Differential 5′-tRNA Fragment Expression in Circulating Preeclampsia Syncytiotrophoblast Vesicles Drives Macrophage Inflammation

The data that support the findings of this study are available from the corresponding author upon reasonable request.

This project was approved by the Central Oxfordshire Research Ethics Committee (07/H0607/74 and 07/H0606/148). All participants provided informed written consent. Preeclampsia was defined using the International Society for the Study of Hypertension in Pregnancy classification.¹⁸ Placentas were obtained at the time of cesarean section and perfused within 10 minutes of delivery. Uterine vein samples were taken during cesarean section, just before uterine incision, ipsilateral to the placental site. Peripheral blood samples were taken from the antecubital fossa; 21-gauge needles and 4.5 mL sodium citrate vacutainers were used for venepuncture (BD Diagnostics, UK). Nonpregnant samples were from female volunteers of reproductive age. Plasma was obtained by centrifugation at 1500g for 15 minutes. Samples were processed within 30 minutes of collection, aliquoted, and stored at −80 °C.

Placentas from 8 women with early onset preeclampsia and 6 normotensive pregnancies were perfused using a well-established dual-lobe perfusion technique.¹⁹ Maternal perfusate was centrifuged at 10 000g for 30 minutes to isolate medium-large EVs (MLEVs). The supernatant was centrifuged at 150 000g for 2 hours to isolate small EVs (SEVs). Biopsies of placental tissue were taken from the maternal surface of a nonperfused lobe. EVs were characterized using Nanoparticle Tracking Analysis, transmission electron microscopy and Western blotting as described previously.¹⁹

RNA was isolated from MLEVs, SEVs, and placental tissue using total RNA Purification Plus Kit (Norgen Biotek Corporation, Canada). After confirming RNA quantity and integrity using Bioanalyzer (Agilent Technologies, Germany), the same amount of input RNA was loaded for library preparation using the NEBNext Multiplex Small RNA library preparation kit (New England Biolabs). Libraries were size selected for fragments 15 to 50 bp by gel electrophoresis; fragment size and concentration were confirmed using high sensitivity D1000 ScreenTape (Agilent, UK). Single-end sequencing by synthesis was undertaken using an Illumina HiSeq 2500 machine (Illumina). One preeclampsia sample was removed from the SEV/placenta groups due to a technical issue.

Sequence reads were analyzed using sRNAnalyzer.²⁰ Briefly, sequencing adaptors and low-quality reads were removed using Cutadapt.²¹ All identical reads in sequence reads were collapsed, thus, generating a set of unique reads (referred to as fragment ID in this study). The number of sequence reads attributed to a fragment ID were defined as the raw expression level of such a fragment ID. The fragment IDs were mapped to the human small RNA databases allowing 2 mismatches. The human small RNA databases comprised miRNA, piRNA, snoRNA, rRNA, and tRNA.^20,22,23 The normalized expression level for each fragment ID within a class of RNA was used for downstream differential expression analyses. The normalized expression level for a fragment ID in a RNA class was calculated by dividing the raw expression level by a per million scaling factor of total reads of that RNA class, expressed as reads per million (RPM). Compared with healthy placentas, differentially expressed fragment IDs in the MLEVs of preeclampsia placentas were determined using the following criteria: (1) At least 1 MLEV sample had expression level of >100 RPM, (2) P value was required to be <0.05 based on Mann-Whitney U test after Benjamini-Hochberg correction. The in-house bioinformatics pipeline was written in Perl and R languages for counting sequence reads, tag annotations, and differential expression analysis. Sequence read archive data for blood cell datasets were downloaded from NCBI, with identifiers shown in Table S3.

Plasma aliquots were thawed at 37 °C and centrifuged at 3000g for 5 minutes to remove cryoprecipitates. For membrane-based affinity isolation of total vesicular RNA, 500 µL plasma was loaded onto exoRNeasy Midi columns (Qiagen, Germany). For magnetic-bead isolation of STB-EV RNA, 500 µL plasma was centrifuged at 10 000g for 30 minutes and the total EV pellet washed once before resuspension with biotin-saturated MojoSort streptavidin magnetic nanobeads (Biolegend) to deplete nonspecific binding. The supernatant was resuspended with nanobeads prebound with biotinylated in-house PLAP (placental alkaline phosphatase) antibody, known as NDOG2.²⁴ Bead-STB-EV complexes were washed 4 times before EV RNA isolation using Trizol LS (Invitrogen). RNA was stored in aliquots at −80 °C.

