Metabolic characterization of diabetic retinopathy: An 1H-NMR-based metabolomic approach using human aqueous humor

Journal of Pharmaceutical and Biomedical Analysis 174 (2019) 414–421

Contents lists available at ScienceDirect

Journal of Pharmaceutical and Biomedical Analysis journal homepage: www.elsevier.com/locate/jpba

Metabolic characterization of diabetic retinopathy: An 1 H-NMR-based metabolomic approach using human aqueous humor Huiyi Jin a,b , Bijun Zhu a,b,∗ , Xia Liu c , Jing Jin d , Haidong Zou a,b,e,∗ a

Department of Ophthalmology, Shanghai General Hospital, Shanghai Jiao Tong University, Shanghai, China Shanghai Key Laboratory of Fundus Disease, Shanghai, China CAS Key Laboratory of Receptor Research, Department of Analytical Chemistry, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai, China d Department of Ophthalmology, Shanghai Children’s Medical Center, Shanghai Jiao Tong University School of Medicine, Shanghai, China e Department of Preventative Ophthalmology, Shanghai Eye Diseases Prevention and Treatment Center/Shanghai Eye Hospital, Shanghai, China b c

a r t i c l e

i n f o

Article history: Received 1 February 2019 Received in revised form 2 June 2019 Accepted 8 June 2019 Available online 10 June 2019 Keywords: Metabolomics NMR analysis Aqueous humor Diabetic retinopathy Metabolite-metabolite correlation

a b s t r a c t Patients with a long duration of diabetes mellitus (DM) usually have accompanied complications such as diabetic retinopathy (DR), which is a leading cause of blindness and visual impairment among workingage persons in developed countries; nevertheless, some patients have no complications. Thus, various studies, including genomic, transcriptomic, and proteomic studies, have been conducted to identify potential biomarkers for predicting DR and to reveal the underlying disease mechanism. Although metabolomics could be a powerful tool for characterizing aqueous eye fluids and revealing the metabolic signatures of common ocular diseases such as DR, studies about its relationship with DR are limited. Moreover, to our knowledge, no previous study has applied a metabolomic approach to investigate the aqueous humor in DR. Therefore, we performed an NMR-based metabolomic study of the aqueous humor of patients with DM and cataract (DM, n = 13), DR and cataract (DR, n = 14), and senile cataract (CON, n = 7) to investigate the metabolic alterations accompanying the development of DR. Principal component analysis, average change analysis, and heatmap analysis revealed that lactate, succinate, 2-hydroxybutyrate, asparagine, dimethylamine, histidine, threonine, and glutamine were the most altered metabolites that potentially play roles in the development and progression of DR. The highly activated alanine, aspartate, and glutamate metabolic pathway was selected using pathway analysis. The phenotypic metabolomic analyses of the aqueous humor indicated an alteration in the metabolic pathways of energy metabolism and amino acids in DR patients which was to some extent suggestive of the pathophysiological process of mitochondrial dysfunction and oxidative stress/endothelial damage. It provides a proof of concept that metabolomic analysis using the aqueous humor of DM patients may be a reliable method to improve the accuracy of predicting the development and progression of DR. © 2019 Elsevier B.V. All rights reserved.

1. Introduction The prevalence of diabetes mellitus (DM) is increasing worldwide, particularly in China, which ranks first in the world with 114 million patients with DM [1]. Diabetic retinopathy (DR), a microvascular complication of DM, is a leading cause of blindness and visual impairment among working-age persons in developed countries. By 2030, the numbers of patients with DR and vision-threatening

∗ Correspondence authors at: No. 100, Haining Road, Shanghai, 200080, China. E-mail addresses: [email protected] (B. Zhu), [email protected] (H. Zou). https://doi.org/10.1016/j.jpba.2019.06.013 0731-7085/© 2019 Elsevier B.V. All rights reserved.

DR have been estimated to increase to 191.0 million and 56.3 million, respectively [2]. Long duration of DM, poor blood glucose, and poor blood pressure control are the major traditional risk factors for DR. However, some patients with both long duration of DM and poor glycemic control do not develop DR. Therefore, understanding the pathophysiology and mechanisms of DR is important for its prevention and treatment. Metabolomics is the qualitative and quantitative assessment of the metabolites (small molecules<1.5 kDa) in body fluids. These metabolites are downstream of the genetic transcription and translation processes as well as downstream of the interactions with environmental exposures; thus, they are considered closely related to the phenotype, especially in multifactorial diseases. While there

