mm). Furthermore, optimal EPI parameters at 3 T were defined in order to gain BOLD sensitivity in the temporal poles and anterior temporal lobes (Weiskopf et al., 2006). Specifically, in order to minimize susceptibility induced artifacts and signal dropouts in the anterior temporal lobes and the temporal poles, the slice tilt was set to þ 30°, denoting a tilt of the anterior edge of the slice towards the feet. The phase encoding (PE) gradient polarity was chosen to be negative with the phase encoding direction going form the anterior part to the posterior part of the brain. Five dummy scans preceded each run, all of which were then discarded prior to data analysis to optimize EPI image signal. For each subject a high-resolution structural image was acquired for means of coregistration, segmentation and spatial normalization of the EPI scans (MPRAGE, 150 slice T1-weighted image, TR ¼8.03 ms, TE ¼4.1 ms; flip angle ¼8°, TA ¼4.8 min, resolution¼1 mm  1mm  1 mm) in the axial plane. 2.5. fMRI data preprocessing Data were preprocessed and analyzed using SPM8 (Statistical Parametric Mapping; Wellcome Department of Cognitive Neurology, London, UK). Prior to analysis, all images for four sessions underwent a series of preprocessing steps. Time series diagnostics using tsdiffana (Matthew Brett, MRC CBU: http://imaging.mricbu. cam.ac.uk/imaging/DataDiagnostics) were run to verify the quality of the functional data in terms of variance of corresponding voxels between slices and between volumes relative to mean intensity

values calculated respectively for each image or the entire time series. The ArtRepair toolbox (version 4; http://spnl.stanford. edu/ tools/ArtRepair; Mazaika et al., 2009) was used to identify slices and volumes with significant artifacts based on scan-to scan motion (1 SD change in head position) and outliers relative to the global mean signal (3 SD from the global mean). Outlier slices or volumes were then repaired by interpolation with the nearest uncorrupted neighbors. The percentage of slices or volumes with artifacts never exceeded 10% of total slices or volumes in each session, thus no subject was discarded. A subject-specific field-map with the same EPI parameters was acquired and used for distortion correction. For each subject, all EPI images were realigned to the first volume in each time series and successively to the mean and then unwarped by applying a Voxel Displacement Map (VDM) computed on the basis of the EPIbased field maps in order to control for susceptibility-by-movement's generated variance in the time series. ArtRepair toolbox (Mazaika et al., 2009) was also used to correct for movement artifacts and remove residual interpolation errors after the realign and reslice operations (Grootoonk et al., 2000). Realigned functional volumes were first motion-adjusted, and outlier volumes (i.e. head position change exceeding 0.5 mm) were then replaced by linear interpolation between the closest non-outlier volumes. Residual interpolation errors after realignment were finally removed on realigned, unwarped and resliced images (Mazaika et al., 2009). The motion-corrected and unwarped EPI dataset for each subject was then coregistered to the T1-weighted structural image using the mutual information algorithm in SPM8 (Collignon et al., 1995) in order to generate normalization parameters through diffeomorphic image registration (DARTEL) in SPM8 (Ashburner, 2007). Structural images for both aMCI and control subjects were first segmented into their gray matter (GM), white matter (WM) and cerebrospinal fluid (i.e., csf) tissue components in both native and DARTEL imported space. The imported GM and WM maps were then used to achieve more accurate inter-subject alignment using DARTEL by generating a population-specific average template image, to which each subject GM and WM map is iteratively aligned. This procedure creates invertible and smooth deformations for each subject's native space gray matter image to a common coordinate space, thereby producing a template consisting of images from all subjects in both groups, thus representative of the brain size and shape of each participants. The flow fields that describe the subject-specific spatial deformations were then applied to each subject's EPI data to warp the images to MNI space. The normalized EPI images were then smoothed using a 6-mm full-width at halfmaximum (FWHM) Gaussian kernel to ensure that the data were normally distributed and to account for any between-subject residual variations prior to entering statistical analysis. 2.6. fMRI data analysis We adopted a two-stage random-effects approach to ensure generalizability of the results at the population level (Penny and Holmes, 2003). At the individual level, normalized and smoothed images were entered in two independent general linear models (GLMs). In the first GLM, we considered the two different types of blocks (HOM and HET) and the two tasks (PN and WR). This model included 4 regressors for each task (pictures/non-pictures  HOM/ HET; word/non-word  HOM/HET), 1 regressor modeling instructions preceding each block and 1 for fixation cross blocks resulting in a total of 6 regressors for each of the 4 experimental sessions (i.e., 2 PN and 2 WR). In the second GLM, two additional factors were considered for

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

the set of homogeneous blocks category (living/non-living) and manipulability (manipulable/non-manipulable) for a total of 4 regressors for the homogenous picture or word blocks and 4 regressors for the corresponding non-picture or non-words blocks. The heterogeneous picture/word blocks, corresponding non-picture/non-word blocks, instructions and fixation cross blocks were modeled as separate regressors as for the first GLM resulting in a total of 12 regressors for each of the 4 experimental sessions (i.e., 2 PN and 2 WR). Head motion regressors were not included in either GLM 1 or GLM 2, given that the motion adjustment algorithm of ArtRepair toolbox (Mazaika et al., 2009) had already accounted for these, by correcting the interpolation errors caused by realignment of large motions. This procedure enables the application of motion correction before scans enter GLM, avoiding to model realignment parameters as extra regressors in the design matrix. The block conditions were modeled by convolving a box-car function of each block with a “canonical” hemodynamic response as the basis function (hrf) to create regressors of interest, while the instructions were time-locked to the onset of the stimulus by convolving a stick function with the hrf. Data were highpass-filtered at 1/128 Hz to remove low-frequency signal components and were then analyzed in SPM8 (http://www.fil.ion.ucl.ac.uk/spm). Temporal autocorrelation was modeled using a first order autoregressive process. The first GLM crossed “block type” (HOM/HET)  “task” (PN/ WR), thus the difference between HOM/HET blocks for pictures or words and the matching non-pictures/non-words blocks was tested with linear contrasts of the parameter estimates, resulting in 4 contrast images for each subject. In the second GLM the factors “category” (living/non-living) and “task” (PN/WR) were considered exclusively for the homogenous set of blocks and a total of 4 linear contrasts coding the difference between living/non-living blocks of pictures or words and their corresponding blocks of non-pictures/ non-words were computed at the individual level. 2.6.1. PN and WR within each group and between groups Second level paired-sample t-tests were performed for each of the two groups (i.e., Controls and Patients) using participant as a random factor including contrast images for PN or WR coding the difference between HET blocks and corresponding non-pictures/ non-words blocks in order to evaluate 1) the pattern of brain activity elicited by each task in the less semantically demanding condition (Hocking et al., 2009) and 2) functional differences between the two tasks (i.e., PN and WR) when naming or reading stimuli of different categories, irrespective of the semantic relationship between exemplars (heterogeneous condition). In addition, independent two-sample t tests were performed on the same first-level contrast images in order to assess betweengroup differences (i.e., controls vs patients and vice versa) for each task. Activations surviving a voxelwise statistical threshold of p o0.001 uncorrected at the voxel level and a FWE-corrected extent threshold are reported. Where no significant pattern of activation was evident, the spatial extent threshold was lowered to clusters of Z10 voxels in order to explore more subtle task effects and differences between groups for each task, given the relatively small sample sizes. Coordinates for cluster peak activations are reported in Table 3. The location of the activation foci was determined in the stereotaxic space of MNI coordinates system with reference to cytoarchitectonical probabilistic maps of the human brain, using the SPM-Anatomy toolbox v.1.7 (Eickhoff et al., 2005). An additional analysis was carried out for picture naming task combining all stimuli from both task-demand blocks (heterogeneousþ homogenous blocks), when low-level baseline activity (i.e. fixation cross) or high-level baseline activity is removed, the latter including both perceptual and articulatory components (i.e.

141

scrambled images to which subjects were asked to covertly respond ok). Statistical methods and results are reported in the Supplementary material. 2.6.2. ANOVA models The 4 “block type  task” contrast images relative to GLM 1, the 4 homogenous set “category  task” contrasts images relative to GLM 2 for each subject were then included in two independent second level group ANOVAs to assess group differences between aMCI and control subjects. A flexible factorial design analysis was designed in SPM8 to highlight differences in activation patterns for the “block type task” two-way interaction condition (i.e. within-subjects factor) (Model 1) and the “category  task” two-way interaction condition (i.e. within-subjects factor) (Model 2) between aMCI and elderly controls (i.e. between-subjects factor) (three-way interaction). The flexible factorial model was chosen because there were multiple scans performed for each subject (i.e. repeated measures), as opposed to a single scan for each subject in the two groups (aMCI/ Controls). The flexible factorial design used a “group by condition” design to model the three-way interaction “Group  block type task” in Model 1 and “Group  category  task” in Model 2 as well as modeling global effects for each subject. The factor matrix included the main effect of “group” and the interaction “group  condition”. The main effect of “group” was added to model the general differences between aMCI and controls (i.e., age) that are not specific to the experimental conditions in order to increase the sensitivity to detect a “group  condition” interaction by removing the main effect from the interaction effect (Gläscher and Gitelman, 2008). The factor “subject” was additionally included in the design matrix in order to model the subject constants and to potentially capture any between-subject variance, however the “subject” main effect was not included in the model. This design was applied to analyze (i) whether the stimulusspecific activation (pictures or words) in the aMCI group differed from the control group and whether the differences were specific to the type of block (i.e., homogeneous or heterogeneous) in terms of semantic computational demands (Model 1); (ii) whether activation levels differed among the living and non-living stimuli between the aMCI group and the control group in the picture naming or the word reading task for homogeneous block sets with a higher degree of semantic similarity (Model 2). F-tests were performed to test whether the “group  condition” interaction effect could be observed at a more stringent threshold (p o0.05 FWE-corrected) to highlight focal regions in which group differences are highly significant. Model 2 was also assessed at a more liberal threshold (po 0.005, uncorrected at the voxel level) in order to pinpoint subtle semantic effects in highly relevant brain areas in accordance with the literature investigating the neural substrates of semantic memory impairments in AD (Zahn et al., 2004; Libon et al., 2013). Effects of interest were plotted to investigate the direction of any potential difference. 2.6.2.1. Age effect on ANOVA models. A three-step procedure was carried out in order to verify any potential confounding effect of “AGE” (i.e., aMCI older than controls, p o0.05) on the “group condition” interaction effects resulting from both Model 1 and Model 2. First, we investigated the effect of age of the whole sample (i.e., irrespective of group membership) on the interaction terms (i.e., single experimental conditions) of the “group condition” interaction in each model. Second, we computed the “age  group” interaction influence on each interaction term separately (i.e., differences between the slope which relates age in each group with the pattern of brain activity for each term). Third, we “exclusively” masked Model 1 and Model 2 results with the