Custom Taqman stem-loop assays were designed for specific small RNA target sequences identified from RNA sequencing analysis (Applied Biosystems). Assay linearity and specificity were verified. Perfusion samples were normalized to TBP (confirmed empirically to be a suitable reference). Plasma samples were normalized to Caenorhabditis Elegans miR-39 spike-in.

Quantitative PCR (qPCR) assays are documented in Table S4. QuantStudio qPCR instruments (Applied Biosystems) automatically determined quantification cycles using standard settings; quantification cycles, >35 was considered undetectable. Relative expression was determined by following the Pfaffl approach, normalizing to median expression in the control group.

THP-1 cells (ATCC) were seeded onto 24-well Nunc plates at 50 000 cells per well in RPMI-1640 medium supplemented with 10% fetal calf serum. Cells were grown with 6.2 ng/mL phorbol 12-myristate-13-acetate for 24 hours to differentiate into macrophages. Transfection experiments were performed using RNA oligonucleotides (IDT) with a 5′-P modification; the sequence for tRF-A is shown in Table S2. The scramble RNA control was the most abundant STB-EV tRF without 5′-P modification and U nucleotides replaced with A (sequence 5′-GCAAAGGAGGAACAGAGGAAGAAAACACGCCA-3′). RNA (9.2 µmol/L) was packaged into lipid vesicles using N-[1-(2,3-Dioleoyloxy)propyl]-N,N,N-trimethylammonium methylsulfate, following manufacturer’s instructions (Roche, Switzerland) and in line with a prior publication.¹⁷ Human umbilical vein endothelial cells were purchased and seeded onto 24-well Nunc plates at 25 000 cells per well in endothelial cell growth medium 2 (PromoCell, Germany). Macrophage supernatant experiments were conducted by preparing 2X Growth Medium and mixing 50:50 with THP-1 supernatant. Cellular RNA was isolated using RNeasy Plus Mini Kit (Qiagen, Germany). Target mRNA expression was quantified and normalized to GAPDH. Preamplification was used to detect IL-12B using TaqMan master mix #4391128 with 10 cycles.

All data within this article are derived from distinct samples. Figures 3B and 4A and the graphical abstract were created using BioRender.com (Toronto, Canada). Statistical analyses and figures were generated using RStudio (RStudio) and Prism9 (GraphPad). Suspected outliers were excluded if over the 85^th centile. Unpaired 2-tailed Mann-Whitney U tests were used throughout unless specifically stated within the figure legend.

We isolated SEVs and MLEVs, alongside placental biopsies, from 8 placentas with early onset preeclampsia and 6 normotensive controls using dual-lobe placental perfusion. Pregnancy characteristics are shown in Table S1. We performed single-end small RNA sequencing (size selecting <50 nucleotides). Sequence length distribution plots confirmed the majority of reads in MLEVs and SEVs from both preeclampsia and normal placentas were 30 to 34 nucleotides long (Figure 1A). In contrast a peak at 22 nucleotides was greater in the placental samples from both groups. The majority of fragments in MLEVs and SEVs mapped to tRNA species, rather than to the ribosomal RNA and micro-RNA species seen in placental tissues (Figure 1B). Coverage plots demonstrated that 5′-tRFs (but not 3′-tRFs) constitute almost all tRNA reads in EVs in preeclampsia, as well as normal pregnancies (Figure 1C).

We used a bespoke bioinformatics pipeline to investigate differential expression of tRFs. To minimize data loss, we assigned each unique fragment an identifier, then annotated fragments after differential expression analysis (Methods). We identified 983 differentially expressed small RNA fragments in preeclampsia MLEVs compared with controls; 182 mapped to 5′-tRFs using GtRNAdb 2.0.²³ No fragments were found to be differentially expressed in preeclampsia SEVs.

The 12 most abundant differentially expressed fragments are shown in Table. Using the sum of preeclampsia MLEV normalized counts as a denominator, these 12 fragments account for 64% of the differentially expressed counts. Five percent of these counts were accounted for by the 626 least abundant fragments. Thus, a small number of abundant fragments represented the majority of the signal in an otherwise complex dysregulated small RNA profile in preeclampsia MLEVs.