H. Jin et al. / Journal of Pharmaceutical and Biomedical Analysis 174 (2019) 414–421

has been an increasing interest in and more research focus on metabolomics of the eye, the application of metabolomics to retinal diseases has been limited [3]. Few studies have examined the metabolomic underpinnings of DR. Vitreous and serum samples are the main tissues examined, and dysregulation in the pathways, such as the pentose phosphate, arginine to proline, polyol, and ascorbic acid pathways, has been reported [4]. The aqueous humor (AH) is a transparent fluid found in the anterior and posterior chambers of the eye. It is continuously secreted by the ciliary epithelium and enters first into the posterior chamber; thereafter, it travels through the pupil toward the anterior chamber and the trabecular meshwork, passively traveling towards the episcleral venous system. The composition of the AH depends not only on the nature of its production, but also on the metabolic interchanges that occur within various tissues throughout its intraocular route. The AH in glaucoma and myopia has been previously studied through a metabolomics approach in murine models [5] and humans [6], but to our knowledge, the AH in DR has never been studied through a metabolomics approach in humans. Thus, in this study, we explored for the first time the differential metabolic profile of the AH of patients with DR by using a systematic 1 H-NMR-based metabolomic analysis. Through principal component analysis, fold change analysis and heatmap analysis, we performed comparative analyses of the AH obtained from patients with DM and cataract, those with DR and cataract, and those with senile cataract (control group) to identify the most changed metabolites during the development of DR. Our findings provide a possible basis for further pathophysiological studies of DR and a potential complementary method for the prognosis of this disease.

415

transferred to dust-free Eppendorf tubes and stored at −80 ◦ C until metabolomic analysis. On the day of NMR analysis, the samples were defrosted and further dissolved in a 500-␮L solution of 0.25 mM sodium3 -trimethylsilylpropionate-2,2,3,3-d4 (TSP) in deuterium oxide (D2 O). This study was approved by the Human Investigation Ethics Committee of Shanghai General Hospital Affiliated to Shanghai Jiao Tong University and was conducted in accordance with the tenets of the Helsinki Declaration. All patients provided their informed consents prior to inclusion in the examination. 2.2. NMR measurements

2. Materials and methods

All NMR spectra were acquired at 298 K on a Bruker AvanceTM III 600-MHz spectrometer (Bruker GmbH, Rheinstetten, Germany) equipped with a cryogenic probe at 600.17 MHz for 1 H observation. NMR data of the AH samples were recorded using a solventsuppressed 1D 1 H Carr-Purcell-Meiboom-Gill (CPMG) pulse sequence [RD-90◦ -(␶-180◦ -␶)n-ACQ]. A fixed total spin-spin relaxation delay 2 n␶ of 120 ms was applied to attenuate the broad NMR signals of slowly tumbling molecules with short T2 relaxation times and to retain signals of low-molecular-weight compounds. The 1 HNMR spectrum was recorded using four dummy scans and 128 transients into 32 K data points by using a spectral width of 20 ppm with a relaxation delay of 10.0 s and an acquisition time of 2.73 s. Additional 2D pulsed-field gradient COrrelationSpectroscopY (gCOSY) together with 2D homonuclear Total Correlation Spectroscopy (TOCSY) were performed using standard Bruker pulse programs on selected samples for confirming the chemical shift assignments.

2.1. Patients collection and aqueous humor preparation

2.3. Data processing for statistical analyses

Thirty-four male subjects were recruited for this study and were divided into three groups: 14 with type 2 DM and cataract, 13 with DR and cataract, and 7 with senile cataract (controls). The three groups were closely matched in terms of age (69.86 ± 9.30 vs. 61.77 ± 8.39 vs. 66.71 ± 12.16, respectively). All 34 subjects met the following inclusion criteria: (a) patients with DR confirmed using previous slit-lamp biomicroscopy and fluorescein angiography examinations; (b) patients with the explicit diagnosis of cataract in the operated eye (nuclear hardness grade≥3; nuclear hardness classification by Emery and Little [7]), who were willing to undergo cataract surgery; and (c) patients without other eye diseases, such as keratopathy, glaucoma, dacryocystitis, uveitis, ocular trauma, and age-related macular degeneration. The exclusion criteria were as follows: history of ocular surgery and retinal photocoagulation; fasting blood glucose greater than 8.5 mmol/L; renal failure (plasma creatinine≥120 ␮M); other chronic diseases apart from DM; and intake of systemic anti-metabolites, immunosuppressants, or corticosteroids. The DR status of each subject was assessed through indirect ophthalmoscopy and color fundus photography (FF 540 Plus; Carl Zeiss Meditec, Jena, Germany). DR in this study was graded according to the Early Treatment Diabetic Retinopathy Study criteria [8]. One experienced ophthalmologist classified the DR status based on the examination results. All patients had fasted 12 h prior to the procedure. Collection of AH was performed by the same operator. Briefly, after rinsing twice with 5% povidine iodine, 1 or 2 drops of proparacaine hydrochloride 0.5% (Alcaine, Alcon, Ft. Worth, TX, USA) were applied twice to the eye about to undergo surgery. Approximately 100–150 ␮L of aqueous humor were collected under the surgical microscope using a 1 ml tuberculin syringe and a 30-gauge blunt needle at the beginning of the surgical intervention. The AH samples were immediately

