The clinical heterogeneity of preeclampsia is related to both placental gene expression and placental histopathology

Original Research

ajog.org

OBSTETRICS

The clinical heterogeneity of preeclampsia is related to both placental gene expression and placental histopathology Samantha J. Benton, PhD1; Katherine Leavey, PhD1; David Grynspan, MD; Brian J. Cox, PhD2; Shannon A. Bainbridge, PhD2

BACKGROUND: Preeclampsia is a life-threatening disorder of pregnancy, demonstrating a high degree of heterogeneity in clinical features such as presentation, disease severity, and outcomes. This heterogeneity suggests distinct pathophysiological mechanisms may be driving the placental disease underlying this disorder. Our group recently reported distinct clusters of placental gene expression in preeclampsia and control pregnancies, allowing for the identification of at least 3 clinically relevant gene expression-based subtypes of preeclampsia. Histopathological examination of a small number of samples from 2 of the gene expressionebased subtypes revealed placental lesions consistent with their gene expression phenotype, suggesting that detailed placental histopathology may provide further insight into the pathophysiology underlying these distinct gene expression-based subtypes. OBJECTIVES: The objective of the study was to assess histopathological lesions in the placentas of patients belonging to each identified gene expressionebased subtype of preeclampsia, characterized in our previous study. Our goal was to further understand the pathophysiologies defining these gene expressionebased subtypes by integrating gene expression with histopathological findings, possibly identifying additional subgroups of preeclampsia patients. STUDY DESIGN: Paraffin-embedded placental biopsies from patients included in the gene expression profiling study (n ¼ 142 of 157, 90.4%) were sectioned, hematoxylin and eosin stained, and imaged. An experienced perinatal pathologist, blinded to gene expression findings and clinical information, assessed the presence and severity of histological lesions using a comprehensive, standardized data collection form. The frequency and severity scores of observed histopathological lesions were compared among gene expressionebased subtypes as well as within each subtype using using Fisher exact tests, Kruskal-Wallis tests, and hierarchical clustering. The histological

findings of the placental samples were visualized using t-distributed stochastic neighbor embedding and phylogenetic trees. Concordance and discordance between gene expression findings and histopathology were also investigated and visualized using principal component analysis. RESULTS: Several histological lesions were found to be characteristic of each gene expressionebased preeclampsia subtype. The overall concordance between gene expression and histopathology for all samples was 65% (93 of 142), with characteristic placental lesions for each gene expressionebased subtype complementing prior gene enrichment findings (ie, placentas with enrichment of hypoxia-associated genes showed severe lesions of maternal vascular malperfusion). Concordant samples were located in the central area of each gene expressionebased cluster when viewed on a principal component analysis plot. Interestingly, discordant samples (gene expression and histopathology not reflective of one another) were generally found to lie at the periphery of the gene expressionebased clusters and tended to border the group of patients with phenotypically similar histopathology. CONCLUSION: Our findings demonstrates a high degree of concordance between placental lesions and gene expression across subtypes of preeclampsia. Additionally, novel integrative analysis of scored placental histopathology severity and gene expression findings allowed for the identification of patients with intermediate phenotypes of preeclampsia not apparent through gene expression profiling alone. Future investigations should examine the temporal relationship between these 2 modalities as well as consider the maternal and fetal contributions to these subtypes of disease. Key words: disease subtypes, gene expression, placenta pathology,

preeclampsia

P

reeclampsia (PE) is a life-threatening disorder of pregnancy, characterized by hypertension and end-organ dysfunction.1 Affecting 3e8% of pregnancies worldwide, PE is responsible for >76,000 maternal deaths annually.2,3

Cite this article as: Benton SJ, Leavey K, Grynspan D, et al. The clinical heterogeneity of preeclampsia is related to both placental gene expression and placental histopathology. Am J Obstet Gynecol 2018;219:604.e1-25. 0002-9378/free ª 2018 Published by Elsevier Inc. https://doi.org/10.1016/j.ajog.2018.09.036

Maternal complications from PE include eclampsia, pulmonary edema, and liver rupture that can permanently alter a mother’s life during and after her pregnancy.4,5 Moreover, it is increasingly recognized that PE is a significant risk factor for premature cardiovascular disease in these women.6e8 Fetal well-being is also significantly affected by PE, increasing the risk of stillbirth, fetal growth restriction, and iatrogenic preterm birth.3,9 Interventions to prevent and treat PE remain

604.e1 American Journal of Obstetrics & Gynecology DECEMBER 2018

largely elusive, with the delivery of the placenta being the only definitive cure for the disorder, despite necessitating preterm delivery of the fetus in some cases.4,10 Causes of PE are multifactorial in nature with genetic, environmental, and maternal constitutional contributions that have yet to be fully elucidated.11 While it is believed that placental dysfunction underlies the majority of PE cases,12 heterogeneity in clinical presentation and outcomes, disease severity, and placental pathology between women with PE suggest that multiple forms of placental disease in PE exist.13 For

ajog.org

OBSTETRICS

AJOG at a Glance Why was the study conducted? The high degree of patient heterogeneity observed in cases of preeclampsia has hindered our ability to identify and develop highly effective therapeutic interventions for this disorder. It is speculated that this clinical heterogeneity may speak to the presence of several preeclampsia disease subtypes, a hypothesis strengthened by our recent identification of at least 3 distinct placental gene expression profiles in preeclampsia. To date, a comprehensive characterization of the placental disease underlying each of these preeclampsia subclasses has yet to be undertaken. Key Findings The integration of placental histopathology findings with prior placental gene expression and clinical profiles confirmed the presence of 3 clinically relevant subtypes of preeclampsia with divergent underlying pathophysiology. The contextually rich data set further identified gradients of disease severity within each subtype as well as cases with mixed forms of placental disease. What does this add to what is known? This body of work has identified distinct subtypes of placental disease in preeclampsia, confirmed at both the molecular and anatomical level, of which at least 2 are not currently well defined in the literature. These findings help explain the high degree of heterogeneity observed in the preeclampsia patient population and highlight relevant avenues for discovery of effective therapeutic interventions for this disorder. example, aspirin is most effective in reducing the recurrence of preterm, but not term, PE and reduces the risk of PE in women with risk factors for the disease.14e16 Likewise, common pathological lesions such as maternal vascular malperfusion are typically seen in preterm PE but have been reported in severe disease at term.17e19 Additionally, there remains a significant proportion of cases that have minimal evidence of overt placental pathology.18,20,21 Recent hypotheses and accumulating evidence suggest that PE is likely a disorder encompassing several distinct subtypes of disease, beyond the traditional separation of early- vs lateonset PE.22e25 Efforts aimed at concretely identifying, understanding, and characterizing these subtypes would not only advance our understanding of this heterogeneous disorder but also offer the opportunity to provide etiology-based interventions, therapeutics, and management in a personalized medicine approach. In an effort to uncover possible subtypes of PE, we recently used genomewide microarray analysis to obtain gene

expressionebased profiles of 157 placentas from PE and control pregnancies.25 Unsupervised clustering identified 5 distinct clusters of placental gene expression, including three clinically significant subtypes of PE: maternal PE patients with placental gene expression profiles similar to women with healthy pregnancies (cluster 1/PE subtype 1); canonical PE patients with high placental expression of known PE markers (ie, fms-like tyrosine kinase-1, endoglin) and genes related to hypoxia and altered hormone secretion (cluster 2/ PE subtype 2); and immunological PE patients exhibiting an overrepresentation of immune and proinflammatory genes (cluster 3/PE subtype 3). The control pregnancies primarily split into 2 gene expressionebased clusters, with the healthy term controls clustering alongside the maternal PE patients (cluster 1) and the preterm controls samples (delivery <34 weeks) forming a unique cluster defined by the overrepresentation of genes related to cell proliferation and stress response (cluster 4). An additional fifth group of PE patients was also discovered in this

Original Research

clustering analysis, with no strong clinical, gene enrichment, or epigenetic26 associations (cluster 5). Further investigation determined this cluster to be the likely result of chromosomal abnormalities related to confined placental mosaicism and not necessarily a distinct gene expressionebased subtype of PE. While clustering was successful in identifying gene expressionebased subtypes of PE, this particular method of analysis allows only for a strict singular classification of each patient that may omit cases of mixed pathophysiology and/or gradients of disease severity. Visualization of the placental gene expression showed that some patient samples were located at the border of 2 neighboring clusters suggesting the following: (1) in some women, multiple pathophysiologies (eg, hypoxia and immune deregulation) may be contributing to the development of PE; and/or (2) placental gene expression profiling on its own was insufficient to properly identify all possible subtypes of PE pathophysiology. In the present study, we sought to further characterize these gene expressionebased subtypes of PE using detailed placental histopathology in an effort to better understand the pathophysiologies underlying our PE subtypes and to test our hypothesis that detailed placental histopathology would reflect gene expression changes associated with each of our previously identified PE subtypes. By investigating the relationship between changes in gene expression and structural modifications in the placenta, with inferred functional alterations, we can gain a better understanding of the damage and pathology in the placentas of specific groups of patients. In addition to offering complementary insight into the underlying placental pathologies observed across the gene expressionebased PE subtypes, the additional contextual information provided through comprehensive placental histopathology offers the possibility of identifying smaller, subtler groups of PE patients, thereby refining the paradigm of divergent underlying placenta pathophysiology in PE.

DECEMBER 2018 American Journal of Obstetrics & Gynecology

604.e2

Original Research

OBSTETRICS

Materials and Methods In this study, we utilized banked placental tissue from the same patients (both PE and controls) included in our gene expression study.25 Detailed histopathological assessment was performed on placental tissue from patients included in our previous study using a standardized data collection form, and previous gene expression data were integrated with histopathology findings for analyses (described in the following text). A Glossary of Terms for terminology used in the following sections is provided in Supplemental Figure 1.

Patient cohort Details pertaining to the patient cohort used in this study have been previously described.25 Briefly, 157 women with singleton normotensive pregnancies (n ¼ 53), pregnancies with chronic hypertension (n ¼ 24), or pregnancies with PE (n ¼ 80) were selected from the Research Centre for Women’s and Infants’ Health (RCWIH) BioBank (Toronto, Canada). PE was defined as the onset of hypertension (systolic pressure 140 mm Hg and/or diastolic pressure 90 mm Hg) after 20 weeks’ gestation with proteinuria (>300 mg protein/d or 2þ by dipstick) according to diagnostic guidelines at the time of the original study.27,28 Chronic maternal hypertension was defined as systolic pressure 140 mm Hg and/or sustained diastolic 90 mm Hg before 20 weeks’ gestation. Appropriate for gestational age was defined as a neonatal birthweight >10th percentile for gestational age and sex,29 and small for gestational age (SGA) was defined as a neonatal birthweight <10th percentile for gestational age and sex.29 Estimated fetal weight, umbilical artery pulsatility index, and middle cerebral artery pulsatility index were converted to a percentile for gestational age at testing, according to respective reference charts.30e32 Women with diabetes (preexisting or gestational), sickle cell anemia, and/or morbid obesity (body mass index 40 kg/m2) were excluded. Clinical data for each woman were

collected by chart review by RCWIH staff. Placental tissue biopsies from each patient were purchased from the RCWIH Biobank. Placenta sampling procedures and gene expression analysis methods used to determine gene expressionebased subtype membership of each placenta have been previously described.25 Briefly, a total of 4 placental core biopsies (1.5 cm  1.5 cm) were randomly excised from representative sites of the placental disc (1 biopsy per quadrant, full thickness excluding the chorionic plate). Areas of gross pathology were avoided. Each core biopsy was rinsed in phosphate-buffered saline to remove maternal blood and cut into smaller pieces. Tissue samples from all 4 sites were pooled and transferred to vials for snap freezing in liquid nitrogen or fixing in paraformaldehyde. From snap-frozen tissue, messenger RNA was extracted and hybridized against Human Gene 1.0 ST array chips from Affymetrix (Affymetrix, Santa Clara, CA). The microarray data (Gene Expression Omnibus database, GSE75010) was processed, normalized, and converted into log2 values in R33 and underwent unsupervised multivariate clustering, identifying 5 unique gene expressionebased subtypes of placental gene expression using the Bayesian information criterion.

