Pan-viral metagenomic sequencing demonstrates that cryptic viral infection is rarely observed in villitis of unknown etiology

Abstract

Introduction:

Aberrant maternal immune responses are implicated in villitis of unknown etiology (VUE), but the underlying cause of this loss of tolerance, including cryptic causative or precipitating infections, has been difficult to define. Herein, we performed pan-viral metagenomic sequencing of placentas with VUE to investigate the possibility of cryptic viral infection as a contributing factor in this inflammatory pathology.

Methods:

Placentas evaluated at a single tertiary medical center between 2010–2024 were included in this study. Overall, the cohort included infectious villitis due to cytomegalovirus (CMV; n=4), VUE (n=25), and a reference group composed of pathologically unremarkable placentas (n=17). Total nucleic acid was extracted from formalin-fixed paraffin embedded (FFPE) placental tissues and subjected to pan-viral metagenomic sequencing (PVMS) to identify viral-associated reads.

Results:

PVMS detected reads mapping to CMV in 4 (of 4) of CMV cases. For VUE cases, 22 (of 25) had no identifiable viral reads, while 1 case demonstrated CMV reads and two had papillomavirus reads. The control samples demonstrated no identifiable reads in 13 (of 17) samples, while 3 cases had reads mapping to human papillomavirus 16 and one case had reads mapping to human Herpesvirus 6.

Discussion:

Utilizing PVMS, we did not identify cryptic viral sequences in 88% of morphologic VUE cases. In one clinical VUE case, CMV sequences were identified, suggesting a misclassification of infectious villitis. Both papillomavirus and herpesvirus sequences have previously been identified in the placenta, with unknown clinical significance. Overall, these findings exclude active viral infection as a potential etiology of VUE.

Introduction

Chronic villitis (CV) is characterized by the infiltration of maternal lymphocytes into the placental villous stroma, with resultant villous injury [1]. The vast majority of chronic villitis cases are termed villitis of unknown etiology (VUE), a reference to the yet determined cause of inflammation. In a minority of CV cases (<5%), an infectious etiology is evident, with a variety of pathogens implicated in causing placental inflammation, including cytomegalovirus (CMV) [2], herpes simplex virus (HSV), varicella-zoster virus (VZV) [3], rubella [4], Toxoplasma gondii [5], and Treponema pallidum [6]. Establishing a mechanistic understanding of VUE origin is critical given the common nature of the disorder, affecting 15% of term placentas with a recurrence risk of ~30–55% in subsequent pregnancies [1–3].

Determining the underlying pathogenesis of the VUE remains an ongoing challenge. Two theories have been proposed to explain the origin of this pathology: (1) aberrant maternal immune response to the fetal allograft or (2) undiagnosed infection [7]. The current prevailing notion of VUE is a maternal alloimmune response similar to organ transplant rejection. A decrease in receptors supporting immune tolerance on trophoblast cells, upregulation of major histocompatibility complex and release of T cell homing chemokines has been reported which promote the infiltration of the villous stroma by maternal CD8+ cytotoxic T cells [7–9]. Yet, viral infections can also lead to similar morphology [10]. Current methods of viral detection including immunohistochemistry, in situ hybridization, and/or PCR, are targeted to a limited number of known pathogens, and may not be fully utilized during clinical review in the absence of viral-specific inclusions or cytopathic effect [11, 12].

In 2021, Ernst et al published a case control study, utilizing a novel viral metagenomics sequencing strategy, which identified viral reads mapping to the herpesvirus family in approximately 50% of VUE cases versus 10% of controls [8]. While these findings were preliminary in nature, they raised the possibility that cryptic viral infections may be contributing more commonly than appreciated in at least a subset of VUE. Therefore, our goal was to build on these findings further by investigating potential DNA and/or RNA viral causes of VUE using pan-viral capture-based metagenomic shotgun sequencing (PVMS) including well characterized cases from our institution diagnosed with infectious and non-infectious CV.

Methods

Case Selection

This study was approved by the Mayo Clinic Institutional Review Board (IRB 20–012379) and utilized clinical residual tissues from patients who provided Minnesota Research Authorization. Cases and controls were selected by searching the pathology database from 2010 to 2024 for infectious chronic villitis identified by immunohistochemistry (IHC), high-grade chronic villitis without an identified etiology (i.e., VUE), and those placentas without significant pathology from the same timeframe to serve as controls. All cases and controls were re-reviewed by a single pathologist using the Amsterdam Diagnostic Criteria [9]. Clinical information including maternal and gestational age, placental weight and fetal sex was abstracted from the electronic medical records.

