Enrollment of women took place in Memphis TN 2005–2010 with informed written consent supervised by two co-investigators (RE and RR) under IRB approval from both the University of Tennessee Health Science Center in Memphis and the University of Kansas School of Medicine. All clinical records were extracted contemporaneously and entered into a computerized database. Entries were checked/rechecked for accuracy on a randomized basis. Investigators could request a review of the source records should an entry be suspect or a parameter missing. The records were finalized within 6 months of the last subject’s delivery and all patient identifiers removed from the working database. The identifying linkage and original files are maintained on a secure server at the University of Tennessee Health Science Center in Memphis under the control of one of the onsite investigators (RR). Further details on the full cohort will be part of a separate report.

Additional participants, whose deidentified samples of myometrium and plasma included in one or more of the studies presented, were obtained after informed written consent for biobank collection under protocols approved by the local IRB at either the University of Kansas School of Medicine or Yale University School of Medicine. The deidentified samples were transferred to and stored in the University of Kansas Pregnancy Biobank (PI CPW). There was no patient, funder or public involvement in the subsequently described research.

Overall, 305 pregnant women were enrolled at 18.3±1.4 weeks (range 16.0–20.9 weeks) and provided blood (serum, plasma [4mL vacuum tubes coated with 7.2mg K2EDTA], and buffy coat fractions), saliva, urine and cervical / buccal swabs. Thereafter, blood and urine samples were obtained every two weeks until delivery, when an umbilical cord blood sample was obtained. The final cohort included 289 participants (305 enrolled; 6 (1.9%) never sampled, 6 (1.9%) lost to follow-up and 4 (1.3%) clinical exclusions (1 “spontaneous” chorioamnionitis at 19 weeks, 1 PPROM 48 hours after elective cerclage placement at 24 weeks, 1 placental abruption at 28 weeks and 1 appropriately grown IUFD at 35 weeks).

Pregnancy dating was based on a reliable last menstrual period (LMP) when available, confirmed by a dating ultrasound conducted ≤20 weeks of gestation. If the two parameters differed by more than 11 days or there was no LMP, the ultrasound-derived gestational age was used.

No laboratory personnel were aware of pregnancy outcomes.

The 24 week plasma samples were used. Discovery subjects were selected at random from the prospectively assembled cohort. Criteria included a clear diagnosis and no reported maternal or pregnancy co-morbidities. Sixty percent were of White and 40% of Black ethnicity. Five sPTB≦32w subjects were selected in chronologic order of enrollment until the target sample size was achieved. Five control subjects were selected by the same process except they delivered >36 weeks and zero days. Prior preterm birth was not considered. PCF cDNA was run on the microarrays as described.

A primer for each selected miR was designed for quantitative (q)RT PCR and tested on the same cDNA samples used for Discovery. Two confirmed miRs were also quantitated across gestation in samples from the same Discovery subjects.

The as yet unconfirmed potential mRNA markers and confirmed miRs were compared to a list of myometrial Preterm Initiator genes that are differentially expressed in sPTB≦32 weeks¹⁷. The Metacore^™ database (Thomson Reuters, release 2010) was then queried to identify whether any selected potential RNA marker was known to interact with Preterm Initiator genes in any reported cell system. The mRNAs so identified were confirmed using qRT-PCR.

The plasma samples tested were obtained 16.0 to 19+5 weeks. Samples from women participating in validation were selected at random from the cohort after excluding Discovery subjects. Requirements included a clear diagnosis and no reported maternal or pregnancy co-morbidities. Ten subjects (50%) of each outcome group were of Black and 10 subjects of White ethnicity. Twenty (20) sPTB≦32 weeks women were selected in chronologic order of enrollment until the target sample size was achieved. The 20 control subjects were selected by the same process except they delivered >36 weeks and 0 days. Prior PTB was not a consideration. qRT-PCR was performed with the same primers used for Confirmation and expression normalized.

Of the 25,934 mRNA and the 847 miR sought, 296 mRNA and 14 miR were differentially expressed in women destined for sPTB≦32 weeks (Table s1). After reordering the 310 RNAs by narrowness of distribution (i.e., lowest variances among group subjects) and fold change, 99 RNAs (86 mRNAs and 13 miRs) were retained. Twenty-two mRNAs were increased and 64 decreased. Three miRs were increased and 10 decreased. None had previously been associated with pregnancy.

Four mRNA and 4 miR, whose expression on microarray was similar in both sPTB≦32 weeks and control groups, were tested across gestational age as potential normalization sequences. One mRNA (PPIA) (Figure s1d) and one miR (snRNA U6) (Figure s2d) fulfilled criteria.

Pathway analysis revealed the top two diseases associated with the retained 99 RNAs (Figure s3) were Hypertension and Neurologic / Neurobehavioral Manifestations. Obesity and Chorioamnionitis ranked 5^th and 6^th. The number one Map was DNA damage and the number one Network was the O-dexadecanoyl-(L)- carnitine pathway.

Ten of the 13 selected miRs were confirmed (Figure s4a) (confirmation rate 76.9%). The expression of miR 99a was increased by 16 weeks in sPTB≦32 weeks (p<0.01); miR 548L expression did not rise until after 20 weeks (consistent with microarrays) (Figure s4b).