Placenta histopathology Wax-embedded placental tissue biopsies and corresponding placental pathology reports from each patient included in the previous gene expression study were purchased through the RCWIH BioBank. In total, 142 patients of the original 157 microarray patients had placental histopathology samples available. Paraformaldehyde-fixed, wax-embedded tissue from the pooled core biopsies randomly selected from each quadrant of the placenta (as described in the previous text) was sectioned (5 mm thick) and stained with hematoxylin and eosin using standard laboratory protocol.34 High-resolution digital images of the stained sections were taken using the

604.e3 American Journal of Obstetrics & Gynecology DECEMBER 2018

ajog.org Aperio ScanScope C2 microscope (Leica; Concord, Ontario, Canada). A single experienced perinatal pathologist, blinded to gene expression results and clinical outcomes (excluding gestational age at delivery), examined the digital images and graded histological lesions using a standardized comprehensive data collection form25,35,36 with prespecified severity criteria derived from clinical practice guidelines and published literature37e44 (Supplemental Table 1). To achieve a quantitative output for the severity of pathology in each placenta, a severity scoring system was derived from published consensus statements and practice guidelines and included in the data collection form. For lesions with a binary designation (ie, present or absent), a score of 0 was assigned when the lesion was absent or a score of 1 was assigned if it was present. For lesions with a grading severity scheme, the scoring system followed the grading scale in linear fashion. For example, if the lesion is graded as focal, patchy, or diffuse, we assigned a score of 1 for focal, 2 for patchy, or 3 for diffuse. A score of 0 was again assigned if the lesion was not present. Individual lesions were grouped according to broad etiological categories, and a maximum severity score for each category was calculated by adding up the highest score possible for each lesion belonging to that category. Lesion categories and maximum severity score for each category are as follows: maternal vascular malperfusion (maximum score of 13), implantation site abnormalities (max score 4), histological chorioamnionitis (maximum score of 11), placental villous maldevelopment (maximum score of 5), fetal vascular malperfusion (maximum score of 6), chronic uteroplacental separation (maximum score of 3), maternal-fetal interface disturbance (maximum score of 5), and chronic inflammation (maximum score of 6). Gross anatomy (e.g., placental weight, umbilical cord length) was obtained from the corresponding historical placental pathology reports, in addition to several microscopic lesions (eg, placental infarction, chronic deciduitis),

ajog.org

TABLE 1

Clinical characteristics of the gene expressionebased clusters, by maternal hypertensive status Cluster 1 (n ¼ 52)

Cluster 2 (n ¼ 52)

HTN (n ¼ 21)

Normo (n ¼ 31)

HTN (n ¼ 51)

31 (4)

33 (5)

34 (6)

57 (12/21)

42 (13/31)

69 (35/51)

Prepregnancy BMI, kg/m

26 (8)

25 (5)

27 (4)

White ethnicity, %

57 (12/21)

68 (21/31)

47 (24/51)

Previous hypertensive pregnancy, %

75 (6/8)

6 (1/16)

57 (8/14)

Characteristic baseline Maternal age, y Nulliparous, % 2

Cluster 3 (n ¼ 11)

Normoa (n ¼ 1) 30 100 (1/1) 22 0 (0/1)

HTN (n ¼ 10) 36 (4)

Cluster 4 (n ¼ 13)

Cluster 5 (n ¼ 14)

Normoa (n ¼ 1)

HTNa (n ¼ 1)

Normo (n ¼ 12)

HTN (n ¼ 10)

39

21

30 (6)

36 (5)

31 (7)

1.00

58 (7/12)

30 (3/10)

25 (1/4)

1.00

26 (6)

24 (4)

24 (3)

1.00

67 (8/12)

70 (7/10)

50 (2/4)

1.00

0 (0/3)

75 (3/4)

0 (0/2)

.03

40 (4/10) 26 (5)

0 (0/1) 19

0 (0/1) 22

30 (3/10)

0 (0/1)

0 (0/1)

—

33 (2/6)

0 (0/1)

100 (1/1)

9 (21)

Normo (n ¼ 4)

P valueb

Fetal and placental assessmentc —

41 (35)

10 (18)

73 (28)

< .01

95 (8)

—

—

51 (28)

93 (13)

72 (29)

.01

50 (21/42) —

50 (4/8)

—

—

0 (0/8)

20 (1/5)

0 (0/2)

.29

20 (23)

—

—

65 (25)

13 (13)

25 (24)

50 (20)

8 (13)

—

Umbilical artery PI percentile

79 (26)

58 (32)

88 (22)

—

Absent or reversed EDV, %

31 (4/13)

MCA PI percentile

0 (0/8)

35 (36)

29 (11)

14 (16)

sBP, mm Hg

159 (18)

114 (14)

162 (22)

100

160 (15)

100

158

118 (10)

158 (26)

108 (18)

< .01

dBP, mm Hg

99 (12)

70 (11)

103 (13)

70

96 (11)

60

98

72 (12)

97 (11)

62 (3)

< .01

14 (1/7)

50 (4/8)

Maternal assessment and diagnosis

Proteinuria, %

67 (10/15) 9

—

—

0 (0/5)

73 (35/48) —

67 (6/9)

—

100 (1/1)

0 (0/2)

190 (57)

215 (47)

191 (61)

—

183 (60)

—

232

238 (34)

208 (34)

342 (78)

301 (73)

402 (90)

—

368 (71)

—

394

258

365 (60)

—

36 (43)

20 (2)

36 (31)

—

32 (28)

—

16

10

16 (5)

—

d

d

.03

266 (4)

1.00 1.00 .09 d

< .01

Antihypertensive treatment, %

67 (14/21)

6 (2/31)

86 (44/51)

0 (0/1)

70 (7/10)

0 (0/1)

100 (1/1)

17 (2/12)

70 (7/10)

25 (1/4)

Preeclampsia. %

67 (14/21)

0 (0/31)

92 (47/51)

0 (0/1)

80 (8/10)

0 (0/1)

100 (1/1)

0 (0/12)

50 (5/10)

0 (0/4)

< .01

Chronic hypertension, %

43 (9/21)

0 (0/31)

29 (15/51)

0 (0/1)

40 (4/10)

0 (0/1)

0 (0/1)

0 (0/12)

60 (6/10)

0 (0/4)

< .01

Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

(continued)

604.e4

Original Research

Platelets ( 10 /L) Uric acid, mM AST, mM

.53

c

OBSTETRICS

DECEMBER 2018 American Journal of Obstetrics & Gynecology

—

EFW percentile

604.e5 American Journal of Obstetrics & Gynecology DECEMBER 2018

Placental weight percentile

47 (37)

13 (4/31)

0 (0/29)

0 (0/31)

6 (2/31)

50 (23)

2899 (900)

5 (17)

55 (28/51)

2 (1/42)

29 (15/51)

67 (34/51)

11 (12)

1247 (564)

55 (28/51)

82 (42/51)

94 (48/51)

31 (3)

HTN (n ¼ 51)

—

0 (0/1)

0 (0/1)

0 (0/1)

0 (0/1)

34

2870

100 (1/1)

0 (0/1)

0 (0/1)

37

Normoa (n ¼ 1)

Cluster 2 (n ¼ 52)

3 (5)

40 (4/10)

0 (0/8)

70 (7/10)

70 (7/10)

8 (15)

1634 (785)

40 (4/10)

40 (4/10)

80 (8/10)

34 (4)

24

0 (0/1)

0 (0/1)

0 (0/1)

0 (0/1)

41

3280

100 (1/1)

0 (0/1)

0 (0/1)

39

Normoa (n ¼ 1)

Cluster 3 (n ¼ 11) HTN (n ¼ 10)

56

0 (0/1)

0 (0/1)

0 (0/1)

0 (0/1)

14

2600

0 (0/1)

0 (0/1)

0 (0/1)

37

67 (39)

42 (5/12)

0 (0/7)

0 (0/12)

8 (1/12)

57 (27)

1422 (707)

58 (7/12)

92 (11/12)

100 (12/12)

29 (3)

Normo (n ¼ 12)

Cluster 4 (n ¼ 13) HTNa (n ¼ 1)

16 (33)

40 (4/10)

20 (2/10)

40 (4/10)

70 (7/10)

9 (13)

1639 (584)

30 (3/10)

40 (4/10)

80 (8/10)

34 (4)

< .01 < .01 1.00

50 (2/4) 50 (2/4) 75 (3/4)

< .01 < .01 < .01 1.00 .24 < .01

58 (13) 0 (0/4) 0 (0/4) 0 (0/4) 25 (1/4) 60 (28)

2468 (1234) < .01

< .01

P valueb

34 (6)

Normo (n ¼ 4)

Cluster 5 (n ¼ 14) HTN (n ¼ 10)

Not included in the statistical analysis; b Calculated using Fisher exact tests for categorical variables and Kruskal-Wallis tests for continuous variables, followed by Bonferroni correction for multiple comparison testing; c Last value taken within 4 weeks of delivery; d Some normotensive patients received nifedipine as a tocolytic. Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

a

AST, aspartate transaminase; BMI, body mass index; dBP, diastolic blood pressure; EDV, end diastolic velocity; EFW, estimated fetal weight; HTN, hypertensive; MCA, middle cerebral artery; NICU, neonatal intensive care unit; Normo, normotensive; PI, pulsatility index; sBP, systolic blood pressure; SGA, small for gestational age.

Continuous variables are expressed as mean (SD) and discrete variables are percentages (n/N).

29 (6/21) 27 (38)

NICU transfer, %

14 (3/21)

SGA <3rd, % 0 (0/20)

33 (7/21)

SGA <10th, %

Apgar <7 at 5 min, %

24 (20)

Birthweight percentile

Birthweight, g

2247 (764)

32 (10/31) 55 (17/31)

29 (6/21) 48 (10/21)

Delivery <34 wks, %

Male infant, %

35 (11/31)

37 (4)

62 (13/21)

35 (3)

Normo (n ¼ 31)

Delivery <37 wks, %

GA at delivery, wks

Pregnancy outcome

Characteristic baseline

HTN (n ¼ 21)

Cluster 1 (n ¼ 52)

Clinical characteristics of the gene expressionebased clusters, by maternal hypertensive status (continued)

TABLE 1

Original Research OBSTETRICS

ajog.org

because the tissue biopsies were collected from areas that appeared grossly normal and included only villous parenchyma (ie, maternal decidua was not sampled).