142

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

“age  group” interaction contrast relative to the interaction term driving the “group  condition” interaction, in order to verify true interaction effects falling “outside” (i.e., those voxels that are not in the mask images) any activity pattern related to age-differences between the aMCI and control groups. At step 1, we used the so-called Sandwich Estimator (SwE) method (Eicker, 1963), which accounts for within-subject correlation existing in a repeated measure design, as in the case of “block type  task” two-way interaction condition (i.e., HOMHET  PNWR) and the “category  task” two-way interaction condition only in homogenous block subsets (LIVNONLIV  PNWR). The SwE method allows to include covariates (i.e., AGE) where repeated measures exist, such as fMRI designs where multiple contrasts are modeled (Guillaume et al., 2014), by splitting the covariate into between- and within-subject components and including them both in the design matrix (Neuhaus and Kalbfleisch, 1998). SwE then estimates a different covariance matrix for each voxel allowing for statistical inference on each one, as opposed to SPM, which estimates a unique covariance structure for all the voxels leading to invalid inferences in voxels where the true covariance structure is different from the one estimated (Guillaume et al., 2014). The SwE design matrices for both Model 1 and Model 2 included 1) a vector defining in which group the subjects belong and share a common covariance matrix, where the subjects were divided into two groups (i.e., 16 Controls and 8 aMCIs); 2) a vector defining the repeated measures (rm) categories (i.e., interaction terms) which in Model 1 result from the decomposition of the “block type  task” two-way interaction condition (i.e. HOMHET  PNWR) and in Model 2 from the decomposition of “category  task” two-way interaction condition only in homogenous block subsets (LIVNON-LIV  PNWR); 3) a vector defining the covariate AGE for each repeated measures category in which the age (i.e., in years) of each subject was entered. Adjusted residuals were used to compute the SwE. After model estimation, the influence of the covariate AGE on (rm)categories was assessed, independently for Model 1 and Model 2, with an F-contrast testing overall differences between age-(rm)category correlations (i.e., those areas showing a significantly different correlation between age and activity for any repeated measure category). A chi-squared-score for each voxel was computed and statistical images were thresholded at p o0.005 uncorrected at the voxel level (k45) (i.e., the more liberal p-value used to assess subtle semantic effects in Model 2). At step 2, multiple regression models were calculated in SPM8 on each interaction term including as predictors the two groups (Controls, aMCI) and the age covariate separately for each group (AGE_Controls; AGE_aMCI). A contrast testing the “age  group” interaction was computed in order to reveal activations that correlate linearly with age significantly more for the aMCI group as compared to the control group. This contrast was used to create an “exclusive mask” at step 3 only for those interaction terms, which resulted to drive the “group  condition” interaction in Model 1 (PN_HOM and PN_HET) or Model 2 (PN_LIV), in order to verify the authenticity of each interaction effect. At step 3, in order to grant the semantic genuineness of activity differences between the two groups resulting from interaction effects in the two ANOVA models, we computed for Model 1 the same F-interaction contrast testing for differences among pictures or words between aMCI and controls, masked “exclusively” by the “age  group” interaction contrast calculated at step 2, separately for the PN_HOM and PN_HET terms; for Model 2 instead, we computed the same F-interaction contrast testing for category differences among living or non-living pictures or words between aMCI and controls in homogeneous sets, masked “exclusively” by the “age  group” interaction contrast calculated at step 2 for the

Table 2 Anatomical parcellation defined by automated anatomical labeling. Regions

Color of the node in Fig. 4

Inferior frontal gyrus (p. Opercularis) Inferior frontal gyrus (p. Orbitalis) Inferior frontal gyrus (p. Triangularis) Middle orbital frontal gyrus Superior medial frontal gyrus Supplementary motor area Anterior inferior temporal gyrus Middle inferior temporal gyrus Posterior inferior temporal gyrus Anterior middle temporal gyrus Mid middle temporal gyrus Posterior middle temporal gyrus Anterior superior temporal gyrus Posterior superior temporal gyrus Anterior fusiform gyrus Middle fusiform gyrus Posterior fusiform gyrus Temporal pole Hippocampus Parahippocampal gyrus Anterior cingulate cortex Posterior cingulate cortex Supramarginal gyrus Angular gyrus Cuneus Lingual gyrus Precuneus

Blue Blue Blue Pale blue Pale blue Pale blue Green Green Green Green Green Green Green Green Red Red Red Green Green Green Pale blue Pale blue Fuchsia Fuchsia Pale blue Red Pale blue

PN_LIV. Results were assessed at a more conservative threshold (i.e., p ¼0.05 FWE-corrected at the voxel level) for Model 1 and Model 2 and lowered to a more liberal p-value (i.e., p o0.005 uncorrected at the voxel level) for Model 2 to verify the presence of more subtle semantic category effects. Anatomical localization of resultant clusters was performed using the Anatomy toolbox (version 1.7; Eickhoff et al., 2005), which provides gray matter cytoarchitectonic probabilities (where available). Macroanatomic labels with the corresponding percentages of voxels belonging to labeled regions were obtained for each significant cluster resulting from step 3 using the Anatomical Automatic Labeling (AAL) Toolbox (Tzourio-Mazover et al., 2002). Significant peak activations are reported as Montreal Neurological Institute (MNI) coordinates. 2.7. Response extraction and graph analysis The parcel-based method was chosen for graph formation. The parcel-based graph was formed using 54 non-overlapping regions of interest (ROI) (see Table 2) based on the anatomical automatic labeled brain (i.e., the 90-parcel AAL atlas, Tzourio-Mazover et al., 2002), a popular method of graph formation. This atlas divides the cortex and subcortical structures into parcels based upon anatomical landmarks. After registration, 54 non-overlapping ROIs were extracted per subject, having relative gray matter content exceeding 50%, as determined by the segmentation of the structural scan into three tissue maps (i.e., gray matter, white matter and csf). Blood oxygen level-dependent (BOLD) signal extracted from each ROI was de-spiked, de-trended with a 4th degree polynomial, low-pass filtered at f-3 dB ¼0.12 Hz and filtered by regressing out movement vectors, average white matter and cerebrospinal fluid signal (Minati et al., 2014). To calculate the coupling of brain activity between the regions defined, single time-series per subject and ROI were calculated by averaging all voxels' time-series within the ROIs. In total, 54 fMRI time-series for homogeneous and heterogeneous block sets for the two tasks (i.e., PN and WR) and for the

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

two high-level baselines specific to each task (i.e., non-pictures and non-words) consisting of 24 time-points each were available per subject. Regional time-courses were extracted by averaging over all voxels within 27 regions of interest (ROIs) for both the left and right hemisphere (i.e., 54 ROIs over the whole brain) (see Table 2 for the ROIs included in the brain network). Functional connectivity graphs were obtained computing the Pearson correlation coefficient and the p-value (permutation test, N ¼ 10,000) for each ROI time course. Correlation values greater than a set of thresholds (taking values in {0.4, 0.5, 0.6}) that were statistically significant (p o0.05) constituted the 1's of three adjacency matrices, one for each threshold value. For the analysis of these graphs, we introduced a set of common statistics from the Complex Network Theory which generally characterize functional brain networks in similar studies: the clustering coefficient (C) and the characteristic path length (L) and the modularity of the community structures (Q). A Matlab toolbox (Brain Connectivity Toolbox, BCT), was used for the network analyses for all three measures (Rubinov and Sporns, 2010). Estimations of each statistics were computed for the three adjacency matrices, respectively obtained by the different correlation thresholds, and averaged. In complex network theory, several graph measures take specific meaning only if they are compared to the same graphs subject to randomization or latticization (often called null networks). Both procedures guarantee that the node degree distributions of the original graphs were preserved. The randomized version of our graphs were computed in order to estimate Cr and Lr. These null network values are required, for instance, to verify the small-world nature of the graphs. If small-worldness (S) (i.e., (C/Cr)/(L/Lr)) is 41, the graph can be considered a small-world network (Rubinov and Sporns, 2010). We further investigated the community structure of our graphs. Communities in this context refer to the aggregated functional units under investigation and are analyzed in order to assess how node graphs are aggregated in each experimental condition by estimating the modularity index Q (Newman, 2006). Graphs in which less than the 99% of nodes were connected or that were outliers (beyond the 5th and 95th percentiles) of the node, edge and density number distributions were discarded in order to obtain a better homogeneity and homoscedasticity.