Comparing to human small RNA databases using conventional mapping, among the 12 most abundant differentially expressed fragments, 3 were identified as 5′-tRFs and 4 as piwi-interacting RNAs. The remaining 5 had no directly matched reference sequences. Review of fragment sequences demonstrated substantial overlap; indeed all 12 showed only 1 or 2 nucleotide insertions or deletions differentiating them from known 5′-tRFs (Table). By adapting the post hoc mapping strategy to incorporate up to 2 insertions or deletions, it was evident that the majority of differentially expressed small RNA in preeclampsia are 5′-tRFs. A complex profile of differential expression in preeclampsia MLEVs was thus dominated by minor variants of a handful of abundant 5′-tRFs.

We sought 3 target 5′-tRFs that were differentially expressed in preeclampsia STB-EVs to validate our findings ex vivo (in placental perfusion samples) and in vivo (in plasma). 5′-tRFs are known to be expressed in other circulating EVs, most notably from immune cells. The most abundant EVs in plasma are blood-cell derived. Hence, we used publicly available blood-cell data sets to quantify possible target 5′-tRFs likely to be within contaminating EVs in plasma. Using these data we selected 3 tRFs for validation which were: abundant in STB-EVs (above 90th centile preeclampsia expression among 983 differentially expressed fragments), upregulated in preeclampsia, and of low relative abundance in potentially contaminating EV source cells (Table S2). The relative abundance of these 3 targets in STB-EVs using RNA sequencing is shown in Figure 2A (5′-tRF targets named A, B, and C for brevity but full sequences and tRNA derivations shown in Table S2).

We used custom small RNA assays with stem-looped reverse transcription primers and specific minor-groove binding TaqMan probes to compare relative 5′-tRF expression in STB-EVs obtained by perfusion using quantitative real-time polymerase chain reaction. Results validated target 5′-tRF upregulation and effect sizes were consistent with RNA sequencing findings (median 1.4-fold upregulated in preeclampsia; Figure 2B).

Total EV RNA was isolated from maternal plasma using membrane-affinity columns. EV size profiles for plasma were comparable to perfusion MLEVs (Figure S1). Extensive characterization of plasma membrane-affinity EVs has previously been published.²⁵ Target 5′-tRFs were significantly more abundant in pregnant peripheral plasma EVs than in nonpregnant matched control samples (median 3.0-fold higher in pregnancy; P<0.05 for all 3; Figure 3A). The difference between these sample groups suggested a pregnancy-specific load of 5′-tRFs in plasma EVs. The high abundance of these 5′-tRFs in nonpregnant samples confirms they were not unique to pregnancy (median quantification cycle values in nonpregnant samples: A 24.0, B 27.9, C 25.9). We then acquired paired plasma samples simultaneously from the uterine and peripheral veins of women without preeclampsia undergoing elective cesarean section for an indication unrelated to preeclampsia (eg, breech presentation) before delivery of the feto-placental unit. The uterine vein directly receives blood from the placenta, thus placentally derived molecules are more abundant in these samples (Figure 3B).²⁶ All three 5′-tRFs were more abundant in uterine vein samples (median, 1.3-fold; P<0.05 for all 3; Figure 3C). These data support a placentally derived load of 5′-tRFs in circulating plasma EVs.

A technique was optimized (Figure 4A) to isolate STB-EV RNA from maternal plasma, targeting the syncytiotrophoblast marker protein PLAP (placental alkaline phosphatase) Streptavidin nanobeads (130 nm diameter) were incubated with a highly specific biotinylated anti-PLAP antibody (NDOG2, in-house). PLAP+ plasma MLEVs were separated from soluble PLAP by centrifugation and NDOG2-nanobeads used to pull STB-EVs from total plasma EVs. Expression of miR518 (from the placental chromosome 19 miRNA cluster) was used to demonstrate assay sensitivity (Figure 4B). Perfusion STB-EVs were spiked into nonpregnant plasma as a positive control, achieving around 2000-fold greater miR518 expression than pregnant plasma. Nanobeads without antibody (saturated with free biotin) were added to pregnant plasma as a negative control; no miR518 expression was detected.