Visualization and clustering analysis of histopathology data

The placental lesion severity scores in the 142 samples were loaded into R 3.1.3. For each individual lesion, a mean severity score was calculated for the cohort as a whole and for each of the 5 gene expression clusters. For each etiological category (9 categories in all), a mean severity score was calculated based on the sum of severity scores for all lesions belonging to the category. Finally, a total pathology severity score was calculated for the whole cohort and each cluster by summing the lesion severity scores for all etiological categories. The global relationships between the 142 placentas based on the histology information alone were visualized using t-distributed stochastic neighbor embedding (t-SNE)45 with a perplexity of 13 (an estimate of each sample’s number of close neighbors). t-SNE is a reduction method for visualizing complex data sets. It uses nonlinear transformation to map high-dimension data onto 2e3 dimensions while simultaneously preserving both the local and global structures of the original data. Because the histology data were not normally distributed, t-SNE was utilized for visualization instead of principal component analysis (PCA; see the following text). Histology scores for the samples belonging to each individual gene expressionebased subtype were also subjected to hierarchical agglomerative clustering, which treats each sample as their own cluster and then, based on distance metrics, successively merges pairs of similar clusters until the entire cohort is one cluster, forming a hierarchical tree. These results were visualized as phylogenetic trees, using the ape package.46

Concordance between placental histopathology and gene expression findings

To determine the degree of concordance between the previous gene expression profiles and histopathology across our

ajog.org

OBSTETRICS

Original Research

TABLE 2

Frequency of observed histological placental lesions, by gene expressionebased cluster Cluster 1 (n ¼ 52)

Cluster 2 (n ¼ 52)

Cluster 3 (n ¼ 11)

Cluster 4 (n ¼ 13)

Cluster 5 (n ¼ 14)

12 (23)

43 (83)

4 (36)

0 (0)

7 (50)

< .01

9 (17)

36 (69)

3 (27)

2 (15)

8 (57)

< .01

Accelerated villous maturity

17 (33)

45 (87)

5 (45)

1 (8)

7 (50)

< .01

Syncytial knots

Histopathology lesion

P value

Maternal vascular malperfusion, n, % Distal villous hypoplasia Placental infarctions

21 (40)

46 (88)

5 (45)

1 (8)

8 (57)

< .01

Focal perivillous fibrin

5 (10)

13 (25)

5 (45)

1 (8)

3 (21)

.03

Villous agglutination

0 (0)

3 (6)

0 (0)

0 (0)

0 (0)

.48

Decidual vasculopathy

3 (6)

7 (13)

2 (18)

0 (0)

2 (14)

.30

Microscopic accreta

0 (0)

0 (0)

0 (0)

0 (0)

1 (7)

.27

Increased basement membrane fibrin

0 (0)

1 (2)

0 (0)

0 (0)

0 (0)

1.00

Implantation site abnormalities

Histological chorioamnionitis Maternal inflammation

4 (8)

0 (0)

0 (0)

9 (69)

1 (7)

< .01

Fetal inflammation

4 (8)

0 (0)

0 (0)

7 (54)

1 (7)

< .01

Vessel thrombosis

0 (0)

0 (0)

0 (0)

1 (8)

0 (0)

.17

Placenta villous maldevelopment Charangoists

1 (2)

1 (2)

1 (9)

0 (0)

1 (7)

.40

Chorangiomas

0 (0)

2 (4)

0 (0)

0 (0)

0 (0)

.73

Delayed villous maturity

8 (15)

1 (2)

5 (45)

4 (31)

1 (7)

< .01

0 (0)

0 (0)

2 (18)

2 (15)

0 (0)

< .01

Fetal vascular malperfusion Avascular fibrotic villi Thrombosis

1 (2)

5 (10)

1 (9)

1 (8)

0 (0)

Intramural fibrin deposition

0 (0)

0 (0)

0 (0)

0 (0)

0 (0)

.27

Chorionic hemosiderosis

2 (4)

1 (2)

1 (9)

0 (0)

1 (7)

0.41

Retroplacental hematoma

1 (2)

5 (10)

2 (18)

0 (0)

0 (0)

0.12

Laminar necrosis

1 (2)

3 (6)

0 (0)

0 (0)

0 (0)

0.89

Massive perivillous fibrin deposition

0 (0)

1 (2)

3 (27)

1 (8)

0 (0)

<0.01

Maternal floor infarction pattern

0 (0)

0 (0)

1 (9)

0 (0)

0 (0)

0.08

Intervillous thrombi

9 (17)

7 (13)

4 (36)

1 (8)

2 (14)

0.43

Infectious villitis

0 (0)

0 (0)

0 (0)

0 (0)

0 (0)

Villitis of unknown etiology

5 (10)

1 (2)

1 (9)

1 (8)

1 (7)

0.31

Chronic intervillositis

0 (0)

1 (2)

1 (9)

0 (0)

0 (0)

0.18

Chronic deciduitis

6 (12)

5 (10)

2 (18)

2 (15)

0 (0)

0.56

Meconium histiocytes/macrophages within membranes

8 (15)

1 (2)

1 (9)

0 (0)

1 (7)

0.09

Meconium-induced myonecrosis

0 (0)

1 (2)

0 (0)

0 (0)

0 (0)

1.00

—

Chronic uteroplacental separation

Maternal-fetal interface disturbance

Chronic inflammation –

Additional features

Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

DECEMBER 2018 American Journal of Obstetrics & Gynecology

604.e6

Pathology severity for the total patient cohort and across each gene expressionebased cluster by individual lesion, etiological category, and total pathology score Total cohort (n ¼ 142)

Histopathology lesion Maternal vascular malperfusion

Cluster 1 (n ¼ 52)

Cluster 2 (n ¼ 52)

Cluster 3 (n ¼ 11)

Cluster 4 (n ¼ 13)

Cluster 5 (n ¼ 14)

P value

Mean (SD)

Distal villous hypoplasia

(0, 2)

0.64 (0.77)

0.25 (0.48)

1.23 (0.73)

0.36 (0.50)

0 (0)

0.71 (0.83)

< .01

Placental infarctions

(0, 2)

0.56 (0.75)

0.19 (0.44)

1.02 (0.80)

0.27 (0.47)

0.23 (0.60)

0.79 (0.80)

< .01

Accelerated villous maturity

(0, 2)

0.53 (0.50)

0.33 (0.47)

0.87 (0.34)

0.45 (0.52)

0.08 (0.28)

0.50 (0.52)

< .01

Syncytial knots

(0, 2)

0.70 (0.68)

0.44 (0.57)

1.15 (0.61)

0.55 (0.69)

0.08 (0.28)

0.64 (0.63)

< .01

Focal perivillous fibrin

(0, 1)

0.20 (0.44)

0.10 (0.30)

0.25 (0.44)

0.64 (0.81)

0.08 (0.28)

0.21 (0.43)

.02

Villous agglutination

(0, 3)

0.02 (0.14)

0 (0)

0.06 (0.24)

0 (0)

0 (0)

0 (0)

.26

Decidual vasculopathy

(0, 1)

0.10 (0.30)

0.06 (0.24)

0.13 (0.34)

0.18 (0.40)

0 (0)

0.14 (0.36)

.37

Category severity score

(0, 13)

2.75 (2.39)

1.37 (1.66)

4.71 (1.90)

2.45 (1.51)

0.46 (0.78)

3.00 (2.48)

< .01

OBSTETRICS

Score range (minimum, maximum)

Original Research

604.e7 American Journal of Obstetrics & Gynecology DECEMBER 2018

TABLE 3

Implantation site abnormalities Microscopic accreta

(0, 2)

0.01 (0.08)

0 (0)

0 (0)

0 (0)

0 (0)

0.07 (0.27)

.06

Increased basement membrane fibrin

(0, 2)

0.01 (0.08)

0 (0)

0.02 (0.14)

0 (0)

0 (0)

0 (0)

.79

Category severity score

(0, 4)

0.01 (0.12)

0 (0)

0.02 (0.14)

0 (0)

0 (0)

0.07 (0.27)

.35

Maternal inflammation

(0, 5)

0.20 (0.64)

0.13 (0.53)

0 (0)

0 (0)

1.46 (1.20)

0.14 (0.53)

< .01

Fetal inflammation

(0, 5)

0.09 (0.31)

0.08 (0.27)

0 (0)

0 (0)

0.62 (0.65)

0.08 (0.27)

< .01

Vessel thrombosis

(0, 1)

0.01 (0.08)

0 (0)

0 (0)

0 (0)

0.08 (0.28)

0 (0)

Category severity score

(0, 11)

0.30 (0.94)

0.21 (0.77)

0 (0)

0 (0)

2.15 (1.72)

0.21 (0.80)

< .01

Charangoists

(0, 1)

0.03 (0.17)

0.02 (0.14)

0.02 (0.14)

0.09 (0.30)

0 (0)

0.07 (0.27)

.53

Chorangiomas

(0, 2)

0.01 (0.12)

0 (0)

0.04 (0.19)

0 (0)

0 (0)

0 (0)

.48

Delayed villous maturity

(0, 2)

0.14 (0.37)

0.15 (0.36)

0.02 (0.14)

0.45 (0.52)

0.38 (0.65)

0.07 (0.27)

< .01

Category severity score

(0, 5)

0.18 (0.42)

0.17 (0.38)

0.08 (0.27)

0.55 (0.69)

0.38 (0.65)

0.14 (0.36)

.02

Avascular fibrotic villi

(0, 3)

0.03 (0.17)

0 (0)

0 (0)

0.18 (0.40)

0.15 (0.38)

0 (0)

< .01

Thrombosis

(0, 1)

0.06 (0.23)

0.02 (0.14)

0.10 (0.30)

0.09 (0.30)

0.08 (0.28)

0 (0)

.40

Intramural fibrin deposition

(0, 2)

0 (0)

0 (0)

0 (0)

0 (0)

0 (0)

0 (0)

Category severity score

(0, 6)

0.08 (0.28)

0.02 (0.14)

0.10 (0.30)

0.27 (0.47)

0.23 (0.44)

0 (0)

Histological chorioamnionitis

.04

Placenta villous maldevelopment

Fetal vascular malperfusion

.01 (continued)

ajog.org

Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

—

ajog.org

TABLE 3

Pathology severity for the total patient cohort and across each gene expressionebased cluster by individual lesion, etiological category, and total pathology score (continued) Histopathology lesion

Total cohort (n ¼ 142)

Cluster 1 (n ¼ 52)

Cluster 2 (n ¼ 52)

Cluster 3 (n ¼ 11)

Cluster 4 (n ¼ 13)

Cluster 5 (n ¼ 14)

P value

Chronic utero-placental separation Chorionic hemosiderosis

(0, 1)

0.04 (0.18)

0.04 (0.19)

0.02 (0.14)

0.09 (0.30)

0 (0)

0.07 (0.27)

.66

Retroplacental hematoma

(0, 1)

0.06 (0.23)

0.02 (0.14)

0.10 (0.30)

0.18 (0.40)

0 (0)

0 (0)

.10

Laminar necrosis

(0, 1)

0.03 (0.17)

0.02 (0.14)

0.06 (0.24)

0 (0)

0 (0)

0 (0)

.58

Category severity score

(0, 3)

0.12 (0.39)

0.08 (0.27)

0.17 (0.51)

0.27 (0.47)

0 (0)

0.07 (0.27)

.21

Massive perivillous fibrin deposition

(0, 2)

0.04 (0.18)

0 (0)

0.02 (0.14)

0.27 (0.47)

0.08 (0.28)

0 (0)

< .01

Maternal floor infarction pattern

(0, 2)

0.01 (0.08)

0 (0)

0 (0)

0.09 (0.30)

0 (0)

0 (0)

.02

Intervillous thrombi

(0, 1)

0.17 (0.39)

0.17 (0.38)

0.13 (0.34)

0.45 (0.69)

0.08 (0.28)

0.14 (0.36)

.31

Category severity score

(0, 5)

0.21 (0.50)

0.17 (0.38)

0.15 (0.41)

0.82 (0.98)

0.15 (0.55)

0.14 (0.36)

.01

Infectious villitis

(0, 1)

0 (0)

0 (0)

0 (0)

0 (0)

0 (0)

0 (0)

Villitis of unknown etiology

(0, 2)

0.09 (0.37)

0.15 (0.50)

0.02 (0.14)

0.18 (0.60)

0.08 (0.28)

0.08 (0.27)

.57

Chronic intervillositis

(0, 2)

0.01 (0.12)

0 (0)

0.02 (0.14)

0.09 (0.30)

0 (0)

0 (0)

.21

Chronic deciduitis

(0, 1)

0.11 (0.31)

0.12 (0.32)

0.10 (0.30)

0.18 (0.40)

0.15 (0.38)

0 (0)

.60

Category severity score

(0, 6)

0.21 (0.58)

0.27 (0.69)

0.13 (0.40)

0.45 (0.93)

0.23 (0.60)

0.07 (0.27)

.60

Meconium histiocytes/macrophages within membranes

(0, 1)

0.08 (0.27)

0.15 (0.36)

0.02 (0.14)

0.09 (0.30)

0 (0)

0.07 (0.27)

.10

Meconium-induced myonecrosis

(0, 1)

0.01 (0.08)

0 (0)

0.02 (0.14)