3. Results 3.1. Behavioral results The mean percentage of correct responses, collected outside the scanner, was high, 99% of correct responses. A 2  2  2 repeated measures ANOVA was carried out with group (controls vs aMCI) as a between factor and tasks (naming and reading) and condition (homogeneous and heterogeneous) as within factors. Only the main effect of task was significant (F¼ 13.494; p o 0.005), with naming (98.6%) less accurate than reading (99.9%). When age was considered as a covariate, no significant main (F¼0.242; p ¼0.628) and interaction effects were evident. Considering only homogeneous blocks, a 2  2  2 ANOVA was carried out with group as between factor and task (naming and reading) and category (living and non-living) as within factors. Task (naming less accurate than reading; F¼ 13.616; p o0.005), category (living less than non-living; F ¼9.348; p o0.01) and the interaction between task and category (with living better in reading; F¼8.494; p o0.01) were significant (p o0.05); however, also in this case when age was considered as a covariate the significant effect

143

disappeared (main effect of task F¼0.247; p ¼0.625, category F¼1.247; p ¼0.277; interaction between task and category F¼1.605; p ¼0.219). Concerning the reaction times, 7% of data were discarded (5% for registrations problems and 2% were outliers). A 2  2  2 Repeated measures ANOVA was carried out with group (controls vs aMCI) as a between factor and tasks (naming and reading) and condition (homogeneous and heterogeneous) as within factors. Only the main effect of task was significant (F¼84.272; p o0.001), with naming slower than reading. When age was considered as a covariate, no significant main (F ¼0.032; p ¼0.859) and interaction effect were evident. In addition, considering only homogeneous blocks, a 2  2  2 ANOVA was carried out with group as between factor and task (naming and reading) and category (living and non-living) as within factors. Task (naming slower than reading; F¼76.951; p o0.001) and the interaction between task and category (with living faster in reading; F¼ 13.843; p o0.005) were significant; also in this case when age was considered as a covariate the significant effect disappeared (main effect of task: F¼0.001; p ¼0.98; interaction between task and category: F¼0.732; p ¼0.402). 3.2. fMRI results 3.2.1. PN and WR within each group and between groups In order to allow a comparison with other naming studies, which included items from different categories, the analysis was conducted only on the heterogeneous blocks compared to the high level baseline; for the analysis carried out combining all stimuli from both task-demand blocks (heterogeneous þhomogenous) compared to either the low-level baseline (i.e., fixation cross) or high-level baseline, please see the Supplementary material. In the Control group, the PN task elicited brain activity involving the fusiform gyrus extensively, the inferior frontal gyrus (pars opercularis and orbitalis) and the SMA in the left hemisphere, while in the right hemisphere activity peaked in the posterior cingulate cortex, thalamus and the inferior frontal gyrus (pars triangularis). WR yielded significant activity mainly in the lingual gyrus, inferior parietal lobule (IPL), postcentral gyrus, precentral gyrus and the inferior frontal gyrus for the left hemisphere, and the cerebellum (lobule VI), calcarine gyrus, precentral gyrus, rolandic operculum, insula and SMA for the right hemisphere (see Table 3). In the aMCI group, there was increased activity for PN extending from the superior parietal lobule (SPL), postcentral gyrus, to the precentral gyrus and inferior frontal gyrus in the left hemisphere, while in the right hemisphere significant activation was detected in the lingual gyrus, the thalamus and precentral gyrus (see Table 3). For WR, the activation involved the left SPL, precentral gyrus and the middle temporal gyrus. The direct comparison between PN and WR within the control group yielded a pattern of increased activity for PN relative to WR in the left middle occipital gyrus extending to the left fusiform gyrus, left SPL as well as activity in the right fusiform gyrus, the right angular gyrus, right inferior frontal gyrus (p. triangularis) and cerebellar vermis. For the aMCI group the direct comparison between the PN and WR yielded increased activity solely for PN relative to WR in left inferior occipital gyrus and in a more distributed network of areas in the right hemisphere, which included the lingual gyrus, the inferior temporal gyrus, the thalamus and the precentral gyrus. When making a direct comparison of brain activity between groups, increased activity for PN was observed for the aMCI group relative to the Control group in the left postcentral gyrus, inferior parietal lobule and supramarginal gyrus and in the right SPL, postcentral and precentral gyri and Heschl gyrus (see Table 3) for

144

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

Table 3 Brain areas active in the contrasts between the heterogeneous blocks of each task (picture naming and reading task) minus the corresponding baseline (respectively non-pictures and non-words) and differences between the two tasks (picture naming vs reading and vice versa) for each group (controls and aMCI), and group differences (i.e., controls vs aMCI and vice versa) for each task. The voxelwise statistical threshold was set at p o 0.001 uncorrected at the voxel level for all comparisons (k 410 voxels). Clusters surviving an FWE-corrected extent threshold appear as marked with an asterisks (n). Regions

Cluster extent

CONTROLS PICTURE NAMING – BASELINE Left Hemisphere Fusiform Gyrus 5417n Inferiori Frontal Gyrus (p. 399n Opercularis) 158n Inferior Frontal Gyrus (p. Orbitalis) SMA 102n Right Hemisphere Posterior Cingulate Cortex Thalamus Inferior Frontal Gyrus (p. Triangularis)

159n 75n 478n

WORD READING – BASELINE Left Hemisphere Lingual Gyrus 133n Inferior Parietal Lobule 12 Postcentral Gyrus 10 Precentral Gyrus 25 Inferiori Frontal Gyrus (p. 19 Opercularis) Right Hemisphere Cerebellum (Lobule VI) Calcarine Gyrus Precentral Gyrus Rolandic Operculum Insula SMA PICTURE NAMING – WORD Left Hemisphere Middle Occipital Gyrus Fusiform Gyrus Superior Parietal Lobule Right Hemisphere Fusiform Gyrus Angular Gyrus Inferior Frontal Gyrus (p. Triangularis) Cerebellar Vermis

Coordinates x

y

z

6.93 4.97

 38  38

 53  18 10 28

4.15

 45

18

3

4.64

3

15

50

4.34 4.08 4.72

3 25 40

 38 8  30 10 25 10

4.29 3.52 3.59 3.80 3.64

3  35  38  53  50

 75  60  33 0 13

5 50 50 40 8

4.06 3.54 3.69 3.44 3.47 4.05

20 20 50 53 45 5

 68  60 3 8 13 15

 25 8 25 15 3 53

47 779n 99n

3.72 6.18 4.05

 33  38  20

 80 18  53  18  73 43

461n 300n 25

5.56 4.45 4.36

40 35 40

 58  20  60 35 25 10

13

3.37

0

 43

4.78 4.30 4.27 4.10

 33  25  48  35

 60 58  18 40  10 48 40  10

4.58 4.86 4.45

18 10 43

 85 18  20 10  15 58

66n 14 17 14 13 47 READING

aMCI PICTURE NAMING – BASELINE Left Hemisphere Superior Parietal Lobule 360n Precentral Gyrus 110n Postcentral Gyrus 359n Inferiori Frontal Gyrus (p. 220n Orbitalis) Right Hemisphere Lingual Gyrus Thalamus Precentral Gyrus

Voxel level Z-score

911n 462n 74n

 35

WORD READING – BASELINE Left Hemisphere Superior Parietal Lobule 21

3.91

 13

 68 40

Right Hemisphere Middle Temporal Gyrus Precentral Gyrus

3.90 3.87

60 40

 20  8 5 43

32 16

PICTURE NAMING – WORD READING Left Hemisphere Inferior Occipital Gyrus 12

3.37

 45

 63

 15

Table 3 (continued ) Regions

Right Hemisphere Lingual Gyrus Inferior Temporal Gyrus Thalamus Precentral Gyrus

Cluster extent

Voxel level Z-score

Coordinates x

y

z

4.04 4.23 4.06 3.95

20 40 10 43

 88  8  60  10  20 10  15 58

PICTURE NAMING: aMCI-CONTROLS Left Hemisphere Postcentral Gyrus 91n Inferior Parietal Lobule 56 Supramarginal Gyrus 70n

4.18 4.02 3.64

 15  33  55

 35 73  33 45  28 35

Right Hemisphere Superior Parietal Lobule Postcentral Gyrus Precentral Gyrus Heschl Gyrus

4.29 4.15 3.84 3.91

28 18 40 38

 50  35  25  28

24 40 31 10

37 50 48 11

WORD READING: aMCI – CONTROLS Right Hemisphere Precuneus 25 Precentral Gyrus 22

3.90 3.80

15 33

 63  13

65 68 55 10

48 45

naming pictures in heterogeneous sets. For WR, significant activation increases were evident for the aMCI group as compared to the control group only in the right hemisphere, involving the precuneus and precentral gyrus. 3.2.2. Results ANOVA 1: homogeneous and heterogeneous sets The F-interaction contrast testing for differences among pictures or words between aMCI and controls (i.e., considered at po 0.05 FWE corrected at the voxel level) revealed one significant activation difference, located in the left fusiform gyrus (L-FG) for both the homogenous and the heterogeneous sets, with a greater activation extent for the latter set, (see Fig. 2). No significant activation pattern emerged for word stimuli. Plots of “group  condition” parameter estimates revealed that this difference in L-FG was driven by picture stimuli and characterized by a significant higher activation level in this region for the homogeneous set as compared to the heterogeneous one in the aMCI group, while an inverse (but not significant) trend with lower activation levels for the homogeneous set was instead evident in the control group (see Fig. 2). 3.2.3. ANOVA 2: living and non-living in homogeneous sets The F-interaction contrast testing for category differences among living or non-living pictures or words between aMCI and controls in homogeneous sets (i.e., considered at p o0.05 FWE corrected at the voxel level) revealed no significant activation difference. When the statistical threshold was lowered to p o0.005 uncorrected at the voxel level for comparison with the existing literature on semantic category effects in terms of brain activations, category-related activation differences were found in the left temporal pole, the left supramarginal gyrus and the left lingual/ fusiform gyrus (see Fig. 3). All these regions have been linked to semantic processing (Binder et al., 2009; Visser et al., 2010). Plots of “group  category” parameter estimates for pictures and words revealed that this difference was related to a greater activation for living as compared to non-living items in all three regions solely for pictures and exclusively in the aMCI group. 3.2.4. Age effect on ANOVA models The F-contrast testing overall differences between the slopes relating age of the whole sample (i.e., controlsþ patients) and

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

145

Fig. 2. “Block type  Task” brain activations. Columns 1–4 refer to the control group and report the parameter estimates for: 1) pictures HOM – non-pictures HOM, 2) pictures HET – non-pictures HET, 3) words HOM – non-words HOM, 4) words HET – non-words HET. Columns 5–8 refer to the same conditions in the aMCI group. Asterisks show significant differences.

brain activity for each level (i.e., HOM/HET x PN/WR) of the “block type  task” within-subject interaction (i.e., Model 1) or each level (LIV/NON-LIV x PN/WR) of “category  task” with-in subject interaction, revealed for Model 1 one single cluster of difference located in the superior parietal lobule (see Table 4), while for Model 2 age-related activation differences mainly involved areas

in the right hemisphere such as the middle frontal gyrus, the inferior frontal gyrus (pars orbitalis), the angular gyrus, the superior temporal gyrus, and the thalamus. Age differences were also evident in the left hemisphere in the parahippocampal gyrus, left insula and the middle cingulate cortex bordering the midline. The same F-“block type  task” interaction contrast computed

Fig. 3. “Category  Task” brain activations. Columns 1–4 refer to the control group and report the parameter estimates for: 1) pictures LIV HOM – non-pictures LIV HOM, 2) pictures NONLIV HOM – non-pictures NONLIV HOM, 3) words LIV HOM – non-words LIV HOM, 4) words NONLIV HOM – non-words NONLIV HOM. Columns 5–8 refer to the same conditions in the aMCI group. Asterisks show significant differences.