Pan-viral metagenomic sequencing

Formalin fixed paraffin embedded (FFPE) blocks were obtained, and 5 × 5-micron scrolls were cut and placed into a 1.5 mL microfuge tube. Total nucleic acids were extracted and purified from each sample using the Quick-DNA/RNA^™ FFPE kit following the manufacturer’s standard instructions (Zymo Research, Irvine, CA). Using the sample workflow published by Twist Bioscience (San Francisco, CA), RNA (3.3 ng/uL) from each sample was mixed with random primer 6 (New England Biolabs, Ipswich, MA) and annealed for 5 minutes at 95 °C. cDNA synthesis was completed using ProtoScript (New England Biolabs) and NEBNext Second Strand Synthesis reagents (New England Biolabs). 50 ng of cDNA from each sample underwent fragmentation, end repair and dA-tailing under the following conditions: 37°C for 20 minutes, 65°C for 30 minutes and 4°C on hold.

The fragments were ligated to Twist Universal Adaptors (Twist Bioscience) and purified to generate libraries for indexing. Each sample was tagged with a unique UDI primer and amplified on a thermocycler: 98°C for 45 seconds, 8 cycles of 98°C for 15 seconds, 60°C for 30 seconds and 72°C for 30 seconds, followed by 72°C for 1 minute and 4°C hold. The amplified libraries were purified with magnetic beads and quantified via Qubit (ThermoFisher Scientific, Waltham, MA). Eight samples with unique UDI primers were pooled together (187.5ng each) and dried. The Comprehensive Viral Research Panel (Twist Bioscience), containing over 1 million 120 base pair probes complementary to pathogenic human and animal viruses, was added to the pools and hybridized at 70°C for 16 hours in a thermal cycler. Streptavidin beads were then used to bind hybridized targets and purify the pools. A post capture PCR amplification step combining the target bound streptavidin beads with KAPA HiFi HotStart Ready Mix (Roche, Indianapolis, IN) was run under the following thermal cycler conditions: 98°C for 45 seconds, 15 rounds of 98°C for 15 seconds, 60°C for 30 seconds, and 72°C for 30 seconds, followed by 72°C for 1 minute and a 4°C hold. PCR products were further purified using magnetic beads prior to validation and quantification of the generated library by a Bioanalyzer (Agilent Technologies, Santa Clara, CA) and Qubit (ThermoFisher).

Metagenomic Sequencing

Viral capture enriched libraries underwent Illumina sequencing (Illumina, San Diego, CA). Using a NovaSeq SP flow cell, 2 pools (total of 16 samples) were loaded per flow cell and subjected to 2 × 150 bp paired-end sequencing with a target minimum of least 1 million reads per sample.

Metagenomic Analysis

Sequencing reads were analyzed with the Chan Zukerberg ID (CZ-ID) platform [10], using analysis pipeline v8.3 with NCBI database Index Date: 2024–02-06. Analyses were filtered using CZ-ID recommended filtering criteria, including nucleotide (NT) bases per million > 100, non-redundant protein (NR) bases per million > 1, nucleotide alignment length (L) > 200 bp, and NT and NR expect values (E-value) < 0.001. For human papillomavirus reads, the viral type of the best aligned reference genome was accepted as the viral type.

Results

A total of 46 placentas were selected for the study, including 25 cases with a clinical diagnosis of high-grade chronic villitis of unknown etiology (VUE), 4 cases with a clinical diagnosis of CMV villitis (confirmed with CMV IHC stain), and 17 control cases without a pathologically significant diagnosis (16 of the 17 controls were previously reported in Norgan et. al [11]). A subset of VUE cases had other concurrent placental pathology (Supplemental Table 1), but high-grade VUE was a major finding in each case. Twelve of the selected controls were from placentas that did not undergo clinical histopathologic examination; for this subset only maternal age and gestational age at delivery were available. Representative histology from CMV villitis and VUE cases is shown in Figure 1. A single VUE case had CMV identified by metagenomic analysis. On review, this case demonstrated high-grade chronic villitis and rare foci of plasma cell villitis but was classified as VUE due to negative immunohistochemical staining for TORCH pathogens (including CMV). This case was reclassified to the CMV cohort for subsequent analyses.