All free induction decay NMR signals were processed using an exponential function with a 0.03-Hz line broadening and zerofilling to 64 K data points. Each spectrum was manually phased, baseline-corrected, and carefully aligned. Then, all the 1D NMR spectra were referenced to the methyl group of TSP at 0.00 ppm. The spectral non-overlapped regions of each metabolite were selected and binned into one integral. The integrals of the resulting 25 metabolites in the AH samples were normalized to the sum of the spectral intensities (excluding the regions of the residual water) to compensate for differences in the concentrations of samples. Then, the well-processed data were imported into SIMCA-P software (Version 14.0; Umetrics AB, Umea, Sweden) for multivariate pattern recognition analyses. Principal component analysis (PCA) was first performed to detect any group separation based on NMR signal variability as well as to identify outliers. Subsequently, the normalized integral values were UV-scaled for orthogonal partial least-squares discriminant analysis (OPLS-DA). OPLS-DA was illustrated using the first predictive (t [1]) and one orthogonal component (to [1]). The terms R2X, R2Y, and Q2 were used to evaluate the quality of the OPLS-DA models. R2X and R2Y were the respective fractions of X and Y variance in the data explained by the model, and they indicated the goodness of fit. Q2 represented the crossvalidated explained variation, and it indicated predictability. The standard seven-round cross validation and permutation test (200 cycles) was carried out to measure the robustness of the model. Variable importance in the projection (VIP) derived from the OPLS-DA model ranked the importance of each variable for the classification, and those variables with VIP > 1.0 were initially considered statistically significant in this model. The correlation coefficients (r) of the variables relative to the first model score value in the OPLS-DA model were also extracted from the S-plot calculated using Pearson correlation. Cutoff values of r with a significance

416

H. Jin et al. / Journal of Pharmaceutical and Biomedical Analysis 174 (2019) 414–421

Table 1 Biochemical parameters of the participants in this study (mean ± SD).

Age (years) Glc (mmol/L) Alb (g/L) TBIL (umol/L) ALT (U/L) AST (U/L) Cr (umol/L) Urea (mmol/L) UA (umol/L) TG (mmol/L) TC (mmol/L)

DM, DR (N = 13)

DM, no DR (n = 14)

Controls (n = 7)

P values

Power (%)

61.77 ± 8.34 5.82 ± 1.58 41.80 ± 6.04 10.20 ± 3.38 26.50 ± 20.40 24.740 ± 11.68 98.40 ± 22.20 7.56 ± 2.77 390.00 ± 76.20 1.40 ± 0.64 3.51 ± 0.76

69.86 ± 9.30 6.29 ± 1.07 43.14 ± 2.14 14.15 ± 4.38 28.49 ± 12.01 24.56 ± 9.30 80.01 ± 15.34 6.88 ± 1.70 388.1 ± 110.2 1.57 ± 1.11 4.25 ± 0.82

66.71 ± 12.16 5.50 ± 0.51 45.85 ± 3.03 11.58 ± 2.47 20.82 ± 8.20 23.80 ± 12.34 84.67 ± 13.71 6.05 ± 2.06 371.00 ± 44.49 1.61 ± 0.70 4.08 ± 0.53

0.11 0.36 0.17 0.14 0.54 0.99 0.15 0.49 0.92 0.92 0.21

8.39 6.73 11.14 12.72 5.61 5.00 10.57 5.83 5.01 5.01 10.00

Glc: glucose; Alb: albumin; TBIL: total bilirubin; ALT: alanine aminotransferase; AST: aspartate aminotransferase; Cr: Creatinine; UA: Uric Acid; TG: triglyerides; TC: total cholesterol.

level of 0.05 were used to identify variables that were responsible for the discrimination of groups [9]. Thus, the integrals of metabolites that had a VIP > 1 and the correlation coefficient |r|>the critical value when P = 0.05 accounted for the detected discrimination. The ROC curve plot of each statistically significant metabolite between the groups and the heatmap was illustrated to validate their discriminative potential. 2.4. Quantitative comparison of the metabolites The average changes in the metabolites between the groups were calculated, and significant differences in the mean values were evaluated using one-way analysis of variance, followed by Bonferroni’s or Turkey’s post-hoc analysis when appropriate. Spearman correlation analysis was used to evaluate the association among all the resulting metabolites. Statistical significance was considered at P < 0.05. All the statistical analyses were performed using SPSS Statistics for Windows/Macintosh, Version 17.0 (SPSS Inc., Chicago, IL). The compound names and KEGG (http:// www.kegg.jp/) numbers of potential biomarkers were performed with MetaboAnalyst 3.0 (http://www.metaboanalyst.ca/) for further enrichment and pathway analysis [10], which was used to identify the perturbed metabolic pathway in DR metabolism. 3. Results The clinical parameters of patients recruited in this study were shown in Table 1. The mean ages of patients in three groups had no statistical significance (P = 0.11). There was no difference regarding serum glucose, albumin, total bilirubin, alanine aminotransferase (ALT), aspartate aminotransferase (AST), creatinine, uric acid, triglyceride, or total cholesterol, which were controlled before the cataract surgery. Representative spectra of the AH samples of patients with DM, patients with DR, and healthy controls are provided in the supplementary material (Fig. S1). The observed metabolites in these AH spectra included ascorbate (Asc), citrate (Cit), creatine (Cr), dimethylamine (DMA), formate (For), glucose (Glc), isobutyrate (IB), lactate (Lac), succinate (Suc), 2-hydroxyisovalerate (2HIV), 2-hydroxybutyrate (2HB), 2-oxoisocaproate (2OIC), and amino acids including alanine (Ala), asparagine (Asn), glutamine (Gln), histidine (His), isoleucine (Ile), leucine (Leu), lysine (Lys), methionine (Met), phenylalanine (Phe), tyrosine (Tyr), threonine (Thr), valine (Val), and 2-aminobutyrate (2AB), as summarized in Table S1. 3.1. Metabolic alterations in patients with DR, patients with DM without DR, and controls Visual inspection of the spectra could hardly reveal the different metabolic fingerprint profiles of the AH metabolites among the