0 (0)

0 (0)

0 (0)

.79

3.96 (2.31)

2.44 (1.83)

5.40 (1.95)

4.91 (2.34)

3.62 (1.45)

3.79 (2.29)

Maternal-fetal interface disturbance

Chronic inflammation —

Total pathology score

< .01

604.e8

Original Research

Maximum severity score by category was as follows: maternal vascular malperfusion (maximum score 13), implantation site abnormalities (maximum score 4), histological chorioamnionitis (maximum score 11), placental villous maldevelopment (maximum score 5), fetal vascular malperfusion (maximum score 6), chronic uteroplacental separation (maximum score 3), maternal-fetal interface disturbance (maximum score 5), and chronic inflammation (maximum score 6). Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

OBSTETRICS

DECEMBER 2018 American Journal of Obstetrics & Gynecology

Additional features

Original Research

OBSTETRICS

sample set, we defined a priori concordance and discordance for each cluster as follows: concordant cluster 1 placentas are those with minimal or no placental pathology, and normal/healthy placental gene expressionebased profiles; concordant cluster 2 placentas are those with a high histopathology score (3 or greater) for lesions of maternal vascular malperfusion, and upregulation of hypoxia-related genes; concordant cluster 3 placentas are those with signs of a maternal-fetal interface disturbance and/or chronic inflammation and enrichment of immune response gene expression profiles; and, concordant cluster 4 samples are those with histological chorioamnionitis and upregulation of genes related to infection including stress response, DNA damage, and inflammation. Samples not meeting the described criteria were classified as discordant. Cluster 5 had no clear defining features (outside the identified chromosomal abnormalities); therefore, all placentas in this cluster were classified as discordant. A PCA plot of the gene expression data from the 142 placenta samples with available histopathology was recolored to demonstrate the gene expressionebased relationships between samples with concordant and discordant placental histological features. Principal component analysis is another dimension reduction method that uses linear transformations of the data and was used in this case for consistency with our prior gene expression study. The goal of PCA is to create a set of new variables, termed principal components (PC) that are independent and composed of mixtures of the original variables. The total number of PC variables is equal to the total number of samples in the data set, and they are ranked from highest variance to lowest. PCA was performed in R using the prcomp function and the first 3 components (PC1-3) were visualized with the plot3d function (rgl library) because these represent the highest proportion of the variation in the data set. To achieve the visual semitransparency of the spheres included in PCA plot, the plot3d alpha was set to 0.2.

A mean value for each of PC1, PC2, and PC3 for each gene expressionebased cluster (ie, the center of the cluster on the 3-dimensional PCA plot) was calculated using the 142 samples with available histopathology. For each sample belonging to a concordant or discordant group of interest, the Euclidean distance between its PC1-3 location on the plot and the center of the cluster of interest was calculated, using the dist function in R. The relative locations of various groups were then compared by Wilcoxon rank sum tests.

Statistics Descriptive data are presented as means with SDs and categorical variables as proportions and counts. Fisher exact tests were used for the comparison of categorical variables across the clusters, and Kruskal-Wallis tests and Wilcoxon rank sum tests were used for continuous variables, followed by Bonferroni correction for multiple comparison testing, as appropriate. The frequency of histological findings as well as severity scores for each individual lesion, each etiological category, and total pathology score were also compared among the clusters using Fisher exact tests and Kruskal-Wallis tests, respectively. c2 goodness-of-fit tests were used to determine whether the degree of gene expressionehistological concordance observed within the entire cohort and each cluster was significant. Values of P < .05 were considered to indicate statistical significance.

Study approval Ethics approval for this study was granted from the Research Ethics Boards of the Ottawa Health Science Network (#2011623-01H), Mount Sinai Hospital (#13-0211-E), and the University of Toronto (#29435). All women provided written informed consent for the collection of biological specimens and clinical data.

Results Patient characteristics Of the 157 samples included in our original microarray study, 142 (90%) had matched placental tissue biopsies available

604.e9 American Journal of Obstetrics & Gynecology DECEMBER 2018

ajog.org for histological assessment. Clinical characteristics of the gene expressionebased clusters, by maternal hypertensive status, are shown in Table 1. Of the clinically relevant PE subtypes (clusters 1e3), hypertensive women in clusters 2 and 3 were more likely to develop PE (Fisher exact test, P ¼ .02), deliver earlier (KruskalWallis test, P <.01), and give birth to SGA neonates (Fisher exact test, P ¼ .03) compared with hypertensive women belonging to cluster 1. As expected, normotensive women in cluster 4 delivered preterm (mean gestational age, 29 weeks). In cluster 5, both normotensive and hypertensive women tended to deliver around the same gestational age (Wilcoxon rank sum test, P ¼ .89); however, there were more SGA neonates in the hypertensive group (Fisher exact test, P ¼ .07).

Placenta pathology findings according to gene expressionebased cluster subtype of preeclampsia The frequency of histological findings (Table 2) and the severity scores of observed lesions (Table 3) were compared across the gene expressionebased clusters. Placentas from cluster 1 demonstrated minimal evidence of placental pathology, with the lowest frequency of observed lesions for each histological category and the lowest total pathology score (mean score, 2.44; Kruskal-Wallis test, P < .01). The most severe placenta pathology was seen in placentas from cluster 2 (total pathology score of 5.40; KruskalWallis test, P < .01), including the 3 placentas with the highest total pathology scores in the cohort (each placenta had total pathology score of 9 because of severe grading in lesions of maternal vascular malperfusion lesions as well as intervillous thrombi and retroplacental hematomas). Interestingly, all 3 cases developed PE and delivered an SGA infant preterm. Overall, histopathology findings enriched in cluster 2 samples were maternal vascular malperfusion lesions (Fisher exact test, P <.01) such as distal villous hypoplasia, placental infarctions, advanced villous maturity, and increased syncytial knots.

ajog.org

OBSTETRICS

Original Research

Placentas from cluster 3 also demonstrated significant placental histopathology (total pathology score of 4.91; Kruskal-Wallis test, P < .01), with the presence of lesions consistent with a maternal-fetal interface disturbance (Kruskal-Wallis test, P ¼ .01) such as massive perivillous fibrin deposition, maternal floor infarct, and/or intervillous thrombi. Placentas belonging to cluster 4 displayed distinct lesions of histological chorioamnionitis (KruskalWallis test, P < .01), while there was no evident enrichment of any placental lesions identified in samples from cluster 5. Additionally, we investigated the total pathology score in each gene expressionebased cluster by maternal hypertensive status (hypertensive vs normotensive; Supplemental Table 2). Histological chorioamnionitis severity was highest for normotensive women in cluster 4 (Kruskal-Wallis test, P < .01). Maternal vascular malperfusion lesion severity was higher in hypertensive women compared with normotensive women, across all clusters (KruskalWallis test, P < .01). Severity scores for placenta villous maldevelopment, fetal vascular malperfusion, and chronic inflammation were increased in hypertensive women belonging to cluster 3 (Kruskal-Wallis test, P < .05). Interestingly, the sole hypertensive woman belonging to cluster 4 demonstrated a similar severity trend as hypertensive cluster 3 women.

FIGURE 1

t-SNE visualization of placenta samples by histology

Samples that plot closer together demonstrate more similar histopathology. A, Overall, the grouping of patients by histology demonstrated a general congruency to those originally identified through gene expressionebased clustering (cluster 1, black; cluster 2, red; cluster 3, green; cluster 4, blue; cluster 5, cyan). B, However, several subsets of samples with similar clinical features such as maternal hypertensive state (normotensive, chronic hypertensive, or preeclampsia), infant birthweight (appropriate for gestational age or small for gestational age), and gestational age at delivery (preterm [<34 weeks] or term) grouped together by histology, regardless of their placental gene expression profile, while other clinical phenotypes scattered throughout the plot. C, Samples with increasingly severe lesions of maternal vascular malperfusion (score range, 0e8, light to dark red) plotted on the lower half of the t-SNE plot, while the group of placentas on the top left of the plot was driven by the presence of histological chorioamnionitis (score range, 0e4, light to dark blue) (D). Samples with lesions of maternal-fetal interface disturbance (E) and/or chronic inflammation (F) were predominantly found along the right side of the plot (both range, 0e3, light to dark green). AGA, appropriate for gestational age; CH, chronic hypertensive; N, normotensive; PE, preeclampsia; SGA, small for gestational age. Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

Relationship among samples based on histopathology scores t-SNE was used to visualize the high dimensional relationships between the placental samples in 2-diimensioanl space, based on the quantitative severity scoring of histopathology. This revealed an overall congruency in the grouping of patients to those originally identified through gene expression clustering, particularly for the samples belonging to gene expression-based clusters 2 and 4 (Figure 1A). Similar to the gene expression clustering findings, samples with similar

DECEMBER 2018 American Journal of Obstetrics & Gynecology

604.e10

Original Research

OBSTETRICS

clinical characteristics such as maternal hypertensive state, infant birthweight and gestational age at delivery were seen to group together by histopathology. However, this was not uniform for all clinical phenotypes (Figure 1B). In general, samples with increasing severity scores for lesions of maternal vascular malperfusion plotted on the lower half of the t-SNE plot (Figure 1C), while placentas with features of histological chorioamnionitis formed a group on the top left of the plot (Figure 1D). Samples with lesions of maternal-fetal interface disturbance and chronic inflammation were predominantly found along the right side of the plot (Figure 1, E and F).

Histopathological subgroups within each gene expressionebased cluster identified by hierarchical clustering Given the observed histological heterogeneity within the gene expressione based clusters (particularly clusters 1, 3, and 5), further hierarchical clustering of the histopathology severity scores within each individual gene expressione based cluster was performed to identify intracluster histological subgroups (Figure 2). Within cluster 1, placentas from healthy pregnancies demonstrated histopathology-driven subgroups with: little to no pathology; meconium histiocytes/macrophages within membranes; or villitis of unknown etiology (Figure 2A). The PE samples in gene expressione based cluster 1 (n ¼ 14) split into 3 histopathology subgroups based on the severity of maternal vascular malperfusion lesions present: placentas with a score of 0e2 clustered with the healthy samples; placentas with a score of 3e4 formed a subgroup associated with term deliveries; and placentas with a score of 4e6 formed a subgroup with preterm deliveries. Approximately half of the normotensive preterm samples in this gene expressionebased cluster formed their own histopathology subgroup, driven by the presence of lesions consistent with histological chorioamnionitis. Gene expressionebased cluster 2 samples were fairly cohesive by

histopathology, with subgroup formation driven by the severity of maternal vascular malperfusion lesions and cooccurrence of chronic deciduitis (Figure 2B). Additionally, a small subgroup of samples in this cluster (n ¼ 6, 3 diagnosed with PE) exhibited little or no evidence of maternal vascular malperfusion histopathology. In gene expressionebased cluster 3, the 11 samples clustered into 3 distinct histopathology subgroups: one with dominant features of massive perivillous fibrin deposition, maternal floor infarct, and/or chronic inflammation (PE/SGA/term); a second with signs of maternal vascular malperfusion (PE/SGA/preterm); and the third with minimal histopathology (Figure 2C). Within this third subgroup, 2 PE/appropriate for gestational age/term associated placentas showed evidence of intervillous thrombi. Furthermore, a single PE/SGA/preterm placenta displayed features of both maternal malperfusion and a maternal-fetal interface disturbance and clustered between the first 2 histopathology-driven subgroups. Within gene expressionebased cluster 4, the majority of placentas demonstrated evidence of histological chorioamnionitis (Figure 2D). However, a small subgroup of samples (n ¼ 4) exhibited either minimal placental histopathology or lesions of maternal-fetal interface disturbance and chronic inflammation. Finally, gene expressionebased cluster 5 samples, previously characterized as having chromosomal abnormalities likely because of confined placental mosaicism,25 split into 3 subgroups by histology: one with little to no pathology (mostly term deliveries); one with severe features of maternal malperfusion (hypertensive with preterm deliveries); and a single placenta from a normotensive woman delivered preterm with evidence of histological chorioamnionitis (Figure 2E).