Figure 1. Representative micrographs of CMV (infectious) villitis and villitis of unknown etiology (VUE) cases.

Panels (A-D) show features of CMV villitis. (A) CMV inclusion (arrowhead) within a chorionic villus with chronic villitis; (B) Plasma cell (arrows) villitis seen in a VUE case which was CMV-positive by metagenomics; (C) Plasma cell (arrows) villitis and hemosiderin deposition (arrowhead); (D) Positive immunohistochemical stain for CMV (arrows); (E) Cluster of chorionic villi with vascular obliteration (arrows) due to VUE. Scale bar represents 0.1 mm.

Clinical characteristics are shown in Table 1. In general, the maternal age of the subjects was not significantly different between groups (p=0.219); however, patients with CMV during pregnancy were generally younger. Additionally, those in the CMV group had a median gestational age at delivery of 34 weeks (p=0.002) compared to VUE (38.5 weeks) and control (39 weeks) participants. As such, median fetal weight at delivery was significantly lower in the CMV infected neonates (1830 grams) and VUE neonates (2790 grams) compared to controls (3680 grams; p=0.015). There were no significant differences in the proportions of male and female fetuses included per group (p=0.153), but the control group had an over-representation of male fetuses while the CMV group was skewed to more female fetuses. Despite changes seen in fetal weights, placental weights were not significantly different between the CMV (265 grams), VUE (426.6 grams) and control groups (550 grams; p=0.132). However, this corresponded with significant differences in the expected placental weights (p=0.002), resulting in 35% of the VUE group having small for gestational age (SGA) placentas, 52% demonstrating appropriate size for gestational age (AGA) placentas and 13% of placentas being considered large for gestational age (LGA). The controls were mainly AGA while the CMV group consisted of 60% SGA and 40% LGA placentas.

Table 1:

Clinical features of study cases

	VUE (n=24)	Control (n=17)†	CMV (n=5)*	p-value
Maternal age, years 1	29.5 (23–41)	28 (26–39)	25 (19–32)	0.219
Gestational age at delivery, weeks 1	38.5 (19–40)	39 (38–40)	34 (31–36)	0.002
Fetal weight, grams 1	2790 (430–3940)	3680 (3420–3860)	1830 (1040–3460)	0.015
Birthweight centile 1	18.9 (0.3–93.1)	72.5 (48.4–100)	18.5 (0.3–94.9)	0.157
Fetal sex 2				0.153
Female	10 (43%)	1 (6%)	4 (80%)
Male	13 (54%)	4 (24%)	1 (20%)
Unavailable	1 (3%)	12 (70%)	0 (0%)
Placental weight, grams 1	426.5 (62–703)	550 (431–667)	265 (170–676)	0.132
Placental centile 2				0.002
Small for gestational age (<10 percentile)	8 (35%)	1 (6%)	3 (60%)
Appropriate for gestational age (10–90th percentile)	12 (52%)	16 (94%)	0
Large for gestational age (>90 percentile)	2 (13%)	0	2 (40%)
Live birth2	21 (88%)	17 (100%)	5 (100%)	0.230

¹Kruskal-Wallis test with Dunn’s correction; median (range)

²Chi-Squared test; percentage

^*One case was reclassified from VUE to CMV based on sequencing results

^†Twelve control cases were selected from pregnancies that did not meet criteria for pathology exam (true normal controls), limiting available obstetrical data on these pregnancies