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

146

Table 4 Age-related activations for ANOVA MODEL 1 and MODEL 2. Regions

EFFECTS OF AGE: MODEL 1 Left Hemisphere Superior Parietal Lobule MODEL 2 Left Hemisphere ParaHippocampal Gyrus Insula Posterior Medial Frontal (middle cingulum) Right Hemisphere Angular Gyrus Thalamus Superior Temporal Gyrus Middle Frontal Gyrus Inferior Frontal Gyrus (pars orbitalis, BA 44/45)

Cluster extent

Voxel level χ² value

Coordinates x

y

z

Table 5 Areas of difference resulting from the F-interaction contrast testing for differences among pictures or words between aMCI and controls, masked “exclusively” by the “age  group” interaction contrast calculated for the PN Homogenous condition and PN Heterogeneous, where a “group  interaction” condition arose as shown in Fig. 2. Regions

14

14.90

 33

 65

53

7 5

11.76 11.46

 18  35

 35 3

 10 8

37

18.33

3

0

48

14 14 7 20 16

11.48 12.71 9.85 8.94 16.13

43 10 48 23 53

 60  28  20 15 21

53 1 5 45 3

p o0.005 uncorrected; k 45.

for Model 1 at step 3 testing for differences among pictures or words between aMCI and controls, masked “exclusively” by the “age  group” interaction contrast calculated at step 2 for the PN_HOM term verified a genuine interaction effect in the fusiform gyrus in two separate small clusters including portions of the inferior occipital and temporal gyrus and extending to middle occipital gyrus and to cerebellum (see Table 5 for cluster label and percentage details). The F-“block type  task” interaction contrast masked “exclusively” by the “age  group” interaction contrast calculated for PN_HET term validated the effect in a large cluster with the highest maxima in the left fusiform gyrus extending though to inferior and middle portions of both occipital and temporal gyri as well as cerebellum (see Table 5 for cluster label and percentage details). The same F-interaction contrast computed for Model 2 at step 3 testing for category differences among living or non-living pictures or words between aMCI and controls in homogeneous sets, masked “exclusively” by the “age  group” interaction contrast calculated at step 2 for the PN_LIV, confirmed the genuineness of the category-related activation peak differences between groups in the left temporal pole, the left supramarginal gyrus and the left lingual gyrus/fusiform gyrus (see Table 6 for cluster label and percentage details). 3.3. Graph analysis The human brain can be seen as a giant complex network from a dual point of view: structural connections among neurons return a stable picture of brain circuits, while neural interactions (functional connections) show a dynamical representation which enables a vast repertoire of transient behaviors. From a functional perspective, coarse-grained measurements of two crucial brain network modalities are traceable by clustering coefficient (C), which is an estimator of the network ability to segregate information (functional segregation), and on the characteristic path length (L), which represents the network capability to integrate information (functional integration). These metrics were assessed for 1) naming pictures or reading words in homogeneous or heterogeneous blocks 2) responding to non-pictures or non-words, in order to assess the degree of network segregation and integration in the two groups in response to different degrees of semantic load (heterogeneous or homogenous blocks) or when no semantic

Cluster extent

% of entire volume

Voxel level Z

Coordinates x

y

z

HOMOGENEOUS CONDITION EXCLUSIVELY MASKED BY EFFECTS OF AGE Left Hemisphere Inferior Occipital 1 60.31 5.39  40  73  10 Gyrus -Fusiform 23.74 - Middle Occipital 8.56 - Inferior 3.11 Temporal Fusiform Gyrus 2 43.58 4.76  40  65  13 -Occipital Inferior 34.24 -Temporal Inferior 10.51 -Middle Occipital 4.28 Fusiform Gyrus 3 61.87 4.82  40  60  15 -Inferior Occipital 15.56 -Inferior Temporal 14.01 - Cerebellum 4.28 HETEROGENEOUS CONDITION EXCLUSIVELY MASKED BY EFFECTS OF AGE Left Hemisphere Fusiform Gyrus 79 43.58 6.19  40  65  13 -Inferior Occipital 34.24 -Inferior Temporal 10.51 -Middle Temporal 4.28 -Middle Occipital 2.72 -Cerebellum 2.72 Inferior Occipital 62.26 5.52  40  75 8 Gyrus -Middle occipital 23.74 -Fusiform 9.34 Fusiform Gyrus 57.20 5.31  38  48  23 -Cerebellum 24.12 -Inferior Temporal 14.01 Labels and percentage of the labelized region belonging to the cluster were obtained for each significant cluster using the automated anatomical labeling (AAL) Toolbox; p o 0.05 FWE; k ¼ 0.

processing gradient is elicited as in the case of non-pictures or non-words. Non-parametric Wilcoxon rank-sum tests were used for all comparisons. There were no significant group differences for picture naming in heterogeneous blocks. In contrast, for homogenous blocks there was a significant difference in terms of network integration (L) only for the picture naming task. No significant group differences were evident for network activity elicited by blocks of non-pictures or non-words either in terms of segregation and integration (see Table 7). Modularity analyses delineate the community structure in complex networks, which can be considered as a complementary coarse-grained measure of the functional segregation. The overall strength of the community organization for the whole brain was investigated by comparing modularity (Q) values for the same conditions between the aMCI and Control group. Overall average modularity differed between aMCI and controls only for network activity elicited in response to homogeneous blocks of pictures, with aMCI showing a greater Q, while no significant difference was evident for the other conditions (see Table 7). The observation that a significant divergence emerged only with the modularity index implies that the greater ability of aMCI brain networks to segregate information was detectable only by inspecting the networks through large groups of nodes instead of small groups (as C was normal), indicating an over-segregated condition, where networks are organized in more numerous and more isolated modules (as L

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

Table 6 Areas of difference resulting from F-interaction contrast testing for category differences among living or non-living pictures or words between aMCI and controls in homogeneous sets, masked “exclusively” by the “age  group” interaction contrast calculated at step 2 for the LIVING category in the picture naming task, where a “group  interaction” condition arose as shown in Fig. 3. LIVING VS NON-LIVING EXCLUSIVELY MASKED BY EFFECT OF AGE Regions

Left Hemisphere Lingual Gyrus -Cerebellum -Fusiform Supramarginal Gyrus -Inferior Parietal -Postcentral Superior Temporal Pole -Middle Temporal Pole

Cluster extent

% of entire volume

Voxel le- Coordinates vel Z x y z

4

2.78

 18

 70

75

78.60 10.89 7.00 30.35

3.71

 45

 30 30

8

21.40 10.12 68.09

3.33

 50 13

8

 20

15.95

Labels and percentage of the labelized region belonging to the cluster were obtained for each significant cluster using the automated anatomical labeling (AAL) Toolbox; po 0.005 uncorrected; k ¼3.

was greater). In particular, in Fig. 4, where differences between average adjacency matrices of fMRI correlation networks in aMCI group and in controls (below) during the homogeneous condition are shown, a greater functional connectivity is clearly visible between temporal areas in aMCI than controls (yellow edges) during the homogeneous condition. At the same time, a generally decreased functional connectivity between distant modules in aMCI vs controls was also observed (fuchsia edges).

4. Discussion The main focus of this study was to investigate the presence of functional changes in the brain networks underpinning semantic memory processing, when, as in the case of aMCI patients, subclinical alterations may already be induced by pathology, although no behavioral evidence is still clinically present. To address this issue, we carried out an fMRI study in which both aMCI and control subjects were required to covertly name pictures and read words in situations where the lexical/semantic demands were experimentally manipulated, using a blocking paradigm, to be lower (heterogeneous condition) or higher (homogenous condition), and considering the possible impact of different semantic categories (living vs non-living items). The main result is that early functional changes of both brain activation and functional connectivity can be found in aMCI subjects engaged in a picture naming task in which they show an unimpaired performance. The comparison between the heterogeneous conditions showed an additional involvement of bilateral parietal and right frontal regions during naming task in aMCI compared to controls. In contrast, a regular word reading task, which is associated with low semantic processing demands (Woollams et al., 2007), was associated only with a localized difference in right parietal activity. Functional changes were more evident in the homogeneous than in the heterogeneous condition, only in the case of the picture naming task. Consistently, the results obtained with network analyses highlighted that aMCI subjects exhibit topological changes, only during the homogeneous naming task, without changes during the reading task.

147

It must again be underlined that all these activation changes in aMCI subjects were observed in the absence of a corresponding behavioral effect, showing that, despite the increased neuronal activity, there was no processing cost involved in terms of naming latencies. The functional changes reported here are in line with an extensive literature about brain activation changes during functional imaging in normal aging (Reuter-Lorenz, 2013). A common finding is that older adults recruit more resources at lower levels of task demand in comparison to younger controls (for example, in the case of working memory: Cappell et al., 2010). It has been proposed that these functional changes represent compensatory recruitment, allowing the aging brain to adapt to structural alterations associated with the normal ageing process (Barulli and Stern, 2013). Elderly subjects showed increased frontal activation during a naming task in comparison to younger subjects (Wierenga et al., 2008). Most of the functional imaging studies in MCI subjects aimed at the investigation of episodic memory. Imaging during a semantic memory task showed increased frontal activation (Gigi et al., 2010). At-risk normal adults (APOE ε4) engaged in a photograph naming task demonstrated a more widespread brain response, with increased activity in the left fusiform gyrus, bilateral medial prefrontal cortex, and right perisylvian cortex (including inferior parietal lobe), compared to a control group (Wierenga et al., 2010). A greater activity during naming in the heterogeneous condition was found in the left parietal regions, in particular in the left supramarginal gyrus, in aMCI with respect to controls. Supramarginal gyrus is markedly affected in AD (Harasty et al., 1999). Given its probable role in phonological processing (Buchsbaum et al., 2011; Romero et al., 2006), this region does not have an explicit role in semantic processing theories. Its activation, however, is frequently reported in imaging studies employing semantic memory tasks (Binder et al., 2009). In addition it is frequently associated with a semantic memory deficit in AD (Zahn et al., 2004; Libon et al., 2013). The central role in lexical-semantic processing of the neighboring angular gyrus region (Luria, 1947; Demonet et al., 1992) raises the issue of a possible increased recruitment of adjacent regions. The interpretation of the other activation difference between aMCI vs controls in sensorimotor areas, i.e. the right pre and postcentral gyri is also difficult. While increased articulation during the covert task is unlikely, because of the presence of a high level baseline and of microphone control during scanning, an increased retrieval effort may be a possible underlying mechanism. Our study allows further insight into the observed pattern of increased activity, as observed when the lexical/semantic demands were increased by the homogeneous condition. Vigliocco et al. (2002) have demonstrated that more closely related items generate stronger interference effects than those more distant. Semantic context effect has been generally attributed to an activation of shared conceptual features among categorically similar objects, leading to co-activated lexical candidates and consequently slowing target selection within the homogeneous context. The effect has been proposed to arise at the conceptual level, at the interface of the conceptual and the lexical level or a the lexical level (Belke, 2013; Kroll and Stewart, 1994; Damian and Als, 2005; Damian et al., 2001). It must be underlined that, while naming latencies in the homogeneous condition, when compared to blocks in the heterogeneous condition, are generally longer (Damian et al., 2001; Levelt et al., 1999; Kroll and Stewart, 1994), in the present study the functional alterations do not appear to be due to the overall difficulty of the task on the basis of the behavioral data, as the experimental conditions were matched for speed and accuracy. The manipulation of the semantic context indicated that aMCI in order to select the appropriate name in the context of blocks of