Relationship between gene expression and histopathology in each PE subtype Overall, the degree of concordance between gene expressionebased classification and histopathology phenotype was 65% (93 of 142 samples; c2 test, P < .01)

604.e11 American Journal of Obstetrics & Gynecology DECEMBER 2018

ajog.org (Figure 3). By individual cluster, 62% of cluster 1 placentas (32 of 52; c2 test, P ¼ .10) demonstrated gene expressione histological concordance, with normal/ healthy placental gene expression profiles and no evidence of significant placental histopathology. These concordant samples were centered in cluster 1 on the PCA plot of gene expression (Figure 3, A and C) and separated distinctly from the other gene expressionebased clusters. Gene expressionehistological discordant samples in cluster 1 demonstrated placental lesions characteristic of gene expressionebased cluster 2 (maternal vascular malperfusion), cluster 3 (chronic inflammation), or cluster 4 (histological chorioamnionitis). Samples with maternal vascular malperfusion lesions or histological chorioamnionitis plotted on the periphery of cluster 1, bordering clusters 2 (Wilcoxon test, P <.01) and 4 (Wilcoxon test, P ¼.02), respectively (Figure 3, B and D). However, the placentas with features of chronic inflammation (specifically villitis of unknown etiology) did not plot significantly closer to gene expressionebased cluster 3 (Wilcoxon test, P ¼ .54). Within cluster 2 (88% concordant, 46 of 52; c2 test, P <.01) and cluster 4 (69% concordant, 9 of 13; c2 test, P ¼ .17) themselves, gene expressionehistological concordant samples plotted farther away from cluster 1 (Figure 3, A and C), while the discordant samples were located closer to cluster 1 (Wilcoxon test, P ¼ .05 and P ¼.15, respectively) (Figure 3, B and D). In cluster 3 (55% concordant, 6 of 11; c2 test, P ¼ .76), the discordant preterm hypertensive patients with severe maternal vascular malperfusion lesions formed a group bordering cluster 2 (Wilcoxon test, P ¼ .41 compared with the concordant cluster 3 samples) (Figure 3, B and D). Lastly, gene expressionebased cluster 5 had no clear defining histological features; however, placentas with minimal to no pathology plotted closer to the center of principal component 1 (PC1), in line with cluster 1, while those with features of maternal malperfusion plotted along the negative edge of PC1, similar to cluster 2 (Wilcoxon test,

ajog.org

OBSTETRICS

Original Research

FIGURE 2

Visualization of histology clustering within each gene expressionebased cluster

Within a given gene expressionebased cluster, histopathology data were subjected to hierarchical clustering, and the results were plotted as phylogenetic trees. Samples plotting closer together demonstrated more similar histological features, with tips colored based on clinical outcome. Each subgroup of placentas identified was assigned a group label, colored based on the gene expressionebased cluster that the subgroup most closely resembled histologically (Panel A: cluster 1, black; Panel B: cluster 2, red; Panel C: cluster 3, green; Panel D: cluster 4, blue; Panel E: cluster 5, cyan). Maximum severity score by category was as follows: maternal vascular malperfusion (maximum score of 13), implantation site abnormalities (maximum score of 4), histological chorioamnionitis (maximum score of 11), placental villous maldevelopment (maximum score of 5), fetal vascular malperfusion (maximum score of 6), chronic uteroplacental separation (maximum score of 3), maternal-fetal interface disturbance (maximum score of 5), and chronic inflammation (maximum score of 6). DVH, distal villous hypoplasia; MPFD, massive perivillous fibrin deposition; MVM, maternal vascular malperfusion; VUE, villitis of unknown etiology. Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

DECEMBER 2018 American Journal of Obstetrics & Gynecology

604.e12

Original Research

OBSTETRICS

FIGURE 3

PCA plots of gene expressionehistological concordance and discordance

These plots are modified from the original PCA plot in our prior gene expression study, restricted to the 142 samples with both gene expression and histological data available (gene expressionebased cluster 1, black; gene expression-based cluster 2, red; gene expressionebased cluster 3, green; gene expressionebased cluster 4, blue; gene expressionebased cluster 5, cyan). Placentas with concordant gene expressionehistological features are shown in full color from the front (A) and from the top (C), while samples with discordant gene expressionehistological features are shown in full color (front) (B) and (top) (D). Concordant patients form tighter groups by gene expression, plotting farther away from the borders, while discordant samples generally plot near the cluster with more similar histological features. Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

P ¼ .03) (Figure 3, B and D). These patterns indicate the existence of blended or intermediate phenotype samples on the cluster borders.

Comment Principal findings In this study, we utilized standardized, quantitative placental histopathology to further characterize our identified distinct gene expressionebased subtypes of PE. In general, placental pathology findings mirrored cluster-specific gene expression/enrichment findings (65%

overall concordance), providing strong evidence for at least 3 distinct forms of placenta disease underlying PE (Figures 1 and 3). Cluster 1 placentas, which appeared the healthiest by gene expression, exhibited minimal histopathology, whereas cluster 2 placentas displayed lesions consistent with maternal vascular malperfusion and similarly demonstrated an overexpression of hypoxia-mediated gene sets. Placental lesions consistent with a maternal-fetal interface disturbance defined several cluster 3 placentas,

604.e13 American Journal of Obstetrics & Gynecology DECEMBER 2018

ajog.org in agreement with the observed enrichment of immune response and proinflammatory gene sets in this group. Additionally, placentas belonging to cluster 4, primarily made up of preterm control samples, were robustly associated with histological chorioamnionitis. A fifth cluster was originally identified, with no clear clinical, epigenetic, or ontological associations but with evidence of chromosomal abnormalities likely resulting from confined placental mosaicism. Unsurprisingly, no unique set of histopathology features was capable of defining this group of placentas. PE placentas within this group tended to display histopathology of other groups, such as those of gene expressionbased subtype 1 or 2, suggesting that placental mosaicism may not have its own histopathology signature. Together with gene expression results, this finding solidifies our notion that this may not be a true distinct subtype of PE but instead be a sampling artifact of isolated placental mosaicism. Cases of discordance between placental gene expression and histology were found to be statistically enriched at the borders of gene expressionebased clusters when visualized on a PCA plot (Figure 3). This result indicates that these patients share some gene expression and histopathology features of both adjacent clusters, suggesting a possible intermediate phenotype. Additionally, concordant subgroups within each gene expressionebased cluster, with agreement between gene expression and histopathology findings, often revealed subtle differences in the severity of observed placental lesions and/or co-occurrence of additional placental features. This finding suggests that a gradient of PE severity may exist within a cluster. Lastly, the innovative integration of these modalities has further allowed us to identify smaller subgroups of patients within each PE subtype with likely mixed pathophysiology responsible for their disease development (Figure 3). Recognizing PE subtypes, both from a gene expression and a histopathology perspective, resulted in groups of patients with less heterogeneity in terms of

ajog.org clinical presentation, disease severity, and pregnancy outcomes, indicating a refinement in the classification of PE populations with this approach.

Results in context Advancing our understanding of the pathophysiologies underlying PE through the identification of clinically relevant disease subtypes would not only explain the heterogeneity observed in this disorder but also improve the care of women and babies affected by PE through etiological-based interventions. In this study, we have used the combination of gene expression profiling and clinical histopathology to confirm the presence of distinct PE patient populations with divergent forms of placental disease, moving beyond the classical separation of early- vs late-onset PE. Furthermore, our study has allowed us to better understand the patients belonging to each identified PE subtype. For example, our gene expression data set identified PE subtype 1 patients (gene expressionebased cluster 1) as those with normal placental gene expression, initially indicating minimal placental contribution to PE establishment/progression. However, histopathology showed that the majority of PE patients found within this healthy placenta cluster did, in fact, demonstrate some evidence of maternal vascular malperfusion, despite no significant global alterations in gene expression patterns. We would speculate that many of these patients may demonstrate a similar placental disease phenotype as PE subtype 2 patients (gene expressionebased cluster 2) but possibly a milder form of placental damage and/or a robust protective response to hypoxic stress. Interestingly, the PE subtype 1 patients would primarily be described as lateonset PE, whereas PE subtype 2 is enriched in early-onset PE patients. The discovery of similar histopathology findings in both groups indicates that these clinically distinct subtypes of PE patients may have similar underlying forms of placental dysfunction, but the complexity and/or severity of the placental pathology present is reflected through the timing and severity of the

OBSTETRICS

clinical manifestation of disease in the mother. Our findings further highlight the existence of women who develop PE despite no obvious placental contribution. Five PE subtype 1 patients (w6% of all PE samples) demonstrated very limited evidence of placental pathology, through both gene expression profiling and histopathology, identifying a true population of PE patients with healthy placentas. These patients likely represent a subtype of PE pathophysiology driven in large part by maternal constitutive factors (ie, subclinical cardiovascular dysfunction), and we can speculate that these patients may be among those with the highest risk of cardiovascular disease in later life because of these subclinical cardiovascular risk factors.7 Furthermore, we have revealed considerable similarity between the placentas from women with PE and those with chronic hypertension who did not develop PE in terms of their scored histological features, similar to our prior gene expression analysis.25 This confirms that these 2 hypertensive states are nearly indistinguishable at the placental level, and matched maternal blood samples will be required to understand the transition from a chronic hypertension to a PE disease phenotype. PE subtype 3 includes the most intriguing group of patients, who are currently poorly described in the literature. The core group of gene expressionehistological concordant PE patients found within cluster 3 demonstrates gene expression profiles and histological findings consistent with profound immune activation, including overexpression of tumor necrosis factora, interferon-g, chemokine ligand-10, and histological evidence of massive perivillous fibrin deposition. It may therefore be pertinent to investigate promising predictive biomarkers and therapeutic interventions used in the field of transplantation medicine for potential application to this unique immune-driven subtype of patients.47,48 Additionally, a significant discordant subgroup of hypertensive patients was also discovered within cluster 3,

Original Research

demonstrating dominant maternal vascular malperfusion lesions, despite gene expression profiles suggesting immune activation. Whether these are intermediate phenotypes of PE with mixed pathophysiology or the result of external factors, such as increased infiltration of maternal immune cells into these placentas, is unclear. Differences in immune cell populations in cluster 3 placentas, compared with the other clusters, is currently under investigation by our group.

Clinical implications The identification of clinically relevant subtypes of PE through novel integration of gene expression and histopathology findings allows one to envision the development and/or application of etiology-focused screening tools and therapies for PE. For example, PE subtype 2 is defined by upregulation of classical angiogenic genes (eg, fms-like tyrosine kinase-1 and endoglin) and maternal vascular malperfusion. As biomarkers, these proteins are likely ideal candidates for identifying this specific subtype of PE patients. Because PE subtypes 1 and 3 are not characterized by placental hypoxia and angiogenic dysfunction, patients belonging to these subtypes will likely not be identified by these markers, likely explaining the overall modest accuracy of these biomarkers in the prediction of PE as a single entity.49e51 Furthermore, the clinical application of a divergent placental disease paradigm for PE has the potential to improve patient counselling and follow-up practices in regard to recurrence risks for subsequent pregnancies and short- and long-term health outcomes for mother and infants. For example, findings of massive perivillous fibrin deposition and maternal floor infarction are associated with a high degree of recurrence.52 Considering that these placental lesions are highly enriched in PE subtype 3, these findings could be used to specifically identify this PE patient population as one at highest risk for PE recurrence. In a similar vein, PE subtype 1 patients may have the highest risk for future cardiovascular disease because many of these

DECEMBER 2018 American Journal of Obstetrics & Gynecology

604.e14

Original Research

ajog.org

OBSTETRICS

patients demonstrated minimal placental contribution to disease establishment and instead may have had underlying subclinical cardiovascular dysfunction at the onset of pregnancy. This information could better inform referral practices to specialized postpartum cardiovascular clinics that aim to modify maternal cardiovascular risk after PE.53 Well-described prospective cohort studies that focus on profiling maternalplacental-fetal contributions to disease are required to validate these hypotheses in future investigations. However, the potential to improve clinical care for women and infants affected by PE in a personalized medicine approach is clear.

characteristics, and placental pathology. For example, by viewing the disorder from the perspective of disease subtypes as opposed to traditional clinical diagnosis, we delineated the variability in PE placental pathology reported in previous studies.19,55e57 Moving forward, the use of a PE subtype paradigm may allow for more homogeneous patient population selection in study designs, leading to greater scientific contributions in regard to understanding of disease pathophysiology, discovery of highly accurate predictive biomarkers, and targeted therapeutic interventions tailored to each disease subtype.