PVMS and metagenomic bioinformatic analyses were used to assess the presence of viral sequences in placental tissues (Table 2). CMV (i.e., human betaherpesvirus 5) sequences were detected in the four known CMV villitis placentas and one VUE case, with viral specific reads ranging from 360,360 reads (3876.8 reads per million [rPM]) to 14,301,797 reads (206,163 rPM) in the CMV cohort cases. The CMV PVMS-positive but IHC negative VUE case, (Case 5) which was reclassified CMV villitis, had 46,234 reads (1153 rPM). In the remaining VUE cohort, two cases (out of 25) had viral associated reads above threshold, corresponding to human papillomavirus 107 (12,488 reads; 104 rPM) and human papillomavirus 16 (4,164,537 reads; 100038 rPM). The HPV 107 case (Case 19) was that of an intrauterine fetal demise at a gestational age of 26-weeks in the setting of high-grade chronic villitis and a vaginal delivery. The HPV 16 case (Case 29) corresponded to a liveborn male infant delivered vaginally at 36 weeks gestation with high-grade chronic villitis and negative TORCH immunostains. To further evaluate HPV16 infection within the placental tissue, in situ hybridization for high-risk HPV and HPV mRNA transcripts E6/E7 were performed, with negative results (data not shown). The remainder of the VUE cases were negative for viral associated reads. For the control cases, as previously reported [11], three (out of 17) demonstrated reads mapping to HPV 16 and one case mapping to HHV-6. All seventeen control cases had uncomplicated pregnancies and no significant findings identified by placental pathology evaluation as defined by Amsterdam Criteria [9].

Table 2:

Viruses detected from placental tissues by PVMS

Case	Cohort	Virus Detected	Reads	rPM	Contigs	Sequence Identity (%)
1	CMV	Cytomegalovirus	360,360	3876.8	136	99.3
2	CMV	Cytomegalovirus	4,256,562	61613.9	2369	99.3
3	CMV	Cytomegalovirus	14,301,795	206163.5	1124	99.3
4	CMV	Cytomegalovirus	4,138,789	94384.1	223	99.7
17	VUE*	Cytomegalovirus	46,234	1153.9	95	99.5
19	VUE	HPV107#	12,488	104.2	13	98.1
29	VUE	HPV16#	4,164,537	100038.2	4404	99.2
37	Control**	HPV16#	8,866	222.1	9	99.4
39	Control**	HPV16#	23,316	629.3	6	99.5
40	Control**	HPV16#	3,724	115.6	9	99.1
45	Control**	HHV-6	4,893	124.2	140	99.3

^*Reclassified as CMV in Table 1.

^**Previously reported in [11].

^#HPV type inferred from the closest alignment to NCBI Human papillomavirus sequence entries.

Abbreviations: Pan-viral metagenomic sequencing (PVMS); Human papillomavirus (HPV); Human herpesvirus 6 (HHV-6); Reads per million reads (rPM)

In the initial analysis, only one herpesvirus, HHV6, was detected in a control case and no non-CMV herpesviruses were found in CMV or VUE cases. To further evaluate the presence of low level herpesvirus reads, we re-analyzed the data with the following read thresholds: nucleotide (NT) bases per million > 1, non-redundant protein (NR) bases per million > 1, nucleotide alignment length (L) > 75 bp, NT and NR expect values (E-value) < 0.001, and a filter for Herpesviridae (i.e., Human alphaherpesvirus 1/2/3, Human gammaherpesvirus 4, and human betaherpesvirus 5/6/7). In this secondary analysis, we identified reads mapping Herpesviridae in 4 (of 17) control cases and 12 (of 24) VUE cases, Supplementary Table 2.

Discussion

CV is a common inflammatory pathology observed in 10–15% of placentas at term and features the infiltration of T lymphocytes in the chorionic villi. Overall, CV is a diagnosis of exclusion; approximately 5% of CV diagnoses are attributable to infection whereas the other 95% are deemed VUE [1]. High-grade VUE, demonstrating diffuse lymphocyte infiltration across the tissue, is associated with adverse pregnancy outcomes and an increased risk of recurrence during a subsequent pregnancy, making this a clinically relevant finding [12–15]. Yet, management of VUE remains a challenge as it currently cannot be identified until after delivery, and there are no clinical trials to suggest effective ways to treat it. To move towards clinical solutions, we must first understand the origin and mechanisms the lead to the establishment of this pathology.

VUE is hypothesized to be caused by aberrant maternal immune response to the fetal allograft or, less likely, cryptic infection. While a few studies have suggested cryptic viruses or bacteria can cause VUE [8, 16], majority of data supports the maternal immune hypothesis. For example, C4d immunostaining of syncytiotrophoblast cells is evident in placentas with VUE, which would indicate the establishment of an antibody-mediated rejection response within the tissue [17, 18]. Placentas with VUE also demonstrate an upregulation of classical major histocompatibility complex (MHC) class I and II, indicating antigen presentation and activation [19]. Furthermore, T cell receptors from cases of VUE and CMV villitis are distinct, with unique clonotypes that (unlike CMV T cells) do not recognize common viral antigens [20]. Additionally, expression of checkpoint receptors on trophoblasts and infiltrating lymphocytes are altered differently in CMV villitis versus VUE [21], suggesting distinct mechanisms for infectious and VUE pathology.