148

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

Fig. 4. The brain networks relative to the parcel-based graph consisting of 54 non-overlapping regions of interest (ROI) representing binarized adjacency matrices of fMRI correlation networks in aMCI (above), Controls (center) and differences between aMCI group and controls (below, where yellow edges represent greater connections for aMCI vs controls, vice versa, fuchsia edges indicate greater connections for controls than aMCI) relative to the homogeneous condition were visualized with the BrainNet Viewer (http://www.nitrc.org/projects/bnv/) (Xia et al., 2013). All graphs have been binarized with a threshold set to 0.5. Nodes representing anatomical landmarks for each region were grouped on the basis of two criteria anatomical proximity and relevance for semantic processing. Node sizes were varied according to significant areas resulting from Model 1 and Model 2 (i.e., L-FG, L-SMG, L-TP). Green nodes include Posterior inferior temporal gyrus, Anterior middle temporal gyrus, Mid middle temporal gyrus, Posterior middle temporal gyrus, Anterior superior temporal gyrus, Posterior superior temporal gyrus, Temporal pole, Hippocampus, Parahippocampal gyrus; red nodes include Anterior fusiform gyrus, Middle fusiform gyrus, Posterior fusiform gyrus and the lingual gyrus; blue nodes include Inferior frontal gyrus (p. Opercularis, p. Orbitalis and p. Triangularis); pale blue include Middle orbital frontal gyrus, Superior medial frontal gyrus, Supplementary motor area, Anterior cingulate cortex, Posterior cingulate cortex and precuneus; fuchsia nodes include Supramarginal gyrus and Angular gyrus. L¼ left; R¼ right.

group. A compatible result was reported by Lenzi et al. (2011). Using a semantic judgment task, they found an increase in the activation of the left inferior temporal gyrus in aMCI subjects compared to normal controls, in association with preserved

pictures closely related to each other (homogeneous condition) rely on a higher degree of brain activity in the left fusiform gyrus (lateral portion), while an inverse trend with lower activation levels for the homogeneous set was instead evident in the control

Table 7 Complex network statistics and statistical comparisons among the groups and tasks. A significant difference between Controls and aMCI is only observable in the picture naming homogeneous task. Specifically, aMCI reported an increased coarse-grain segregation (Q) and a reduced integration capacity. TASK

Conditions

Segregation

Integration

C

PN WR PN WR BASELINE BASELINE

Heterogeneous Heterogeneous Homogeneous Homogeneous Non-Pictures Non-Words

Q

L

CONTROLS

aMCI

p-Value

CONTROLS

aMCI

p-Value

CONTROLS

aMCI

p-Value

0.63 0.64 0.65 0.65 0.62 0.62

0.63 0.63 0.62 0.62 0.61 0.61

0.9216 0.92 0.5047 0.5047 0.7678 0.8941

0.28 0.28 0.26 0.26 0.33 0.33

0.3 0.3 0.31 0.32 0.36 0.37

0.2475 0.4772 0.0428 0.09 0.3998 0.3877

1.9 1.9 1.86 1.86 2.1 2.11

1.92 1.92 1.99 1.97 2.25 2.27

0.8014 0.8 0.0378 0.3411 0.3526 0.2978

C¼ clustering coefficient; Q ¼ modularity; L ¼ characteristic path length; PN ¼ Picture Naming; WR ¼ Word Reading. p-Value computed using the non-parametric Wilcoxon rank-sum test.

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

behavioral performance. Given the imperfect age-matching between the groups, particular care was taken to exclude a possible confound of the results by the age factor. While modest age related effects were observed, in agreement with the available evidence (Grady et al., 2006), the exclusive masking procedure indicated that they were not responsible for the observed semantic effect. The increased activity of the left fusiform gyrus may be related to the computation of visual/semantic representations of the items. When aMCI are asked to name a picture in a context where the close presentation of a semantically related word has occurred, a greater cognitive effort will translate in greater activity in this area which will be tailored at discriminating between viable semantic candidates. The fusiform gyrus is strongly activated by tasks requiring participants to discriminate an object from many similar visual or semantic competitors (Joseph and Gathers, 2003; Price et al., 2003). Rogers et al. (2005), using a category verification paradigm, showed an activation of the lateral fusiform gyrus when the subjects had to classify at a more specific level (e.g., Labrador or BMW) with respect of an intermediate (or basic, e.g., dog or car) level of categorization, suggesting a role in a finer-grained discrimination required to differentiate between visually or semantically similar objects. Other studies (de Zubicaray et al., 2014) have hypothesized that the peririnhal cortex may be the candidate region in the case in which the interference effect originates at conceptual feature level (e.g., Belke, 2013) or in feature-to-lexical connections (e.g., Damian and Als, 2005; Oppenheim et al., 2010). Both lesion and neuroimaging evidence from picture naming tasks implicates this region in processing feature overlap among objects (e.g., Bussey et al., 2005; Chan et al. 2011; Clarke et al., 2013; Hocking et al., 2009; Moss et al., 2005; Tyler et al., 2004). Hocking et al. (2009) reported a greater involvement of the peririnhal cortex in naming visually similar items in homogeneous vs heterogeneous blocks. Recently, Tyler et al. (2013) proposed different roles for the peririnhal cortex and the fusiform gyrus. While the differentiation between highly similar objects, enabling object-specific representations, was associated with bilateral peririnhal activity, the activity of the lateral portion of the fusiform gyrus is driven by similarity of conceptual structure reflecting category structure (i.e., the relative amount of shared features within a concept, enabling for example matching task). Our experimental paradigm does not allow to draw a clear distinction between the different roles of the fusiform gyrus and perirhinal cortex. However, the increased activity in aMCI in the lateral fusiform gyrus for the homogeneous condition, without additional activities in the perirhinal cortex, confirms the role of the fusiform gyrus in elaborating pictures sharing several features. It may also support a role in the discrimination of similar pictures, as also suggested by previous studies (Rogers et al., 2005; Joseph and Gathers, 2003; Price et al., 2003). These results are in line with the main hypothesis of this study, i.e., that the homogeneous condition could be particularly apt to show subtle changes in brain activity in aMCI subjects. A detailed analysis at the functional network level also highlighted signs of functional connectivity alterations only in the case of the homogeneous condition in picture naming. If a fine-grained measure of segregation (clustering coefficient, C) did not result in statistical differences, a coarsegrained statistics, the modularity, showed increments of the putative numbers of the functional modules constituting the entire set of ROIs, suggesting the presence of a over-partitioned condition associated with a reduced ability to integrate information previously computed in other brain areas. Temporal areas, including the fusiform gyrus, were more connected with each other in aMCI than controls during the homogeneous condition. At the same time, a generally decreased functional connectivity between distant modules in aMCI vs controls was also observed.

149

The few graph theoretical studies conducted on aMCI revealed alterations of brain network organization, but with conflicting results (Dai and He 2014). Increased path length was reported during resting state in an fMRI study (Wang et al., 2013), while a decreased path length and a lower modularity were reported in a MEG study during a memory task (Sternberg's letter-probe task) (Buldú et al., 2011). Differences in the ability to integrate information have been attributed to cognitive load, i.e. respectively resting state, “where fewer connections may reflect essential disconnections of spontaneous neural activity”, and memory task, where “more connections may be required to achieve the same level of cognitive output” (Wang et al., 2013). In particular, the results reported by Buldú et al. (2011), i.e. decreased path length and a lower modularity, stand in contrast to our present findings. This discrepancy may be attributed to multiple factors, i.e. differences in methodological approaches, in task employed, and/or in severity of cognitive impairment. In particular, the role of task is probably major, as different compensation mechanisms may be engaged according to the main sites of pathological involvement. In addition, while we included subjects with amnestic MCI, in Buldú et al. (2011) the subjects had multi-domain MCI. Modularity is decreased in AD (de Haan et al., 2012), and decreasing modularity was observed with increasing cognitive impairment (Brier et al., 2014; Sun et al., 2014). In a recent study, cognitively preserved patients with Multiple Sclerosis (Gamboa et al., 2014), during resting state, showed increased modularity, suggesting “that brain adaptive mechanisms operate by the strengthening of intra-modular functional connectivity, which may compensate for the decrease of inter-modular connectivity”. Consistently, a recent work claimed that networks gain more modules in response to an increasing adaptability demand (Clune et al., 2013). An additional factor we considered is the role of semantic category. No areas survived to correction for multiple comparisons. With the aim of detecting possible small category-related differences (Tyler et al., 2003), for comparison to the existing literature investigating the neural substrates of semantic memory impairments in AD (Zahn et al., 2004; Libon et al., 2013) we investigated the pattern of activation at a more lenient threshold (0.005 uncorrected). This of course increases the chance of false positives, and the results should be interpreted with caution, exclusively in relation to previous findings reporting differences between semantic categories in the same brain regions. At this lower threshold, aMCI showed a greater activation for living vs non-living in left lingual/fusiform gyrus (in its medial portion), in the supramarginal gyrus and in the temporal pole in the picture naming task, in comparison with controls. Processing living items in the homogeneous condition may require a greater cognitive effort in aMCI, involving neighboring temporo-occipital areas. Many functional neuroimaging studies in healthy adults have indeed supported an opposite specialization of the different portions of the fusiform gyrus for processing category knowledge, with animals eliciting more activation than tools in the lateral portion, while tools involve the medial regions (e.g., Chao et al., 1999). The anterolateral aspect of the temporal lobe is generally considered as an amodal hub, contributing to semantic processing for all kinds of concepts and for all modalities of reception and expression (Patterson et al., 2007). Rogers et al. (2006) reported that ATL responds when the task requires differentiation of items with similar representations. A greater recruitment of the left ventral temporal areas (BA 19 and 37) was observed in AD relative to healthy subjects during an unimpaired pleasantness judgment task for living with respect to non-living items (Grossman et al., 2003). A previous PET study in AD patients showed an association between impaired knowledge of the visual properties of living concepts and hypometabolism in the posterior fusiform gyrus (Zahn et al., 2006). In line with the proposal of Garrard et al. (1998), it is