This technique was used to quantify the abundance of 5′-tRF-Glu-CTC (tRF-A) in peripheral plasma from a new cohort of 14 women with early onset preeclampsia and 12 gestation-matched normotensive controls (Table S5). 5′-tRF-Glu-CTC was upregulated (median, 1.4-fold; P=0.017) in preeclampsia plasma STB-EVs (Figure 4C). The effect size was comparable to preeclampsia perfusion STB-EVs (median, 1.4-fold upregulated).

Inflammation is a key feature of preeclampsia.²⁷ Tissue-resident macrophages are known to be activated in preeclampsia.²⁸ Recent studies suggest STB-EVs may underlie this activation.²⁹ We treated macrophages and monocytes in culture with the most abundant 5′-tRF upregulated in preeclampsia (5′-tRF-Glu-CTC, labeled A for brevity; full sequence in Table S2). RNA was packaged within otherwise undecorated lipid vesicles, to distinguish the effect of 1 tRF from the accompanying RNA, lipids, and proteins in perfusion-derived STB-EVs. We used a scrambled version of the most abundant STB-EV 5′-tRF sequence as a negative control, following validation against untreated cells (Figure S2). Macrophages were activated to a type 1 phenotype after 12 hours treatment with tRF-A, increasing expression of proinflammatory cytokines (Figure 5A). This proinflammatory 5′-tRF action was confined to macrophages, with no changes observed when the same experiment was repeated in undifferentiated monocytes (Figure 5B).

Endothelial damage is considered a common factor in the wide-ranging maternal organ dysfunction of preeclampsia.¹ Vascular macrophages are known to modulate adjacent endothelial cell function.³⁰ We treated human umbilical vein endothelial cells with vesicle-bound tRF-A (direct) and with the supernatants from tRF-A-activated macrophages (indirect) alongside corresponding scramble controls. tRF-A activated endothelial cells (quantified through increased expression of adhesion molecules) indirectly through macrophage stimulation, but not directly (Figure 5C).

The primary presenting feature of preeclampsia is high blood pressure. Immediate blood pressure regulation involves constitutive release of NO from endothelial cells, which maintains a state of relative vasodilation. In preeclampsia, circulating nitrite levels are reduced, which may contribute to hypertension.³¹ tRF-A reduced expression of eNOS (endothelial NO synthase) in endothelial cells, indirectly via macrophage activation (Figure 5C).

Preeclampsia is a multifactorial condition, with diverse clinical features. Syncytiotrophoblast stress is common to all cases, but remains poorly understood. Here we present 5′-tRFs, a novel and highly abundant class of RNA differentially released by the preeclamptic syncytiotrophoblast. We demonstrate placentally derived 5′-tRFs in the maternal circulation. The most abundant preeclampsia-upregulated STB-EV 5′-tRF was also increased in preeclampsia plasma. We find proinflammatory effects of this tRF on macrophages. Together these data suggest 5′-tRFs may play a role as transducers of an inflammatory signal from placenta to periphery in preeclampsia. Our findings offer a novel category of signaling molecule released by the placenta, warranting further investigation. We speculate that 5′-tRFs may dysregulate other cells in preeclampsia. Future studies will consider: other putative 5′-tRF targets in preeclampsia, such as liver sinusoids and pericytes; the actions of additional 5′-tRFs which are differentially expressed in preeclampsia; whether STB-EV 5′-tRFs play a role in other pregnancy-related diseases. Our ongoing work also focuses on the optimization of techniques to isolate low-abundance placental EV 5′-tRF signal from complex biofluids such as plasma, with attention to their clinical relevance. 5′-tRFs may join other better-studied stress markers such as sFlt-1 and PlGF in explaining the pathogenesis of preeclampsia.

The study was conceived and designed by W.R. Cooke, C. Redman, and M. Vatish Bioinformatic analyses were performed by P. Jiang, L. Ji, J. Bai, W.R. Cooke, and G.D. Jones. Supervision was provided by P. Jiang, Y.M.D. Lo, C. Redman, and M. Vatish. Experimental work was performed and article written by W.R. Cooke. All authors edited and approved the final article.

Figures S1–S2

Tables S1–S5