In a previous pan viral metagenomic analysis of VUE tissues, herpesvirus sequences corresponding to CMV, EBV, HSV-1, HHV-6B and HHV-7 were detected in 50% of VUE cases versus 10% of controls [8]. In our primary analysis of VUE cases in this study, we detected no non-CMV Herpesviridae in any VUE cases, and a single case of HHV-6 in the control group. Reanalysis of the data presented here with relaxed bioinformatics thresholds did show a trend towards increased detection of viral reads corresponding to Herpesviridae in VUE cases (12/24; 50%) compared with controls (4/17; 24%; p=0.0514); however, reads detected in the secondary analysis (mean reads = 294; median reads = 122) were several orders of magnitude lower in abundance that those detected in clinical CMV or other primary analysis cases (mean reads = 2,483,778; median reads = 46,234), raising a question regarding clinical significance. The detection of low-level Herpesviridae reads in peripheral blood of healthy individuals has been previously reported and thought to be largely incidental due to the persistent nature of Herpesviridae infections [22]. Lymphocytes can serve as host cells of EBV (typically B lymphocytes) and T lymphocytes can host CMV, HHV-6 and HHV-7. HSV-1/2 and VZV, by contrast, are neurotropic but may transiently cause viremia during primary infection and clinical or subclinical reactivation. One potential explanation of the trend towards increased detection of low-level Herpesviridae reads in this study (read abundance was not provided in the work by Ernst et al. [8]) is the increased overall abundance of lymphocytes in VUE cases. Another possibility is that the inflammatory milieu of VUE cases correlates with increased expression of latent Herpesviridae. While it is formally possible that the low levels of Herpesvirdiae reads correlate with VUE causation in a subset of cases, this seems a less likely possibility than the detection of low abundance Herpesviridae reads being incidental to the actual cause of VUE.

HHV-6 was detected in a single control case in the primary analysis. The clinical significance of this finding in this case is unclear as morphologically the placenta was without evident histopathology and clinically the pregnancy was without serious complications. HHV-6 has been previously detected in placental tissues, with connections to adverse outcomes in pregnancy proposed, but not fully supported [23, 24]. In at least a subset, and perhaps a majority, of congenital HHV-6 cases, the ability of HHV-6 to chromosomally integrate has been implicated in placental infection and vertical transmission. While much remains to be understood about the role of HHV-6 in pregnancy and neonates, as it relates to this study, HHV-6 is not correlated with the presence of VUE pathology.

As in prior directed studies and in the metagenomics analysis by Ernst et al [8, 25, 26], human papillomavirus sequences were identified in a subset of placentas, including three HPV16 detections in the control group, as well as one HPV16 and one HPV107 in the VUE group. The significance of HPV detection in the placenta remains unclear; however, HPV has been associated with a variety of adverse outcomes inclusive of congenital transmission [25]. We were unable to detect HPV by HPV in situ hybridization on sections of the VUE case with the highest read abundance of HPV. Despite this, we believe the results of this study support prior observations demonstrating HPV within the placenta. Given that HPV studies in the placenta are rarely (if ever) performed clinically, the incidence of HPV within the placenta is likely underestimated.

In conclusion, we find that the vast majority of VUE diagnosed placentas are negative for viral infection by PVMS. This work helps to exclude the possibility of cryptic viral infection as a direct cause of morphologic VUE; however, the possibility that VUE results from a transient infection earlier in gestation cannot be directly addressed by this study. This work also supports prior observations that viruses of the family Herpesviridae and Papillomaviridae infect the placenta, although additional studies are required to understand the clinical significance of this infection.

Abbreviations:

VUE

villitis of unknown etiology

CMV

cytomegalovirus

PVMS

pan-viral metagenomic sequencing

rPM

reads per million

Data Availability

Data is available through the Sequence Read Archive (SRA) under accession PRJNA1332810.