150

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

plausible that category specific effect in AD may be due to differences in the regional distribution of the underlying pathology. The featural representation of non-living items may be more distributed, involving also fronto-parietal areas not affected in the early stages. This mechanism may result in less impairment in AD, and in a less demanding condition in aMCI. In conclusion, the present findings suggest that functional activation with a simple naming task can be a sensitive tool to detect subclinical changes in brain activity in subjects at risk for progression to AD. These changes, in particular those reported in the more demanding conditions, involve regions which are part of the lexical-semantic network, known to be involved in semantic memory impairment in AD, and may reflect compensation to incipient neuropathological involvement.

Acknowledgments We thank Silvia Gilardoni for assistance with behavioral data collection in healthy subjects; Elisa Scola and Monia Cabino for their assistance with fMRI acquisition; Ludovico Minati for methodological advices and Giada Caramatti, Paola Frasson, Valeria Golzi, Alessandra Marcone, Valentina Plebani and Michele Zamboni for assistance with neurological and neuropsychological data collection.

Appendix A. Supplementary material Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.neuropsychologia. 2015.01.009.

References Adlam, A.L., Bozeat, S., Arnold, R., Watson, P., Hodges, J.R., 2006. Semantic knowledge in mild cognitive impairment and mild Alzheimer's disease. Cortex 42 (5), 675–684. Ahmed, S., Arnold, R., Thompson, S.A., Graham, K.S., Hodges, J.R., 2008. Naming of objects, faces and buildings in mild cognitive impairment. Cortex 44 (6), 746–752. Alathari, L., Trinh Ngo, C., Dopkins, S., 2004. Loss of distinctive features and a broader pattern of priming in Alzheimer's disease. Neuropsychology 18, 603–612. Anchisi, D., Borroni, B., Franceschi, M., et al., 2005. Heterogeneity of brain glucose metabolism in mild cognitive impairment and clinical progression to Alzheimer disease. Arch. Neurol. 62 (11), 1728–1733. Apostolova, L.G., Lu, P., Rogers, S., Dutton, R.A., Hayashi, K.M., Toga, A.W., Cummings, J.L., Thompson, P.M., 2008. 3D mapping of language networks in clinical and preclinical Alzheimer's disease. Brain Lang. 104, 33–41. Ashburner, J., 2007. A fast diffeomorphic image registration algorithm. Neuroimage 38 (1), 95–113. Balthazar, M.L.F., Cendes, F., Damasceno, B.P., 2008. Semantic error patterns on the Boston Naming Test in normal aging, amnestic mild cognitive impairment, and mild Alzheimer’s disease: is there semantic disruption? Neuropsychology 22 (6), 703. Barca, L., Burani, C., Arduino, L.S., 2002. Word naming times and psycholinguistic norms for Italian nouns. Behav. Res. Methods Instrum. Comput. 34, 424–434. Barulli, D., Stern, Y., 2013. Efficiency, capacity, compensation, maintenance, plasticity: emerging concepts in cognitive reserve. Trends Cogn. Sci. 17 (10), 502–509. Belke, E., 2013. Long-lasting inhibitory semantic context effects on object naming are necessarily conceptually mediated: implications for models of lexical-semantic encoding. J. Memory Lang. 69, 228–256. Binder, J.R., Desai, R.H., Graves, W.W., Conant, L.L., 2009. Where is the semantic system? A critical review and meta-analysis of 120 functional neuroimaging studies. Cereb. Cortex 19, 2767–2796. Bokde, A.L., Lopez-Bayo, P., Meindl, T., Pechler, S., Born, C., Faltraco, F., et al., 2006. Functional connectivity of the fusiform gyrus during a face-matching task in subjects with mild cognitive impairment. Brain 129, 1113–1124. Brier, M.R., Thomas, J.B., Fagan, A.M., Hassenstab, J., Holtzman, D.M., Benzinger, T.L., Ances, B.M., 2014. Functional connectivity and graph theory in preclinical Alzheimer's disease. Neurobiol. Aging 35 (4), 757–768.

Buldú, J.M., Bajo, R., Maestú, F., Castellanos, N., Leyva, I., Gil, P., Boccaletti, S., 2011. Reorganization of functional networks in mild cognitive impairment. PLoS One 6 (5), e19584. Buchsbaum, B.R., Baldo, J., Okada, K., Berman, K.F., Dronkers, N., D’Esposito, M., Hickok, G., 2011. Conduction aphasia, sensory-motor integration, and phonological short-term memory – an aggregate analysis of lesion and fMRI data. Brain Lang. 119 (3), 119–128. Bussey, T.J., Saksida, L.M., Murray, E.A., 2005. The perceptual-mnemonic/feature conjunction model of perirhinal cortex function. Q. J. Exp. Psychol. 58B, 269–282. Cappell, K.A., Gmeindl, L., Reuter-Lorenz, P.A., 2010. Age differences in prefontal recruitment during verbal working memory maintenance depend on memory load. Cortex 46 (4), 462–473. Catricalà, E., Della Rosa, P.A., Ginex, V., Mussetti, Z., Plebani, V., Cappa, S.F., 2013. An Italian battery for the assessment of semantic memory disorders. Neurol. Sci. 34 (6), 985–993. Catricalà, E., Della Rosa, P.A., Plebani, V., Vigliocco, G., Cappa, S.F., 2014. Abstract and concrete categories? Evidences from neurodegenerative diseases. Neuropsychologia 64, 271–281. Chan, A.M., Baker, J.M., Eskandar, E., Schomer, D., Ulbert, I., Marinkovic, K., Cash, S. S., Halgren, E., 2011. First-pass selectivity for semantic categories in human anteroventral temporal lobe. J. Neurosci. 31, 18119–181129. Chao, L.L., Haxby, J.V., Martin, A., 1999. Attribute-based neural substrates in temporal cortex for perceiving and knowing about objects. Nat. Neurosci. 2, 913–919. Chertkow, H., Bub, D., 1990. Semantic memory loss in dementia of Alzheimer's type. What do various measures measure? Brain 113 (Pt2), 397–417. Choi, H., Kim, J.H., Lee, C.M., Kim, J.I., 2013. Features of semantic language impairment in patients with amnestic mild cognitive impairment. Dement. Neurocogn. Disord. 12 (2), 33–40. Clague, F., Graham, K.S., Thompson, S.A., Hodges, J.R., 2011. Is knowledge of famous people compromised in mild cognitive impairment? Cogn. Behav. Neurol. 24 (3), 134–144. Clarke, A., Taylor, K.I., Devereux, B., Randall, B., Tyler, L.K., 2013. From perception to conception: How meaningful objects are processed over time. Cereb. Cortex 23, 187–197. Clune, J., Mouret, J.B., Lipson, H., 2013. The evolutionary origin of modularity. Proc. R. Soc. B 280, 20122863. Collignon, A., Maes, F., Delaere, D., Vandermeulen, D., Suetens, P., Marchal, G., 1995. Automated multi-modality image registration based on information theory. Inf. Process. Med. Imaging 3 (6), 263–274. Colombo, L., Brivio, C., Benaglio, I., Siri, S., Cappa, S.F., 2000. Alzheimer patients' ability to read words with irregular stress. Cortex 36 (5), 703–714. Convit, A., De Asis, J., De Leon, M.J., Tarshish, C.Y., De Santi, S., Rusinek, H., 2000. Atrophy of the medial occipitotemporal, inferior, and middle temporal gyri in non-demented elderly predict decline to Alzheimer's disease. Neurobiol. Aging 21 (1), 19–26. Dai, Z., He, Y., 2014. Disrupted structural and functional brain connectomes in mild cognitive impairment and Alzheimer's disease. Neurosci. Bull. 30 (2), 217–232. Damian, M.F., Als, L.C., 2005. Long-lasting semantic context effects in the spoken production of object names. J. Exp. Psychol.: Learn. Memory Cogn. 31, 1372–1384. Damian, M.F., Vigliocco, G., Levelt, W.J., 2001. Effects of semantic context in the naming of pictures and words. Cognition 81, B77–B86. de Haan, W., van der Flier, W.M., Koene, T., Smits, L.L., Scheltens, P., Stam, C.J., 2012. Disrupted modular brain dynamics reflect cognitive dysfunction in Alzheimer's disease. Neuroimage 59 (4), 3085–3093. de Zubicaray, G., Johnson, K., Howard, D., McMahon, K., 2014. A perfusion fMRI investigation of thematic and categorical context effects in the spoken production of object names. Cortex 54, 135–149. Dell’Acqua, R., Lotto, L., Job, R., 2000. Naming times and standardized norms for the Italian PD/DPSS set of 266 pictures: direct comparisons with American, English, French and Spanish published databases. Behav. Res. Methods Instrum. Comput. 32 (4), 588–615. Della Rosa, P.A., Catricalà, E., Vigliocco, G., Cappa, S.F., 2010. Beyond the abstract— concrete dichotomy: mode of acquisition, concreteness, imageability, familiarity, age of acquisition, context availability, and abstractness norms for a set of 417 Italian words. Behav. Res. Methods 42 (4), 1042–1048. Demonet, J.F., Chollet, F., Ramsay, S., Cardebat, D., Nespoulous, J.L., Wise, R., Frackowiak, R., 1992. The anatomy of phonological and semantic processing in normal subjects. Brain 115 (6), 1753–1768. Devlin, J.T., Moore, C.J., Mummery, C.J., Gorno-Tempini, M.L., Phillips, J.A., Noppeney, U., Price, C.J., 2002. Anatomic constraints on cognitive theories of category specificity. Neuroimage 15 (3), 675–685. Domoto-Reilly, K., Sapolsky, D., Brickhouse, M., Dickerson, B.C., 2012. Naming impairment in Alzheimer’s disease is associated with left anterior temporal lobe atrop. NeuroImage 63, 348–355. Duarte, L.R., Marquié, L., Marquié, J.C., Terrier, P., Ousset, P.J., 2009. Analyzing feature distinctiveness in the processing of living and non-living concepts in Alzheimer's disease. Brain Cogn. 71, 108–117. Eicker, F., 1963. Asymptotic normality and consistency of the least squares estimators for families of linear regressions. Ann. Math. Stat. 34 (2), 447–456. Eickhoff, S.B., Stephan, K.E., Mohlberg, H., Grefkes, C., Fink, G.R., Amunts, K., Zilles, K., 2005. A new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data. Neuroimage 25 (4), 1325–1335. Estévez-González, A., García-Sánchez, C., Boltes, A.O.N., Otermín, P., Pascual-