groups. To perform a comprehensive comparison of the metabolic profiles among the groups, PCA and OPLS-DA with the first two principal components (t[1], t[2]) were employed (Fig. 1), and these revealed diverse trends among the different groups. The healthy control individuals and patients with DM clustered to the right section, whereas patients with DR clustered much closer to the left section. However, partial overlaps among the groups were also observed in the PCA score plots. To further identify the metabolic characteristics of patients, pairwise OPLS-DA was performed. Obvious separations between DM and DR, DM and CON, and DR and CON (Fig. 2A, B, C), with the confirmation of the cross-validation results of the permutation tests (Fig. 2A’, B’, C’), indicated that significant metabolic alterations occurred during the development of DR. The VIP and p(corr) values extracted from the OPLS-DA plot revealed that compared to the patients with DM, those with DR showed significantly decreased levels of lactate, succinate, and 2HB, and significantly increased levels of asparagine, histidine, glutamine, threonine, and DMA in the AH samples. Compared to the controls, the patients with DR showed significantly downregulated levels of lactate, succinate, ascorbate, and formate, and obviously upregulated levels of asparagine and isoleucine in the AH samples. The levels of threonine, glutamine, histidine, 2HB, and DMA lost their statistical significance in patients with DR and the controls. We calculated the average changes in the metabolites based on the 1 H-NMR variables between the groups and summarized the data in Table S2. All these above-mentioned discriminative metabolites were further validated and confirmed using the AUROC values (Fig. 3A, B, C). Additionally, by using the identified significant metabolites, the heatmap plots were drew to classify the upregulated and downregulated metabolites (Fig. 3A’, B’, C’). 3.2. Metabolite-metabolite correlation analysis and metabolic network variation based on the altered metabolites Spearman correlation analysis was used to identify potential links among all the resulting metabolites to reveal the possible metabolic network of the AH in DR. The results revealed that the correlation of lactate and succinate and the correlation of asparagine and glutamine were remarkable between the DR group and the DM group (Fig. 3A”), while these correlations disappeared in other two-pair groups (Fig. 3B”, C”), implying that the alteration of glucose metabolism and glutamine and asparagine metabolism played a role in the development of DR. To explore metabolic pathways influenced between DR group and DM group, pathway analysis was performed by MetaboAnalyst 3.0, which combines results from powerful pathway enrichment analysis with the topology analysis. Metabolic pathway analysis revealed that over 11 pathways were influenced such as alanine, aspartate and glutamate metabolism, aminoacyl-tRNA biosynthesis, propanoate metabolism, nitrogen metabolism, and so on

H. Jin et al. / Journal of Pharmaceutical and Biomedical Analysis 174 (2019) 414–421

417

Fig. 1. PCA score scatter plot (A) and OPLS-DA scatter plot (A’) for the 1 H-NMR data collected using the 1D CPMG spectra of the aqueous humor samples by using 25 measured metabolites from 14 patients with DM without DR ( ), 13 with DM and DR ( ), and 7 controls ( ).

Table 2 Result from pathway analysis with MetaboAnalyst 3.0. Pathway name

Total

Expected

Hits

Raw p

-log(p)

Impact

Alanine, aspartate and glutamate metabolism Aminoacyl-tRNA biosynthesis Propanoate metabolism Nitrogen metabolism D-Glutamine and D-glutamate metabolism Cyanoamino acid metabolism Citrate cycle (TCA cycle) Valine, leucine and isoleucine biosynthesis beta-Alanine metabolism Glycolysis or Gluconeogenesis Pyruvate metabolism

24 75 35 39 11 16 20 27 28 31 32

0.079767 0.24927 0.11633 0.12962 0.03656 0.053178 0.066473 0.089738 0.093062 0.10303 0.10636

3 4 3 3 1 1 1 1 1 1 1

0.0000472 0.0000554 0.00015017 0.00020838 0.036032 0.052031 0.064663 0.086414 0.089485 0.098643 0.10168

9.9601 9.8005 8.8037 8.4762 3.3233 2.9559 2.7386 2.4486 2.4137 2.3162 2.2859

0.25261 0 0.00134 0.00763 0.02674 0 0.01446 0 0 0 0.13756

The totals are the number of compounds in the pathway; the hits are the matched number from the user uploaded data; the raw p is the original p value calculated from the enrichment analysis; and the impact is the pathway impact value calculated from pathway topology analysis.