Strengths and limitations Research implications Placental pathology is an important modality for understanding placentamediated diseases such as PE.54 While histopathology is classically a descriptive technique, we utilized a quantitative approach to comprehensively and systematically grade the presence and severity of placental lesions using a datacollection tool that reflects current consensus guidelines for placental pathology. The generation of quantitative histopathology outputs allowed for data analysis modalities traditionally reserved for omics-type data sets (ie, clustering analysis) and, importantly, allowed for the novel integration of placental histopathology, placental gene expression, and clinical profile data sets, an approach integral to the validation of 3 clinically relevant PE subtypes. Our approach demonstrates that clinical placental pathology can be meaningfully incorporated with traditional basic science data sets to better understand disease pathophysiology and should be included as important variables in future studies investigating PE and other placentalmediated diseases. A considerable strength of our research approach is the ability to reduce heterogeneity observed in PE, an issue that has plagued research endeavors in the field. By utilizing the disease subtype paradigm, we were able to characterize more homogeneous patient groups in terms of gene expression, clinical

A major strength of this study is our well-defined cohort, spanning the clinical spectrum of PE presentation, with robust clinical, gene expression, and histological data. To our knowledge, this study is the first to attempt to link gene expression data with histopathology to identify and described subpopulations of PE patients. While this approach has not been previously undertaken in the context of placental-mediated diseases, transcriptional and histological assessment to better phenotype tumors has been successful in the field of oncology.58e61 Moreover, we limited the variability in our placental pathology assessments through the use of a standardized data collection form, established using evidence-based criteria for assessing the severity of placental lesions. There likely exists some bias in our initial sample selection that may have minimized our identification of concordant cluster 3 (immune-driven) PE patients. The placentas originally selected for microarray analysis were limited to those associated with live births, and given that the histological features characteristic of cluster 3 PE patients (massive perivillous fibrin deposition and maternal floor infarct) are often lethal and observed earlier in pregnancy,62e64 we have likely excluded the more severe and common form of this pathology. Additionally, it is possible that some of the gene expression-histological discordance observed in our study is the result

604.e15 American Journal of Obstetrics & Gynecology DECEMBER 2018

of sampling limitations. The placenta is a redundant structure composed of hundreds of chorionic villi over a large surface area. While we utilized 4 biopsies randomly drawn from each quadrant of the placenta, the histological and molecular analysis was not performed on identical tissue samples. It is conceivable that sampling could have occurred in regions of the placenta that have different pathology or cellular composition. The use of 4 biopsies should help to mitigate this effect, but higher sampling rates could be warranted in future studies. Furthermore, while we applied a robust data clustering method, there is potential, albeit small, for patients to be misclassified based on the false-positive error inherent to this type of statistical modelling. For example, one PE woman who delivered at term with no evidence of intrauterine infection was assigned to gene expressionebased cluster 4, a cluster defined by normotensive patients with preterm deliveries and gene expression and histopathology findings consistent with chorioamnionitis. Further validation of integrated gene expression and histopathology data sets unique to distinct PE subtypes in additional patient cohorts will help clarify the issue of patient misclassification in future work.

Conclusions Collectively, this body of work has confirmed and contributed to an expanded understanding of the divergent pathophysiologies underlying a PE diagnosis. Critically, histopathology can be used as a quantitative measure when applied using a systematic scoring system. Specifically, our findings in this current study demonstrate that distinct PE phenotypes of placental disease can be described by integrating histopathology and gene expression. Histopathology provided considerable contextual information, allowing for the identification of gradients of disease severity in each PE subtypes. Importantly, it further discovered potential intermediate phenotypes of PE not found through gene expression profiling alone. Being able to adequately identify and understand

ajog.org these distinct patient populations within the clinical setting is a critical first step in the generation of effective predictive and diagnostic biomarkers, targeted therapeutic approaches, and clinical management of these PE subtypes. n References 1. Steegers EA, von Dadelszen P, Duvekot JJ, Pijnenborg R. Pre-eclampsia. Lancet 2010;376: 631–44. 2. Duley L. Pre-eclampsia and the hypertensive disorders of pregnancy. Br Med Bull 2003;67: 161–76. 3. Hutcheon JA, Lisonkova S, Joseph KS. Epidemiology of pre-eclampsia and the other hypertensive disorders of pregnancy. Best Pract Res Clin Obstet Gynaecol 2011;25: 391–403. 4. Magee LA, Pels A, Helewa M, Rey E, von Dadelszen P; Canadian Hypertensive Disorders of Pregnancy (HDP) Working Group. Diagnosis, evaluation, and management of the hypertensive disorders of pregnancy. Pregnancy Hypertens 2014;4:105–45. 5. Eastabrook G, Brown M, Sargent I. The origins and end-organ consequence of preeclampsia. Best Pract Res Clin Obstet Gynaecol 2011;25:435–47. 6. Bellamy L, Casas JP, Hingorani AD, Williams DJ. Pre-eclampsia and risk of cardiovascular disease and cancer in later life: systematic review and meta-analysis. BMJ 2007;335:974. 7. Smith GN, Pudwell J, Walker M, Wen SW. Ten-year, thirty-year, and lifetime cardiovascular disease risk estimates following a pregnancy complicated by preeclampsia. J Obstet Gynaecol Can 2012;34:830–5. 8. Brown MC, Best KE, Pearce MS, Waugh J, Robson SC, Bell R. Cardiovascular disease risk in women with pre-eclampsia: systematic review and meta-analysis. Eur J Epidemiol 2013;28: 1–19. 9. Xiao R, Sorensen TK, Williams MA, Luthy DA. Influence of pre-eclampsia on fetal growth. J Matern Fetal Neonatal Med 2003;13:157–62. 10. Backes CH, Markham K, Moorehead P, Cordero L, Nankervis CA, Giannone PJ. Maternal preeclampsia and neonatal outcomes. J Pregnancy 2011;2011:214365. 11. Giannakou K, Evangelou E, Papatheodorou SI. Genetic and non-genetic risk factors for pre-eclampsia: an umbrella review of systematic reviews and meta-analyses of observational studies. Ultrasound Obstet Gynecol 2018;51:720–30. 12. Schneider H. Placental dysfunction as a key element in the pathogenesis of preeclampsia. Dev Period Med 2017;21:309–16. 13. Gillon TE, Pels A, von Dadelszen P, MacDonell K, Magee LA. Hypertensive disorders of pregnancy: a systematic review of international clinical practice guidelines. PLoS One 2014;9:e113715.

OBSTETRICS

14. Roberge S, Bujold E, Nicolaides KH. Aspirin for the prevention of preterm and term preeclampsia: systematic review and metaanalysis. Am J Obstet Gynecol 2018;218: 287–93.e1. 15. Rolnik DL, Wright D, Poon LC, et al. Aspirin versus placebo in pregnancies at high risk for preterm preeclampsia. N Engl J Med 2017;377: 613–22. 16. Xu TT, Zhou F, Deng CY, Huang GQ, Li JK, Wang XD. Low-dose aspirin for preventing preeclampsia and its complications: a meta-analysis. J Clin Hypertens (Greenwich) 2015;17: 567–73. 17. Egbor M, Ansari T, Morris N, Green CJ, Sibbons PD. Morphometric placental villous and vascular abnormalities in early- and late-onset pre-eclampsia with and without fetal growth restriction. BJOG 2006;113:580–9. 18. Ogge G, Chaiworapongsa T, Romero R, et al. Placental lesions associated with maternal underperfusion are more frequent in early-onset than in late-onset preeclampsia. J Perinat Med 2011;39:641–52. 19. Falco ML, Sivanathan J, Laoreti A, Thilaganathan B, Khalil A. Placental histopathology associated with pre-eclampsia: systematic review and meta-analysis. Ultrasound Obstet Gynecol 2017;50:295–301. 20. Nelson DB, Ziadie MS, McIntire DD, Rogers BB, Leveno KJ. Placental pathology suggesting that preeclampsia is more than one disease. Am J Obstet Gynecol 2014;210:66. e1–7. 21. Pathak S, Lees CC, Hackett G, Jessop F, Sebire NJ. Frequency and clinical significance of placental histological lesions in an unselected population at or near term. Virchows Arch 2011;459:565–72. 22. Huppertz B. Placental origins of preeclampsia: challenging the current hypothesis. Hypertension 2008;51:970–5. 23. Ness RB, Roberts JM. Heterogeneous causes constituting the single syndrome of preeclampsia: a hypothesis and its implications. Am J Obstet Gynecol 1996;175:1365–70. 24. Schlembach D. Pre-eclampsia—still a disease of theories. Fukushima J Med Sci 2003;49: 69–115. 25. Leavey K, Benton SJ, Grynspan D, Kingdom JC, Bainbridge SA, Cox BJ. Unsupervised placental gene expression profiling identifies clinically relevant subclasses of human preeclampsia. Hypertension 2016;68: 137–47. 26. Leavey K, Wilson SL, Bainbridge SA, Robinson WP, Cox BJ. Epigenetic regulation of placental gene expression in transcriptional subtypes of preeclampsia. Clin Epigenetics 2018;10:28. 27. Magee LA, Helewa M, Rey E; Hypertension Guideline Committee; Strategic Training Initiative in Research in the Reproductive Health Sciences (STIRRHS) Scholars. Diagnosis, evaluation, and management of the hypertensive disorders of pregnancy. J Obstet Gynaecol Can 2008;30: S1–2.

Original Research

28. National High Blood Pressure Education Program Working Group on High Blood Pressure in Pregnancy. Report of the National High Blood Pressure Education Program Working Group on High Blood Pressure in Pregnancy. Am J Obstet Gynecol 2000;183:S1–22. 29. Kramer MS, Platt RW, Wen SW, et al. A new and improved population-based Canadian reference for birth weight for gestational age. Pediatrics 2001;108:E35. 30. Hadlock FP, Harrist RB, Martinez-Poyer J. In utero analysis of fetal growth: a sonographic weight standard. Radiology 1991;181:129–33. 31. Acharya G, Wilsgaard T, Berntsen GK, Maltau JM, Kiserud T. Reference ranges for serial measurements of blood velocity and pulsatility index at the intra-abdominal portion, and fetal and placental ends of the umbilical artery. Ultrasound Obstet Gynecol 2005;26: 162–9. 32. Baschat AA, Gembruch U. The cerebroplacental Doppler ratio revisited. Ultrasound Obstet Gynecol 2003;21:124–7. 33. R Core Team. R: a language and environment for statistical computing. Vienna, Austria: Foundation for Statistical Computing; 2015. 34. Warrander LK, Batra G, Bernatavicius G, et al. Maternal perception of reduced fetal movements is associated with altered placental structure and function. PLoS One 2012;7: e34851. 35. Benton SJ, McCowan LM, Heazell AE, et al. Placental growth factor as a marker of fetal growth restriction caused by placental dysfunction. Placenta 2016;42:1–8. 36. Benton S, Fellus I, et al. Utilization a standardised pathology examination tool to improve placental pathology examination and reporting. Reprod Sci 2016;23:319A. 37. Redline RW. Placental pathology: a systematic approach with clinical correlations. Placenta 2008;29(Suppl A):S86–91. 38. Redline RW, Boyd T, Campbell V, et al. Maternal vascular underperfusion: nosology and reproducibility of placental reaction patterns. Pediatr Dev Pathol 2004;7:237–49. 39. Redline RW, Ariel I, Baergen RN, et al. Fetal vascular obstructive lesions: nosology and reproducibility of placental reaction patterns. Pediatr Dev Pathol 2004;7:443–52. 40. Redline RW, Faye-Petersen O, Heller D, et al. Amniotic infection syndrome: nosology and reproducibility of placental reaction patterns. Pediatr Dev Pathol 2003;6:435–48. 41. Redline RW. Villitis of unknown etiology: noninfectious chronic villitis in the placenta. Hum Pathol 2007;38:1439–46. 42. Benirschke K, Burton G, Baergen R. Pathology of the human placenta. Berlin, Germany: Springer-Verlag; 2012. 43. Khong TY, Mooney EE, Ariel I, et al. Sampling and definitions of placental lesions: Amsterdam Placental Workshop Group Consensus Statement. Arch Pathol Lab Med 2016;140:698–713. 44. Khong TY, Bendon RW, Qureshi F, et al. Chronic deciduitis in the placental basal plate:

DECEMBER 2018 American Journal of Obstetrics & Gynecology

604.e16

Original Research

OBSTETRICS

definition and interobserver reliability. Hum Pathol 2000;31:292–5. 45. van der Maaten L, Hinton G. Visualizing data using t-SNE. J Machine Learn 2008;9: 2431–56. 46. Paradis E, Claude J, Strimmer K. APE: analyses of phylogenetics and evolution in R language. Bioinformatics 2004;20: 289–90. 47. Romero RWA, Korzeniewski SJ, Than NG, et al. Maternal floor infarction/ massive perivillous fibrin deposition: a manifestation of maternal antifetal rejection? Am J Reprod Immunol 2013;70:285–98. 48. Israni AK, Leduc R, Jacobson PA, et al. Inflammation in the setting of chronic allograft dysfunction post-kidney transplant: phenotype and genotype. Clin Transplant 2013;27: 348–58. 49. Levine RJ, Maynard SE, Qian C, et al. Circulating angiogenic factors and the risk of preeclampsia. N Engl J Med 2004;350: 672–83. 50. Levine RJ, Lam C, Qian C, et al. Soluble endoglin and other circulating antiangiogenic factors in preeclampsia. N Engl J Med 2006;355:992–1005. 51. Kleinrouweler CE, Wiegerinck MM, RisStalpers C, et al. Accuracy of circulating placental growth factor, vascular endothelial growth factor, soluble fms-like tyrosine kinase 1 and soluble endoglin in the prediction of preeclampsia: a systematic review and meta-analysis. BJOG 2012;119:778–87. 52. Chen A, Roberts DJ. Placental pathologic lesions with a significant recurrence risk—what not to miss! APMIS 2018;126:589–601. 53. Smith GN. The maternal health clinic: improving women’s cardiovascular health. Semin Perinatol 2015;39:316–9.

54. Redline RW. The clinical implications of placental diagnoses. Semin Perinatol 2015;39: 2–8. 55. Roberts JM, Escudero C. The placenta in preeclampsia. Pregnancy Hypertens 2012;2: 72–83. 56. Stanek J. Placental pathology varies in hypertensive conditions of pregnancy. Virchows Arch 2018;472:415–23. 57. Vinnars MT, Wijnaendts LC, Westgren M, Bolte AC, Papadogiannakis N, Nasiell J. Severe preeclampsia with and without HELLP differ with regard to placental pathology. Hypertension 2008;51:1295–9. 58. Leong AS, Zhuang Z. The changing role of pathology in breast cancer diagnosis and treatment. Pathobiology 2011;78:99–114. 59. Soikkeli J, Lukk M, Nummela P, et al. Systematic search for the best gene expression markers for melanoma micrometastasis detection. J Pathol 2007;213:180–9. 60. Guan Z, Zhang J, Song S, Dai D. Promoter methylation and expression of TIMP3 gene in gastric cancer. Diagn Pathol 2013;8:110. 61. Kim NH, Lim HY, et al. Evaluation of clinicopathological characteristics and oestrogen receptor gene expression in oestrogen receptor-negative, progesterone receptorpositive canine mammary carcinomas. J Comp Pathol 2014;151:42–50. 62. Mandsager NT, Bendon R, Mostello D, Rosenn B, Miodovnik M, Siddiqi TA. Maternal floor infarction of the placenta: prenatal diagnosis and clinical significance. Obstet Gynecol 1994;83:750–4. 63. Katzman PJ, Genest DR. Maternal floor infarction and massive perivillous fibrin deposition: histological definitions, association with intrauterine fetal growth restriction, and risk of

604.e17 American Journal of Obstetrics & Gynecology DECEMBER 2018

ajog.org recurrence. Pediatr Dev Pathol 2002;5: 159–64. 64. Bane AL, Gillan JE. Massive perivillous fibrinoid causing recurrent placental failure. BJOG 2003;110:292–5. 65. Fitzgerald B, Kingdom J, Keating S. Distal villous hypoplasia. Diagnost Histopathol 2012;18:195–200.

Author and article information From the Departments of Cellular and Molecular Medicine (Drs Benton and Bainbridge) and Pathology and Laboratory Medicine (Dr Grynspan), Faculty of Medicine, and Interdisciplinary School of Health Sciences (Dr Bainbridge), Faculty of Health Sciences, University of Ottawa, Ottawa; and the Departments of Physiology (Drs Leavey and Cox) and Obstetrics and Gynaecology (Dr Cox), Faculty of Medicine, University of Toronto, Toronto, Ontario, Canada. 1 These authors contributed equally to this article. 2 Drs. Cox and Bainbridge are both seniors authors of this article. Received June 20, 2018; revised Sept. 7, 2018; accepted Sept. 24, 2018. The funding sources played no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. This study was supported by a Canadian Institutes of Health Research operating grant (128369) to Drs Bainbridge and Cox; a Physicians’ Services Incorporated Foundation grant to Drs Grynspan, Bainbridge, and Benton; and the Preeclampsia Foundation Vision grants (to Drs Bainbridge and Cox). Dr Benton is supported by a Molly Towell Research Fellowship, Dr Leavey is supported by an Ontario Graduate Scholarship, and Dr Cox is supported by a Tier 2 Canada Research Chair in Placental Development and Maternal-Fetal Health. The authors report no conflicts of interest. Corresponding authors: Brian Cox. [email protected]; or Shannon Bainbridge. [email protected]

ajog.org

OBSTETRICS

Original Research

SUPPLEMENTAL FIGURE 1

Glossary of Terms ape package: R package with functions for reading, writing, plotting, and manipulating phylogenetic trees. Appropriate-for-gestational-age (AGA): Neonatal birth weight >10th percentile for gestational age and sex. Bayesian Information Criterion (BIC): Criterion for model selection that works by rewarding models that fit the data well while penalizing overly complicated models with a lot of parameters to avoid over-fitting the data. The model with the lowest BIC is the optimal model. Chronic hypertension: Systolic pressure 140 mmHg and/or sustained diastolic 90 mmHg before 20 weeks gestation. Concordant samples: Placentas with agreement between gene expression and histopathological findings. Discordant samples: Placentas where gene expression and histopathology are not reflective of one another. dist function: Base function in R that computes the distances between the rows of a data matrix. Euclidean distance: The “straight-line” distance between two points (i.e. the length of a line connecting the two points). Hierarchical agglomerative clustering: A method of clustering analysis that treats each sample as their own cluster and then, based on distance metrics, successively merges pairs of similar clusters until the entire cohort is one cluster, forming a hierarchical tree. Perplexity: A parameter that can be modified while performing t-SNE analysis in R. An estimate of the optimal number of neighbours (recommended values are between 5 and 50). Phylogenetic trees: A visualization tool for demonstrating the degree of relatedness of samples obtained from clustering analysis. It is a type of dendrogram/hierarchical tree. plot3d function: Function in R (included in the rgl package) that plots the first three principal components (principal component 1 (PC1), PC2, and PC3) obtained from principal component analysis (PCA). These three components represent the most variability in the data set. Samples that plot closer together are more similar. prcomp function: Base function in R that performs principal component analysis. PCA is a linear method of data reduction that converts high dimensional data (ex. gene expression data) into new independent weighted variables called “principal components” (PCs). Genes with highly linearly correlated expression and responsible for the most variance in the data will contribute the most to the first principal component (PC1). Those weighted strongly in principal component 2 (PC2) are those that contribute the second greatest to the data variance and exhibit a significant linear correlation to each other, but are independent of PC1. This continues for PC3 onwards, depending on the number of samples. Once the contribution of each gene to each principal component is determined, these are used in combination with the original expression values to calculate a weighted score for each sample for each component. Preeclampsia: The onset of hypertension (systolic pressure 140 mmHg and/or diastolic pressure 90 mmHg) after 20 weeks’ gestation with proteinuria (>300 mg protein/day, or 2+ by dipstick). Preterm: Delivery before 34 weeks of gestation. Small-for-gestational-age (SGA): Neonatal birth weight <10th percentile for gestational age and sex. t-distributed stochastic neighbor embedding (t-SNE): A non-linear method of data reduction that works by calculating distance-based similarity scores between all the samples in high-dimensional space and then randomly projecting the sample points onto a two or three-dimensional plot. With each iteration, a given sample is moved closer to other samples with high similarity scores and farther from those with low similarity scores until the relationships between the points (i.e. the matrix of similarity scores) in low-dimensional space reflects the relationships between the points in the original high-dimensional space, or the set maximum number of iterations is reached. t-SNE plot: Visualization of t-SNE results. Unlike with PCA, the number of dimensions to be plotted (2 or 3) is pre-established and are all used to demonstrate the relationships between the points. Samples that plot closer together are more similar. Unsupervised clustering analysis: The identification of groups of samples based solely on similarities or dissimilarities in gene expression (or histological/proteomic/metabolomic, etc.) profiles, independent of clinical diagnosis or characteristics. Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

DECEMBER 2018 American Journal of Obstetrics & Gynecology

604.e18

Original Research

ajog.org

OBSTETRICS

SUPPLEMENTAL TABLE 1

Pathology data collection form Gross examination Indicate if examination was performed on fresh or fixed placenta , Fresh or , Fixed Trimmed placental weight : _______________________ grams Placenta weight percentile for gestational age at delivery : _______________________ *derived from local population standards34 Prior sampling? Yes or no Note disruptions to basal plate: Placental disk34 Maximal linear length

_________ cm

Maximal width (perpendicular to liner length)

_________ cm

Minimal thickness

_________ cm

Maximal thickness

_________ cm

Accessary lobes?

Yes or no Description:_______________________________

Umbilical cord34 Average diameter

_________ cm

Cord length

_________ cm

Velamentous cord insertion?

Yes or no Description:_______________________________

Marginal cord insertion? <1 cm from nearest margin

Yes or no

Peripheral cord insertion? <3 cm from nearest margin

Yes or no

Handedness of coiling

, Right or , left or , Could not be determined

Hypercoiling of cord? >3 coils per 10 cm

Yes or no

Segmented hypercoiling?

Yes or no

Presence of deep grooves between coils?

Yes or no

Hypocoiling of the cord? <1 coil per 10 cm

Yes or no

Two vessel cord?

Yes or no

True knots?

Yes or no Description:_______________________________

Membranes Color

___________________________________

Completeness

___________________________________

Extrachorialis?