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

Sedano, B., Gironell, A., Kulisevsky, J., 2004. Semantic knowledge of famous people in mild cognitive impairment and progression to Alzheimer’s disea. Dement. Geriatr. Cogn. Disord. 17 (3), 188–195. Folstein, M.F., Folstein, S.E., McHugh, P.R., 1975. Mini-mental state: a practical method for grading the cognitive state of patients for the clinician. J. Psychiatr. Res. 12, 189–198. Frings, L., Kloppel, S., Teipel, S., Peters, O., Frolich, L., Pantel, J., Schroder, J., Gertz, H. J., Arlt, S., Heuser, I., Kornhuber, J., Wiltfang, J., Maier, W., Jessen, F., Hampel, H., Hull, M., 2011. Left anterior temporal lobe sustains naming in Alzheimer's dementia and mild cognitive impairme. Curr. Alzheimer Res. 8, 893–901. Fung, T.D., Chertkow, H., Murtha, S., Whatmough, C., Peloquin, L., Whitehead, V., et al., 2001. The spectrum of category effects in object and action knowledge in Alzheimer's disea. Neuropsychology 15, 371–379. Funnell, E., Hodges, J.R., 1991. Progressive loss of access to spoken word forms in a case of Alzheimer's disease. Proc. R. Soc. Lond. Ser. B: Biol. Sci. 243 (1307), 173–179. Gainotti, G., Quaranta, D., Vita, M.G., Marra, C., 2014. Neuropsychological predictors of conversion from mild cognitive impairment to Alzheimer's disease. Journal of Alzheimer's Disease 38 (3), 481–495. Gamboa, O.L., Tagliazucchi, E., von Wegner, F., Jurcoane, A., Wahl, M., Laufs, H., Ziemann, U., 2014. Working memory performance of early MS patients correlates inversely with modularity increases in resting state functional connectivity networks. NeuroImage 94, 385–395. Gardini, S., Cuetos, F., Fasano, F., Ferrari Pellegrini, F., Marchi, M., Venneri, A., Caffarra, P., 2013. Brain structural substrates of semantic memory decline in mild cognitive impairment. Curr. Alzheimer Res. 10 (4), 373–389. Garrard, P., Lambon Ralph, M.A., Hodges, J.R., Patterson, K., 2001. Prototypicality, distinctiveness and intercorrelation: analyses of the semantic attributes of living and nonliving concepts. Cogn. Neuropsychol. 18, 125–174. Garrard, P., Lambon Ralph, M.A., Patterson, K., Pratt, K.H., Hodges, J.H., 2005. Semantic feature knowledge and picture naming in dementia of Alzheimer's type: a new approa. Brain Lang. 93, 79–94. Garrard, P., Patterson, K., Watson, P.C., Hodges, J.R., 1998. Category specific semantic loss in dementia of Alzheimer's type: functional-anatomical correlations from cross-sectional analyses. Brain 121, 633–646. Gerlach, C., Law, I., Paulson, O.B., 2002. When action turns into words. Activation of motor-based knowledge during categorization of manipulable objects. J. Cogn. Neurosci. 14, 1230–1239. Giffard, B., Desgranges, B., Nore-Mary, F., Lalevee, C., Beaunieux, H., de la Sayette, V., Pasquier, F., Eustache, F., 2002. The dynamic time course of semantic memory impairment in Alzheimer's disease: clues from hyperpriming and hypopriming effects. Brain 125, 2044–2057. Giffard, B., Desgranges, B., Nore-Mary, F., Lalevée, C., da la Sayette, V., Pasquier, F., Eustache, F., 2001. The nature of semantic memory deficits in Alzheimer's disease. New insights from hyperpriming effec. Brain 124, 1522–1532. Gigi, A., Babai, R., Penker, A., Hendler, T., Korczyn, A.D., 2010. Prefrontal compensatory mechanism may enable normal semantic memory performance in mild cognitive impairment (MCI). J. Neuroimaging 20 (2), 163–168. Gläscher, J., & Gitelman, D., 2008. Contrast weights in flexible factorial design with multiple groups of subjects. SPM@ JISCMAIL. AC. UK Sml, editor. Gonnerman, L.M., Andersen, E.S., Devlin, J.T., Kempler, D., Seidenberg, M.S., 1997. Double dissociation of semantic categories in Alzheimer's disease. Brain Lang. 57 (2), 254–279. Grady, C.L., Springer, M., Hongwanishkul, D., McIntosh, A., Winocur, G., 2006. Agerelated changes in brain activity across the adult lifespan. J. Cogn. Neurosci. 18 (2), 227–241. Grossman, M., Koenig, P., Glosser, G., DeVita, C., Moore, P., Rhee, J., et al., 2003. Neural basis for semantic memory difficulty in Alzheimer’s disease: an fMRI stu. Brain 126 (2), 292–311. Grootoonk, S., Hutton, C., Ashburner, J., Howseman, A., Josephs, O., Rees, G., Friston, K., Turner, R., 2000. Characterization and correction of interpolation effects in the realignment of fmri time series. NeuroImage 11, 49–57. Guillaume, B., Hua, X., Thompson, P.M., Waldorp, L., Nichols, T.E., 2014. Fast and accurate modelling of longitudinal and repeated measures neuroimaging data. NeuroImage 94, 287–302. Harasty, J.A., Halliday, G.M., Kril, J.J., Code, C., 1999. Specific temporoparietal gyral atrophy reflects the pattern of language dissolution in Alzheimer's disease. Brain 122 (4), 675–686. Hirono, N., Mori, E., Ishii, K., Imamura, T., Tanimukai, S., Kazui, H., Hashimoto, M., Takatsuki, Y., Kitagaki, H., Sasaki, M., 2001. Neuronal substrates for semantic memory: a positron emission tomography study in Alzheimer's disease. Dement. Geriatr. Cogn. Disord. 12, 15–21. Hocking, J., McMahon, K.L., de Zubicaray, G.I., 2009. Semantic context and visual feature effects in object naming: an fMRI study using arterial spin labeling. J. Cogn. Neurosci. 21, 1571–1583. Hodges, J.R., 2006. Alzheimer's centennial legacy: origins, landmarks and the current status of knowledge concerning cognitive aspects. Brain 129, 2811–2822. Hodges, J.R., Salmon, D.P., Butters, N., 1992. Semantic memory impairment in Alzheimer’s disease: failure of access or degraded knowledg. Neuropsychologia 30, 301–314. Joseph, J.E., Gathers, A.D., 2003. Effects of structural similarity on neural substrates for object recognition. Cogn. Affect. Behav. Neurosci. 3, 1–16. Joubert, S., Brambati, S.M., Ansado, J., Barbeau, E.J., Felician, O., Didic, M., et al., 2010. The cognitive and neural expression of semantic memory impairment in mild cognitive impairment and early Alzheimer’s disease. Neuropsychologia 48 (4), 978–988.