(Table 2 and Fig. 4). An impact value>0.1 and a hit value>3 were used as the threshold to identify the significantly altered metabolic pathways [11]. Accordingly, the alanine, aspartate, and glutamate metabolic pathway was obviously changed between DR group and DM group (Fig. 4A), and three metabolites (glutamine, succinate, and asparagine) involved in this pathway (Fig. 4A’).

4. Discussion Metabolomics is a powerful approach for studying pathophysiological processes, and metabolomic analysis of human samples may shed light on the mechanisms of DR and help identify potential therapeutic targets. To date, this field remains in its infancy with few studies published and little replication of results. In the present study, the AH was used for the first time in a metabolomic analysis of non-diabetic controls and patients with DM with and without DR. In addition to being actively secreted by the ciliary epithelium, the AH can be produced by diffusion and ultrafiltration from the blood. It moves through the ocular chambers and drains from the eye to the venous blood through its continuous formation. Therefore, we believe AH samples can be another important tissue to reflect the metabolomic abnormality in DR. Our results revealed clear group separations between DR versus CON, DR versus DM, and DM versus CON when using the OPLS-DA score plots. Among the identified perturbed metabolites, the levels of lactate and succinate dramatically decreased, and their positive correlation was observed in DR samples which was missed in DM group and control group. Recent evidence suggests that lactate is not only a terminal product of glycolysis but also an important intermediary in numerous metabolic processes, a particularly mobile fuel for aerobic metabolism [12]. It can be converted to pyruvate which would be oxidized to acetyl-CoA and continue through

the tricarboxylic acid (TCA) cycle. Succinate is an intermediate of the TCA cycle and plays a crucial role in adenosine triphosphate (ATP) generation in mitochondria. Trudeau et al revealed that high-glucose-induced mitochondrial dysfunction resulted in fragmentation and lowered capacity for cellular respiration [13]. Thus, one possible explanation for the simultaneous decrease in lactate and succinate might be related to mitochondrial damages and concomitantly reduced energy metabolism. In addition, studies revealed that succinate was concentrated in body fluids and it served as a universal metabolic signal of local stress and immunologic danger [14]. However, Matsumoto et al [15] reported that there were no significant differences found for the aqueous humor or serum succinate levels between epiretinal membrane patients and proliferative diabetic retinopathy (PDR) patients, although succinate significantly increased in the vitreous fluid of PDR patients. The succinate change in AH may be due to the difference between retina and ciliary body and iris in oxygen demand and mitochondrial reserve capacity [16]. As such, succinate might accumulate in the retina where the oxygen supply is insufficient, and not in either the corneal endothelium or the iris in PDR patients. Furthermore, compared to Matsumoto’s study [15], the severity of DR in our study was milder so that no vitrectomy surgery was needed. Thus, we speculated that another possible reason for lactate and succinate decrease was that ciliary body and iris might lower its constitutively metabolic activity and try to maintain energy homeostasis and to reduce the production of ROS in the relative early or middle stage of DR. Further study will be needed to clarify the mechanism of their decrease. Furthermore, increased levels of asparagine, glutamine, histidine, and threonine were observed in the AH of patients with DR than in those with DM. Depending on their metabolic fates, these four amino acids are all glucogenic because they can be broken

418

H. Jin et al. / Journal of Pharmaceutical and Biomedical Analysis 174 (2019) 414–421

Fig. 2. The OPLS-DA score plots (left panel) and permutation tests (right panel) with the seven-round cross-validation of the 1 H-NMR spectra derived from the aqueous humor samples in the groups (A, A’) DM without DR ( ) vs. DM with DR ( ), (B, B’) DM without DR ( ) vs. CON ( ), and (C, C’) DM with DR ( ) vs. CON ( ).

down into oxaloacetate, ␣-ketoglutarate or succinyl CoA which can be oxidized into CO2 and H2 O to generate ATP via the TCA cycle and oxidative phosphorylation (Fig. 5). It is important to note that TCA cycle is controlled by the enzymes isocitrate dehydrogenase and ␣-ketoglutarate dehydrogenase which are inhibited by increased concentration of ATP and NADH. Previous studies have shown that hyperglycemia and hypoxia have additive effects on accumulation of electrons and protons in a common pool of free intracellular NADH and that NAD/NADH ratio in diabetic retina was decreased dramatically [17,18]. Thus, the increase of these amino acids might be relevant to the high levels of NADH and NADH/NAD which reduced TCA cycle and downregulated catabolism in patients with DR.