Yes or no Description:_______________________________

Macroscopic lesions Retroplacental hematoma/hemorrhage? Yes or no Hemorrhage on the maternal surface of the disk, with congestion/compression of the overlying parenchyma Number: ____________ Estimated volume(s) as a percent of total disc volume : __________ % Location: ____________ Maternal surface fibrin? Yes or no Greatest thickness: ____________ mm Estimated volume(s) as a percentage of total disc volume: __________ % Location: Plaques? Yes or no Diffuse fibrin occupying entire maternal surface? Yes or no Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

604.e19 American Journal of Obstetrics & Gynecology DECEMBER 2018

(continued)

ajog.org

OBSTETRICS

Original Research

SUPPLEMENTAL TABLE 1

Pathology data collection form (continued) Impression of intervillous fibrin? Yes or no Estimate volume(s) as a percent of total disc volume: __________ % Location: ____________ Presence of lesions resembling infarcts? Yes or no Number: ____________ Size(s): ____________ Estimate volume(s) as a percentage of total disc volume: __________ % Location: ____________ Presence of lesions resembling intervillous thrombi? Yes or no Number: ____________ Size(s): ____________ Estimate volume(s) as a percentage of total disc volume: __________ % Location: ____________ Recent, remote or mixed? ____________ Indeterminate lesions? Yes or no Number: ____________ Size(s): ____________ Estimate volume(s) as a percentage of total disc volume: __________ % Location: ____________ Recent, remote, or mixed? ____________ Microscopic lesions Evidence of maternal vascular malperfusion Distal villous hypoplasia34  Reduction in size of intermediate villi with dispersed terminal villi and reduced number that appear thin and elongated, widening of intervillous space; adjusted for gestational age at delivery; involves at least 30% of full-thickness slide 0, not present 1, focal (1 slide only) 2, diffuse (2 slides)

Grade: _____________ * Qualify infarct as recent or remote or mixed2

Placental infarct(s)  Refer to gross description, exclude marginal infarctions in a term placenta 0, no infarcts present 1, focal infarctions (1e3 peripherally located, <3 cm in size) 2, multifocal and/or diffuse infarctions (>3 peripherally located) and/ or any infarct 3 cm in size; >10% of villous volume

Grade: _____________

Accelerated villous maturation pattern34  Presence of term-appearing/hypermature villi for gestational age, not in areas adjacent to infarction 0, villi structure and vessel pattern appropriate for gestational age 1, focal hypermature for gestational age 2, Diffuse hypermature for gestational age

Grade: _____________

Syncytial knots29  Aggregates of syncytiotrophoblast nuclei along stem and/or at terminal villi 0, focal and infrequent presence of syncytial knots, expected for gestational age (<30% terminal villi with knots) 1, syncytial knots excessively increased for gestational age (30% parenchyma) 2, syncytial knots excessively increased for gestational age (>30% parenchyma)

Grade: _____________

Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

(continued)

DECEMBER 2018 American Journal of Obstetrics & Gynecology

604.e20

Original Research

ajog.org

OBSTETRICS

SUPPLEMENTAL TABLE 1

Pathology data collection form (continued) Focal perivillous fibrin deposition29  Increased amounts of fibrin coating proximal stem villi and/or terminal villi 0 ¼ not present 1 ¼ present, seen on <2 slides; increased for gestational age Estimated % volume occupied:

Grade: _____________

Villous agglutination29  Clusters of adherent terminal villi (>2, <20), enmeshed by fibrin and/ or bridging syncytial knots 0, not present 1, focal 2, patchy 3, diffused

Grade: _____________

Decidual vasculopathy29,34 0, not present 1, present as either insufficient vessel remodeling and/or fibrinoid change

Grade: _____________

Implantation site abnormalities Microscopic accreta33  Bundles of myometrium adherent to the basal plate without intervening decidua 0, not present 1, focal 2, multifocal or diffuse (more than 1 focus)

Grade: _____________

Increased basement membrane fibrin65 0, not present 1, patchy fibrin on the maternal surface (basal plate) 2, diffuse fibrin on the maternal surface (basal plate)

Grade: _____________

Evidence of histological chorioamnionitis31,34 Maternal inflammatory response Stage: 0, not present 1, stage 1: neutrophils in subchorionic fibrin and/or trophoblast layer of membrane 2, stage 2: diffuse or patchy neutrophils in fibrous chorion or amnion 3, stage 3: membrane or chorionic plate necrosis

Stage: _____________

Grade 0, not present 1, mild or moderate: lacks criteria for grade 2 2, severe: confluent neutrophils between chorion and decidua, greater than 10  20 cells in extent with greater than 3 foci or a large continuous band

Grade: _____________

Fetal inflammatory response Stage: 0, not present 1, stage 1: chorionic vessel vasculitis or umbilical venous vasculitis 2, stage 2: umbilical vasculitis with umbilical arteritis 3, stage 3: necrotizing funititis/concentric umbilical perivasculitis

Stage: _____________

Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

604.e21 American Journal of Obstetrics & Gynecology DECEMBER 2018

(continued)

ajog.org

OBSTETRICS

Original Research

SUPPLEMENTAL TABLE 1

Pathology data collection form (continued) Grade 0. not present 1, mild to moderate: lacks criteria for grade 2 2, severe: heavy inflammation of vessel within the umbilical cord or chorionic plate vessel with vessel wall damage

Grade: _____________

Thrombosis of any of the umbilical or chorionic fetal vessels present

Yes or no

Specific patterns Candida spp  Gross punctate white nodules on umbilical cord (yes or no); refer to gross findings  Subamniotic microabcesses on umbilical cord (yes or no) B Grocott stain: , not done , negative , positive Histochemical pseudohyphae and yeast forms (yes or no) Listeria  Gross intervillous abscesses (yes or no); refer to gross findings  Histological intervillous abscesses B Gram stain: , not done , negative , positive Gram-negative rods within abscesses (yes or no) Evidence of placenta villous maldevelopment Chorangiosis33  Hypercapillarized terminal villi 0, not present 1, present with >10 terminal villi with 10 capillaries, seen in 3 foci seen

Grade: _____________

Chorangiomas33 0, not present 1, present and <3 cm in size 2, present and 3 cm in size or >5 total nodules

Grade: _____________

Delayed villous maturation34  Monotonous villi (10) with centrally placed capillaries and decreased vasculosyncytial membranes resembling villi in early pregnancy, present in at least 30% of full-thickness section 0, no villous immaturity 1, focal: lesion seen on 1 slide only 2, diffuse: seen on 2 slides

Grade: _____________

Evidence of fetal vascular malperfusion Avascular fibrotic villi34 0, none present 1, small foci: 3 or more foci of 2e4 terminal villi showing complete loss of villous capillaries and bland hyaline fibrosis of the villous stroma 2, intermediate foci: 3 or more foci of 5e10 terminal villi 3, large foci: 3 or more foci of >10 villi

Grade: _____________

Thrombosis34 0, not present 1, present Location: umbilical, chorionic plate, stem vessel Number : ______________ , Occlusive or , Nonocclusive

Grade: _____________

Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

(continued)

DECEMBER 2018 American Journal of Obstetrics & Gynecology

604.e22

Original Research

ajog.org

OBSTETRICS

SUPPLEMENTAL TABLE 1

Pathology data collection form (continued) Intramural fibrin deposition34  Subendothelial or intramuscular fibrin or fibrinoid deposition within the wall of large fetal vessel (recent), with calcifications (remote) 0, not present 1, recent, isolated (only 1 seen per slide) 2, remote, isolated (only 1 seen per slide)

Grade: _____________

Evidence of chronic uteroplacental separation33,34 Chorionic hemosiderosis 0, no 1, yes

Grade: _____________

Presence of retroplacental adherent hematoma  Refer to gross description, confirm histologically 0, no 1, yes

Grade: _____________

Laminar necrosis of decidua capsularis 0, no 1, yes

Grade: _____________

Evidence of maternal-fetal interface disturbance Massive perivillous fibrin deposition pattern33 0, not present 1, diffusely present, mild/moderate / 30e50% of intervillous volume, seen on at least 2 slides 2, diffusely present, severe / >50% of intervillous volume, must be seen on all slides

Grade: _____________

Maternal floor infarct pattern33 0, not present 1, present in 1e2 slides 2, whole floor, present in all slides Thickness:_______________________

Grade: _____________

Intervillous thrombi 0, not present 1, present Number : ____________ Size(s) : ____________ Estimate volume(s) as a percentage of total disc volume: __________ %

Grade: _____________

Evidence of chronic inflammation Infectious villitis 0, not present 1, placental villous inflammation with features suggesting an infectious etiology:  Plasma cell villitis  Viral cytopathic effect/CMV  Viral cytopathic effect: HSV  Viral cytopathic effect: NOS  Toxoplasmosis Immunohistochemistry or ISH positive for an infectious agent, specify:

Grade: _____________

Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

604.e23 American Journal of Obstetrics & Gynecology DECEMBER 2018

(continued)

ajog.org

OBSTETRICS

Original Research

SUPPLEMENTAL TABLE 1

Pathology data collection form (continued) Villitis of unknown etiology32,34 0, not present 1, low-grade, inflammation affecting <10 contiguous villi in any 1 focus, >1 focus * Denote focal (1 slide only) or multifocal (>1 slide) 2, high-grade VUE: inflammation affecting >10 contiguous villi, seen in multiple foci on >1 section *Denote patchy (multiple foci, 1 with >10 contiguous villi) or diffuse (>30% of all terminal villi involved) * With or without vascular damage

Grade: _____________

Chronic intervillositis 0, not present 1, infiltration of the intervillous space by histocytes, < 50% of the total placental intervillous volume 2, infiltration of the intervillous space by histocytes, > 50% of the total placental intervillous volume

Grade: _____________

Chronic deciduitis35 0, not present 1, present 1 or more, plasma cells present

Grade: _____________

Additional findings Meconium histiocytes/macrophages within membranes 0, not present 1, present

Grade: _____________

Meconium-induced myonecrosis 0, not present 1, present

Grade: _____________

Note any significant lesions observed that are not listed above CMV, cytomegalovirus; HSV, herpes simplex virus; ISH, in situ hybridization; NOS, not otherwise specified. Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

DECEMBER 2018 American Journal of Obstetrics & Gynecology

604.e24

Original Research

ajog.org

OBSTETRICS

SUPPLEMENTAL TABLE 2

Mean histological severity scores of each lesion category for each gene expressionebased cluster, by maternal hypertensive status Cluster 1

Cluster 2

Cluster 3

Cluster 4

Cluster 5

HTN Normo (n ¼ 21) (n ¼ 31)

HTN Normoa (n ¼ 51) (n ¼ 1)

HTN Normoa (n ¼ 10) (n ¼ 1)

HTNa Normo (n ¼ 1) (n ¼ 12)

HTN Normo (n ¼ 10) (n ¼ 4)

P value

Maternal vascular malperfusion (maximum)

2.48

0.61

4.78

1.00

2.60

1.00

2.00

0.33

4.00

0.50

< .01

Implantation site abnormalities

0

0

0.02

0

0

0

0

0

0

0.25

.01

Histological chorioamnionitis 0.14

0.26

0

0

0

0

0

2.33

0

0.75

< .01

Placenta villous maldevelopment

0.26

0.08

0

0.40

2.00

0

0.42

0.10

0.25

.03

Fetal vascular malperfusion 0.05

0

0.10

0

0.30

0

0

0.25

0

0

.02

Chronic uteroplacental separation

0.10

0.06

0.18

0

0.20

1.00

0

0

0

0.25

.46

Maternal-fetal interface disturbance

0.10

0.23

0.16

0

0.90

0

2.00

0

0.10

0.25

< .01

Chronic inflammation

0.24

0.29

0.14

0

0.50

0

1.00

0.17

0.10

0

Lesion category

0.05

.71

Maximum severity score by category are as follows: maternal vascular malperfusion (maximum score of 13), implantation site abnormalities (maximum score of 4), histological chorioamnionitis (maximum score of 11), placental villous maldevelopment (maximum score of 5), fetal vascular malperfusion (maximum score of 6), chronic uteroplacental separation (maximum score of 3), maternalfetal interface disturbance (maximum score of 5), and chronic inflammation (maximum score of 6). HTN, hypertensive; Normo, normotensive. a

Not included in the statistical analysis. Benton et al. Clinical subtypes of preeclampsia, placental gene expression, and histopathology. Am J Obstet Gynecol 2018.

604.e25 American Journal of Obstetrics & Gynecology DECEMBER 2018