151

Joubert, S., Felician, O., Barbeau, E.J., Didic, M., Poncet, M., Ceccaldi, M., 2008. Patterns of semantic memory impairment in Mild Cognitive Impairment. Behav. Neurol. 19 (1), 35–40. Kroll, J., Stewart, F., 1994. Category interference in translation and picture naming: Evidence for asymmetric connections between bilingual memory representations. J. Memory Lang. 33, 149–174.. Lenzi, D., Serra, L., Perri, R., Pantano, P., Lenzi, G.L., Paulesu, E., et al., 2011. Single domain amnestic MCI: A multiple cognitive domains fMRI investigation. Neurobiology of aging 32 (9), 1542–1557. Levelt, W.J.M., Roelofs, A., Meyer, A.S., 1999. A theory of lexical access in speech production. Behavioral and Brain Sciences 22, 1–38. Libon, D.J., Rascovsky, K., Powers, J., Irwin, D.J., Boller, A., Weinberg, D., Grossman, M., 2013. Comparative semantic profiles in semantic dementia and Alzheimer's disease. Brain 136, 2497–2509. Lin, C.Y., Chen, T.B., Lin, K.N., Yeh, Y.C., Chen, W.T., Wang, K.S., Wang, P.N., 2014. Confrontation naming errors in Alzheimer’s disease. Dement. Geriatr. Cogn. Disord. 37 (1–2), 86–94. Luria, A.R., 1947. Traumatic Aphasia. Mouton, The Hague, reprinted in translation 1970. Martin, A., Fedio, P., 1983. Word production and comprehension in Alzheimer's disease: the breakdown of semantic knowled. Brain Lang. 19, 124–141. Mazaika, P.K., Hoeft, F., Glover, G.H., 2009. Methods and software for fMRI analysis for clinical subjects. In: Proceedings of the Annual Meeting of the Organization for Human Brain Mapping. Melrose, R.J., Campa, O.M., Harwood, D.G., Osato, S., Mandelkern, M.A., Sultzer, D.L., 2009. The neural correlates of naming and fluency deficits in Alzheimer's disease: an FDG-PET study. Int. J. Geriatr. Psychiatry 24, 885–893. Minati, L., Chan, D., Mastropasqua, C., Serra, L., Spanò, B., Marra, C., Bozzali, M., 2014. Widespread alterations in functional brain network architecture in amnestic mild cognitive impairment. J. Alzheimer’s Dis. 40 (1), 213–220. Moss, H.E., Rodd, J.M., Stamatakis, E.A., Bright, P., Tyler, L.K., 2005. Anteromedial temporal cortex supports fine-grained differentiation among objects. Cereb. Cortex 15, 616–627. Nebes, R.D., 1992. Semantic memory dysfunction in Alzheimer’s disease: disruption of semantic knowledge or information-processing limitation?. In: Squire, L.R., Butters, N. (Eds.), Neuropsychology of Memory, 2nd ed The Guilford Press, New York, pp. 233–240. Nebes, R.D., Brady, C.B., Huff, F.J., 1989. Automatic and attentional mechanisms of semantic priming in Alzheimer's disease. J. Clin. Exp. Neuropsychol. 11, 219–230. Neuhaus, J., Kalbfleisch, J., 1998. Between- and within-cluster covariate effects in the analysis of clustered data. Biometrics, 638–645. Newman, M.E., 2006. Modularity and community structure in networks. Proc. Natl. Acad. Sci. USA 103, 8577–8696. Nilsson, L.G., 2003. Memory function in normal aging. Acta Neurol. Scand. Suppl. 179, 7–13. Ober, B.A., Shenaut, G.K., 1988. Lexical decision and priming in Alzheimer’s disease. Neuropsychologia 26 (2), 273–286. Oppenheim, G.M., Dell, G.S., Schwartz, M.F., 2010. The dark side of incremental learning: a model of cumulative semantic interference during lexical access in speech production. Cognition 114, 227–252. Patterson, K., Nestor, P.J., Rogers, T.T., 2007. Where do you know what you know? The representation of semantic knowledge in the human brain. Nat. Rev. Neurosci. 8 (12), 976–987. Penny, W.D., Holmes, A.P., 2003. Random effects analysis. In: Frackowiak, R.S., Friston, K.J., Frith, C.D., Dolan, R.J., Price, C.J., Ashburner, J., Penny, W.D., Zeki, S. (Eds.), Human Brain Function. Academic Press, London. Petersen, R.C., 2004. Mild cognitive impairment as a diagnostic entity. J. Intern. Med. 256, 183–194. Petersen, R.C., Doody, R., Kurz, A., et al., 2001. Current concepts in mild cognitive impairment. Arch. Neurol. 58, 1985–1992. Price, C.J., Devlin, J.T., Moore, C.J., Morton, C., Laird, A.R., 2005. Meta-analyses of object naming: effect of baseline. Hum. Brain Mapp. 25 (1), 70–82. Price, C.J., Friston, K.J., 1999. Scanning patients with tasks they can perform. Hum. Brain Mapp. 8, 102–108. Price, C.J., Noppeney, U., Phillips, J., Devlin, J.T., 2003. How is the fusiform gyrus related to category-specificity? Cogn. Neuropsychol. 20, 561–574. Reuter-Lorenz, P.A., 2013. Aging and cognitive neuroimaging: a fertile union. Perspect. Psychol. Sci. 8 (1), 68–71. Rogers, T.T., Hocking, J., Mechelli, A., Patterson, K., Price, C., 2005. Fusiform activation to animals is driven by the process, not the stimulus. J. Cogn. Neurosci. 17 (3), 434–445. Rogers, T.T., Hocking, J., Noppeney, U., Mechelli, A., Gorno-Tempini, M.L., Patterson, K., Price, C.J., 2006. Anterior temporal cortex and semantic memory: reconciling findings from neuropsychology and functional imaging. Cogn. Affect. Behav. Neurosci. 6 (3), 201–213. Romero, L., Walsh, V., Papagno, C., 2006. The neural correlates of phonological short-term memory: a repetitive transcranial magnetic stimulation study. J. Cogn. Neurosci. 18 (7), 1147–1155. Rubinov, M., Sporns, O., 2010. Complex network measures of brain connectivity: uses and interpretations. NeuroImage 52, 1059–1069.. Sperling, R.A., Dickerson, B.C., Pihlajamaki, M., Vannini, P., La Violette, P.S., Vitolo, O. V., et al., 2010. Functional alterations in memory networks in early Alzheimer's disease. Neuromolecular medicine 12 (1), 27–43. Sugarman, M.A., Woodard, J.L., Nielson, K.A., Seidenberg, M., Smith, J.C., Durgerian, S., Rao, S.M., 2012. Functional magnetic resonance imaging of semantic

152

E. Catricalà et al. / Neuropsychologia 76 (2015) 136–152

memory as a presymptomatic biomarker of Alzheimer's disease risk. Biochimica et Biophysica Acta (BBA)-Molecular Basis of Disease 1822 (3), 442–456. Sun, Y., Yin, Q., Fang, R., Yan, X., Wang, Y., Bezerianos, A., Sun, J., 2014. Disrupted functional brain connectivity and its association to structural connectivity in amnestic mild cognitive impairment and Alzheimer’s disease. PLoS One 9 (5), e96505. Teipel, S.J., Willoch, F., Ishii, K., Burger, K., Drzezga, A., Engel, R., Bartenstein, P., Moller, H.J., Schwaiger, M., Hampel, H., 2006. Resting state glucose utilization and the CERAD cognitive battery in patients with Alzheimer's disease. Neurobiol. Aging 27, 681–690. Tyler, L.K., Chiu, S., Zhuang, J., Randall, B., Devereux, B.J., Wright, P., et al., 2013. Objects and categories: feature statistics and object processing in the ventral stream. J. Cogn. Neurosci. 25 (10), 1723–1735. Tyler, L.K., Stamatakis, E.A., Bright, P., Acres, K., Abdallah, S., Rodd, J.M., et al., 2004. Processing objects at different levels of specificity. J. Cogn. Neurosci. 16, 351–362. Tyler, L.K., Bright, P., Dick, E., Tavares, P., Pilgrim, L., Fletcher, P., Greer, M., Moss, H., 2003. Do semantic categories activate distinct cortical regions? Evidence for a distributed neural semantic system. Cogn. Neuropsychol. 20 (3–6), 541–559. Tzourio-Mazover, N., Landeau, B., Papthanassiou, D., Crivello, F., Etard, O., Delcroix, N., et al., 2002. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 15, 273–289. Varon, D., et al., 2014. Visual rating and volumetric measurement of medial temporal atrophy in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort: baseline diagnosis and the prediction of MCI outcome. Int. J. Geriatr. Psychiatry , http://dx.doi.org/10.1002/gps.4126 (May). Viggiano, M.P., Vannucci, M., Righi, S., 2004. A new standardized set of ecological pictures for experimental and clinical research on visual object processing. Cortex 40 (3), 491–509. Vigliocco, G., Vinson, D.P., Damian, M.F., Levelt, W., 2002. Semantic distance effects on object and action naming. Cognition 85 (3), B61–B69. Visser, M., Jefferies, E., Lambon Ralph, M.A., 2010. Semantic processing in the anterior temporal lobes: a meta-analysis of the functional neuroimaging literature. J. Cogn. Neurosci. 22, 1083–1094. Wang, J., Zuo, X., Dai, Z., Xia, M., Zhao, Z., Zhao, X., He, Y., 2013. Disrupted functional brain connectome in individuals at risk for Alzheimer's disease. Biol. Psychiatry

73 (5), 472–481. Weiskopf, N., Hutton, C., Josephs, O., Deichmann, R., 2006. Optimal EPI parameters for reduction of susceptibility-induced BOLD sensitivity losses: a whole-brain analysis at 3T and 1.5 T. Neuroimage 33 (2), 493–504. Whatmough, C., Chertkow, H., Murtha, S., Templeman, D., Babins, L., Kelner, N., 2003. The semantic category effect increases with worsening anomia in Alzheimer's type dementia. Brain Lang. 84 (1), 134–147. Wierenga, C.E., Benjamin, M., Gopinath, K., Perlstein, W.M., Leonard, C.M., Rothi, L.J. G., et al., 2008. Age-related changes in word retrieval: role of bilateral frontal and subcortical networks. Neurobiol. Aging 29 (3), 436–451. Wierenga, C.E., Bondi, M.W., 2007. Use of functional magnetic resonance imaging in the early identification of Alzheimer's disease. Neuropsychol. Rev. 17, 127–143. Wierenga, C.E., Stricker, N.H., McCauley, A., Simmons, A., Jak, A.J., Chang, Y.L., et al., 2010. Increased functional brain response during word retrieval in cognitively intact older adults at genetic risk for Alzheimer's disease. Neuroimage 51 (3), 1222–1233. Woodard, J.L., Seidenberg, M., Nielson, K.A., Antuono, P., Guidotti, L., Durgerian, S., Zhang, Q., Lancaster, M., Hantke, N., Butts, A., Rao, S.M., 2009. Semantic memory activation in amnestic mild cognitive impairment. Brain 132, 2068–2078. Woollams, A.M., Ralph, M.A.L., Plaut, D.C., Patterson, K., 2007. SD-squared: on the association between semantic dementia and surface dyslexia. Psychol. Rev. 114 (2), 316. Xia, M., Wang, J., He, Y., 2013. BrainNet Viewer: a network visualization tool for human brain connectomics. PLoS One 8 (7), e68910. Zahn, R., Garrard, P., Talazko, J., Gondan, M., Bubrowski, P., Juengling, F.D., et al., 2006. Patterns of regional brain metabolism associated with knowledge of semantic features and categories in Alzheimer’s disease. J. Cogn. Neurosci. 18, 2138–2151. Zahn, R., Juengling, F., Bubrowski, P., Jost, E., Dykierek, P., Talazko, J., Huell, M., 2004. Hemispheric asymmetries of hypometabolism associated with semantic memory impairment in Alzheimer’s disease: a study using positron emission tomography with fluorodeoxyglucose-F18. Psychiatry Res.: Neuroimaging 132 (2), 159–172.