Moreover, DMA level was higher in patients with DR than in those with DM. DMA is one of the most important metabolites of asymmetric dimethylarginine (ADMA), which is an endogenous inhibitor of nitric oxide synthase (NOS) and may decrease NO availability [19]. Studies have shown that ADMA levels were elevated in both aqueous humor from diabetic rats and culture medium in hypoxia-treated cells [20] and that serum levels of ADMA have been found elevated in diabetic patients with DR [21]. ADMA could contribute to oxidative stress by causing endothelial nitric oxide synthase (eNOS) uncoupling [22], which is an important mechanism underlying the pathological phenomenon of reduced NO production and enhanced free radical generation. Though there is no direct study about DMA and DR, we speculated that increased

H. Jin et al. / Journal of Pharmaceutical and Biomedical Analysis 174 (2019) 414–421

419

Fig. 3. The AUROC plots (left panel) and heatmap analysis (middle panel) of discriminative metabolites obtained from the OPLS-DA in Fig. 2, and Spearman correlation analysis (right panel) of all the resulting metabolites.

Fig. 4. Summary of pathway analysis with MetaboAnalyst 3.0 (A). (1) Alanine, aspartate and glutamate metabolism; (2) Aminoacyl-tRNA biosynthesis; (3) Propanoate metabolism; (4) Nitrogen metabolism; (5) D-Glutamine and D-glutamate metabolism; (6) Cyanoamino acid metabolism; (7) Citrate cycle (TCA cycle); (8) Valine, leucine and isoleucine biosynthesis; (9) beta-Alanine metabolism; (10) Glycolysis or Gluconeogenesis; (11) Pyruvate metabolism. Pathway analysis shows that the alanine, aspartate, and glutamate metabolic pathway (A’) is more likely to be involved in DR development. Three metabolites (glutamine, succinate, and asparagine) are involved in this pathway.

420

H. Jin et al. / Journal of Pharmaceutical and Biomedical Analysis 174 (2019) 414–421

Fig. 5. An overall metabolic network. The red-colored metabolites represented metabolites higher in DR group than in DM group; the blue-colored metabolites represented metabolite lower in DR group than in DM group (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

levels of DMA may, at least in part, reflect a state of enhanced oxidative stress and endothelial dysfunction in patients with DR. Ascorbate, a primary antioxidant that detoxifies exogenous radicals or superoxides generated by mitochondrial metabolism, has long been known to show an obvious reduction in the plasma and vitreous humor of patients with diabetic complications [23,24]. The AH is partly ultrafiltered from the blood and diffuses to the posterior chambers, and our findings obtained using the AH samples are consistent with those of the above studies. Several studies suggest that the oxidative stress in DM causes a high turnover of ascorbate [25], and ascorbate’s systemic or local deficiency may promote pericyte and endothelial dysfunction in diabetic retinas [26]. This implied the interplay between ascorbate decrease and oxidative stress/endothelial dysfunction in DR. Hence, these perturbed metabolites seem to suggest energy metabolism deficiencies and oxidative stress damage in patients with DR. ␣-Hydroxybutyrate (␣-HB) is an organic acid derived from ␣ketobutyrate (␣-KB), which is produced by amino acid catabolism (threonine and methionine) and glutathione anabolism (cysteine formation pathway) [27]. In the present study, we found that ␣-HB levels in patients with DR were significantly lower than those in patients with DM but were not significantly different from those in the control group. A previous study revealed that ␣-HB levels were higher and could be an early biomarker in insulin resistant subjects, suggesting that the underlying biochemical mechanisms may involve increased lipid oxidation and oxidative stress [28]. This is consistent with our findings that ␣-HB levels in patients with DM were much higher than those in the control group. Moreover, with increased oxidative stress and endothelial dysfunction in patients with DR, amino acid catabolism and antioxidant synthesis, such as those of threonine and glutathione, were decreased, thereby which reduced ␣-KB and ␣-HB formation in our patients with DR. A consideration when examining the AH metabolome of DR in comparison with that of patients without DM presented here is the relative age and sex of the cohort profiled. As age and sex are wellestablished factors impacting the metabolome of systemic fluids like the plasma and urine [29,30], we prospectively recruited only age-matched male patients. However, the small size of each group and weakly matched sample size between each group were the limitations of this study which made the power of the study insufficient. In view of the suggestive results obtained, we are recruiting more patients and collecting more detailed clinical characteristics of patients to confirm our metabolomic findings. 5. Conclusion Our NMR-based metabolomic study identified 25 principal metabolites in the AH samples of patients with cataract and DM or

DR. The decreases in lactate and succinate levels and the increases in asparagine, histidine, glutamine, and threonine levels were evident in the DR group than in the DM and control groups. The highly activated alanine, aspartate, and glutamate metabolic pathway was identified using pathway analysis. All these results provide a preliminary indication of an alteration in the metabolic pathways of energy metabolism and amino acids in the AH of patients with DR, to some extent, consistent with the pathophysiological process of mitochondrial dysfunction and oxidative stress/endothelial damage, thereby indicating that NMR-based metabolomics may be used to distinguish patients with DR from healthy controls and patients with DM. Despite the invasive nature of AH and vitreous sampling, the AH is a more accessible sample to predict the risk of DR development after surgery, especially in patients with cataract and DM. Hence, it would be more widely utilized for risk prediction and would have greater clinical applications in the future. Acknowledgements We thank all the patients for their participation in this study. This study was also supported by Chinese National Natural Science Foundation (No. 81670898, 81400388), National Key R&D Program of China (No. 2016YFC0904800), The Chronic Diseases Prevention and Treatment Project of Shanghai Shen Kang Hospital Development Centre (Grant No. SHDC12015315, SHDC2015644), The Shanghai Outstanding Academic Leader Program (Project No.16XD1402300), Shanghai Municipal Education Commission—Gaofeng Clinical Medicine Grant Support (Project No. 20172022), Project of Shanghai Science and Technology (Grant No. 17411950200 & 17411950202). The authors declared no conflict of interest. Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.jpba.2019.06. 013. References [1] International Diabetes Federation, the IDF Diabetes Atlas, 2017 Global Fact Sheet, 8th edition, 2017 (Accessed 22 January 2019) https://www.idf.org/ aboutdiabetes/what-is-diabetes/facts-figures.html. [2] S. Wild, G. Roglic, A. Green, R. Sicree, H. King, Global prevalence of diabetes: estimates for the year 2000 and projections for 2030, Diabetes Care 27 (5) (2004) 1047–1053. [3] I. Laíns, M. Gantner, S. Murinello, J.A. Lasky-Su, J.W. Miller, M. Friedlander, D. Husain, Metabolomics in the study of retinal health and disease, Prog. Retin. Eye Res. 69 (2019) 57–79.

H. Jin et al. / Journal of Pharmaceutical and Biomedical Analysis 174 (2019) 414–421 [4] G. Liew, Z. Lei, G. Tan, N. Joachim, I.V. Ho, T.Y. Wong, P. Mitchell, B. Gopinath, B. Crossett, Metabolomics of diabetic retinopathy, Curr. Diab. Rep. 17 (11) (2017) 102. [5] M. Mayordomo-Febrer, J.M. López-Murcia, D. Morales-Tatay, M.D. Monleón-Salvado, Pinazo-Durán, Metabolomics of the aqueous humor in the rat glaucoma model induced by a series of intracamerular sodium hyaluronate injection, Exp. Eye Res. 131 (2015) 84–92. [6] C. Barbas-Bernardos, E.G. Armitage, A. García, S. Mérida, A. Navea, F. Bosch-Morell, C. Barbas, Looking into aqueous humor through metabolomics spectacles - exploring its metabolic characteristics in relation to myopia, J. Pharm. Biomed. Anal. 127 (2016) 18–25. [7] J.M. Emery, J.H. Little, Phacoemulsification and Aspiration of Cataracts; Surgical Techniques, Complications, and Results, CV Mosby, St. Louis, 1979, pp. 45–48. [8] C.P. Wilkinson, F.L. Ferris 3rd, R.E. Klein, P.P. Lee, C.D. Agardh, M. Davis, D. Dills, A. Kampik, R. Pararajasegaram, J.T. Verdaguer, Global Diabetic Retinopathy Project Group, Proposed international clinical diabetic retinopathy and diabetic macular edema disease severity scales, Ophthalmology 110 (9) (2003) 1677–1682. [9] O. Cloarec, M.E. Dumas, J. Trygg, A. Craig, R.H. Barton, J.C. Lindon, J.K. Nicholson, E. Holmes, Evaluation of the orthogonal projection on latent structure model limitations caused by chemical shift variability and improved visualization of biomarker changes in 1 H NMR spectroscopic metabonomic studies, Anal. Chem. 77 (2) (2005) 517–526. [10] J. Xia, D.S. Wishart, Using MetaboAnalyst 3.0 for comprehensive metabolomics data analysis, Curr. Protoc. Bioinformatics 55 (2016), 14.10.1-14.10.91. [11] Z. Chen, Y. Zhu, Y. Zhao, X. Ma, M. Niu, J. Wang, H. Su, R. Wang, J. Li, L. Liu, Z. Wei, Q. Zhao, H. Chen, X. Xiao, Serum Metabolomic profiling in a rat model reveals protective function of Paeoniflorin against ANIT induced cholestasis, Phytother. Res. 30 (4) (2016) 654–662. [12] L.B. Gladden, Lactate metabolism: a new paradigm for the third millennium, J. Physiol. 558 (Pt 1) (2004) 5–30. [13] K. Trudeau, A.J. Molina, S. Roy, High glucose induces mitochondrial morphology and metabolic changes in retinal pericytes, Invest. Ophthalmol. Vis. Sci. 52 (12) (2011) 8657–8664. ´ S.Y. Sundier, E.L. Robb, A. [14] E.T. Chouchani, V.R. Pell, E. Gaude, D. Aksentijevic, Logan, S.M. Nadtochiy, E.N.J. Ord, A.C. Smith, F. Eyassu, R. Shirley, C.H. Hu, A.J. Dare, A.M. James, S. Rogatti, R.C. Hartley, S. Eaton, A.S.H. Costa, P.S. Brookes, S.M. Davidson, M.R. Duchen, K. Saeb-Parsy, M.J. Shattock, A.J. Robinson, L.M. Work, C. Frezza, T. Krieg, M.P. Murphy, Ischaemic accumulation of succinate controls reperfusion injury through mitochondrial ROS, Nature 515 (7527) (2014) 431–435. [15] M. Matsumoto, K. Suzuma, T. Maki, H. Kinoshita, E. Tsuiki, A. Fujikawa, T. Kitaoka, Succinate increases in the vitreous fluid of patients with active proliferative diabetic retinopathy, Am. J. Ophthalmol. 153 (5) (2012) 896–902. [16] K. Kooragayala, N. Gotoh, T. Cogliati, J. Nellissery, T.R. Kaden, S. French, R. Balaban, W. Li, R. Covian, A. Swaroop, Quantification of oxygen consumption in retina ex vivo demonstrates limited reserve capacity of photoreceptor mitochondria, Invest. Ophthalmol. Vis. Sci. 56 (13) (2015) 8428–8436.

421

[17] J.R. Nyengaard, Y. Ido, C. Kilo, J.R. Williamson, Interactions between hyperglycemia and hypoxia: implications for diabetic retinopathy, Diabetes 53 (11) (2004) 2931–2938. [18] S.S. Zhu, Y. Ren, M. Zhang, J.Q. Cao, Q. Yang, X.Y. Li, H. Bai, L. Jiang, Q. Jiang, Z.G. He, Q. Chen, Wld(S) protects against peripheral neuropathy and retinopathy in an experimental model of diabetes in mice, Diabetologia 54 (9) (2011) 2440–2450. [19] J. Bełtowski, A. Kedra, Asymmetric dimethylarginine (ADMA) as a target for pharmacotherapy, Pharmacol. Rep. 58 (2) (2006) 159–178. [20] M.R. Du, L. Yan, N.S. Li, Y.J. Wang, T. Zhou, J.L. Jiang, Asymmetric dimethylarginine contributes to retinal neovascularization of diabetic retinopathy through EphrinB2 pathway, Vascul. Pharmacol. 108 (2018) 46–56. [21] S. Abhary, N. Kasmeridis, K.P. Burdon, A. Kuot, M.J. Whiting, W.P. Yew, N. Petrovsky, J.E. Craig, Diabetic retinopathy is associated with elevated serum asymmetric and symmetric dimethylarginines, Diabetes Care 32 (11) (2009) 2084–2086. [22] K. Sydow, T. Münzel, ADMA and oxidative stress, Atheroscler. Suppl. 4 (4) (2003) 41–51. [23] L. Stankova, M. Riddle, J. Larned, K. Burry, D. Menashe, J. Hart, R. Bigley, Plasma ascorbate concentrations and blood cell dehydroascorbate transport in patients with diabetes mellitus, Metabolism 33 (4) (1984) 347–353. [24] N.R. Haines, N. Manoharan, J.L. Olson, A. D’Alessandro, J.A. Reisz, Metabolomics analysis of human vitreous in diabetic retinopathy and rhegmatogenous retinal detachment, J. Proteome Res. 17 (7) (2018) 2421–2427. [25] H. Chen, R.J. Karne, G. Hall, U. Campia, J.A. Panza, R.O. Cannon 3rd, Y. Wang, A. Katz, M. Levine, M.J. Quon, High-dose oral vitamin C partially replenishes vitamin C levels in patients with Type 2 diabetes and low vitamin C levels but does not improve endothelial dysfunction or insulin resistance, Am. J. Physiol. Heart Circ. Physiol. 290 (1) (2006) H137–H145. [26] J.M. May, Ascorbic acid repletion: A possible therapy for diabetic macular edema? Free Radic. Biol. Med. 94 (2016) 47–54. [27] S. Landaas, The formation of 2-hydroxybutyric acid in experimental animals, Clin. Chim. Acta 58 (1) (1975) 23–32. [28] W.E. Gall, K. Beebe, K.A. Lawton, K.P. Adam, M.W. Mitchell, P.J. Nakhle, J.A. Ryals, M.V. Milburn, M. Nannipieri, S. Camastra, A.V. Natali, E. Ferrannini, RISC Study Group, Alpha-hydroxybutyrate is an early biomarker of insulin resistance and glucose intolerance in a nondiabetic population, PLoS One 5 (5) (2010) e10883. [29] M.J. Rist, A. Roth, L. Frommherz, C.H. Weinert, R. Krüger, B. Merz, D. Bunzel, C. Mack, B. Egert, A. Bub, B. Görling, P. Tzvetkova, B. Luy, I. Hoffmann, S.E. Kulling, B. Watzl, Metabolite patterns predicting sex and age in participants of the Karlsruhe Metabolomics and Nutrition (KarMeN) study, PLoS One 12 (8) (2017), e0183228. [30] S. Trabado, A. Al-Salameh, V. Croixmarie, P. Masson, E. Corruble, B. Fève, R. Colle, L. Ripoll, B. Walther, C. Boursier-Neyret, E. Werner, L. Becquemont, P. Chanson, The human plasma-metabolome: reference values in 800 French healthy volunteers; impact of cholesterol, gender and age, PLoS One 12 (3) (2017), e0173615.