The 10 confirmed miRs and 86 potential mRNA markers were then compared to a set of myometrial genes differentially regulated in women with sPTB≦32 weeks¹⁷. There were no commonalities between lists. However, the Metacore^™ query identified five (5) PCF RNAs (PSME2, NAMPT, APOA1, APOA4, and Hsa-Let-7g) that in silica interacted directly or indirectly with 7 Preterm Initiator genes (Figure 1a, b). The five RNAs originate from genes on autosomes 3, 7, 11 and 14. Figure 1c illustrates the mean RNA expression measured by qRT-PCR in Discovery samples. APOA1 approached but did not reach significance (confirmation rate 80%).

This is the first plasma transcriptome wide discovery / validation study for sPTB≦32w in the second trimester conducted without prior assumptions. The result was hundreds of potential PCF RNA markers whose expressions were significantly altered by both P value and fold change in women destined for sPTB≦32w. Candidate RNAs were progressively narrowed to a panel of 5 PCF RNA selected after in silica study revealed an interaction between them and 7 myometrial genes also differentially expressed in women with sPTB≦32 weeks.¹⁷ Four PCF RNAs were confirmed by qRT-PCR, and their expression levels measured in a 40 subject case-control validation using plasma obtained 16.0 to 19+5 weeks. sPTB≦32 weeks was associated with significant overexpression of the 4 confirmed RNA markers, suggesting the potential for a strong performance in future cohort studies.

We found after that after panel selection that the RNAs were highly expressed in placenta from women with sPTB≦32 weeks and associated with molecular pathways not previously linked to sPTB. Each marker in silica had the potential to increase intracellular calcium either directly or indirectly. Overexpression of APOA4 mRNA in IPHM increased APOA4 protein, intracellular calcium concentration, and cell contraction frequency. The increased contraction frequency was reduced by a siRNA blocking IFNGR, a predicted intracellular target of APOA4 (see Supporting Information, Discussion). While the remaining markers require similar testing, it may be that sPTB is preceded by myometrial quiescence dysfunction.

The present study has several strengths. 1. The samples were derived from a prospectively assembled cohort enrolled and first sampled at 16–20 weeks followed by sampling every 2 weeks. Patient follow-up was excellent. 2. It is the first study to utilize Rosetta Signaling Laboratory’s proprietary plasma RNA extraction which increases total RNA yield per milliliter of plasma to microgram amounts compared to the nanograms achieved with commercial kits³⁹. 3. We focused on sPTB≦32w as these neonates are at highest risk for mortality, and with survival, at highest risk for near and long-term complications that account for the majority of health care costs in the first 12 months of life^1,2. Further, selecting a gestational age cutoff ≦32 weeks should have reduced mixing of subjects whose labor was triggered by ill-timed ‘term’ labor mechanisms. 4. The discovery study specifically included PTB≦32 weeks preceded by PPROM rather than assuming PPROM prior 32 weeks is an independent event. 5. PCF RNA marker selection was unaffected by prior assumptions of previously proposed mechanisms (e.g. inflammation, progesterone metabolism or oxytocin regulation). 6. Our objective was to identify potential markers present before 20 weeks for sPTB≦32 weeks regardless of race. However, other potential PCF RNAs confirmed and validated could prove of future interest specifically in terms of racial differences in rates of sPTB≦32 weeks. 7. The last step in the selection of the PCF RNAs for validation was the result of a novel strategy- identification in-silica of potential PCF markers that interact with one or more myometrial genes differentially regulated in women who suffered sPTB≦32 weeks.¹⁷

There are several limitations to the investigation. 1. The studies were conducted in 2010–2012 and many of the discovery tools widely available today were not widely available then. The delay in publication is unfortunate, but does not undermine the well-studied validity of microarrays. It is possible that newer tools like RNAseq could identify other, relevant RNAs not part of the arrays used, but that in no way diminishes the potential relevance of the PCF RNA markers identified. 2. The discovery sample size is small, certainly based on the cumulative experience with genomics. Yet, the parameters for the bioinformatic analysis were strict and achieved. Further, the plasma transcriptome is a direct reflection of genome activity rather than genomic potential and our markers were selected for narrow variance among cases. Ngo⁴⁰ et al described the application of RNAseq to 15 women sampled at 24–30 weeks, 8 of whom experienced sPTB (26.4 ± 2.3 weeks) and all sampled ≦7 days from delivery. More recently, Del Veccio⁴¹ et al conducted a plasma transcriptome discovery study on 26 pregnant women. Perhaps just as important as sample size is the mechanism of disease. In addition, the final markers selected were ultimately based on their potential to interact with a group of myometrial RNAs differentially expressed in women with sPTB≦32 weeks and survived their initial validation. While conventional wisdom holds that sPTB≦32 weeks is a syndrome with multiple causes⁴², the fact it occurs at similar rates worldwide² suggests the number of causes may be limited. Been⁴³ et al examined the impact of COVID-19 mitigation on preterm birth in the Netherlands. In a million plus births during 4 months in 2020, they found a significant drop in PTB between 32 and <37 weeks within 2 months of initiating mitigating measures that lasted at least 4 months. Yet there was no significant change in the rate of PTB<32 weeks suggesting early and late PTB have different causes. 3. The case-control initial validation study was subject to selection bias inherent to this study design. However, the analytic approach avoids the risk of overfitting with regression analysis and the strong association of increased PCF RNA expression with sPTB≦32w by 16w supports our working hypotheses. 4. The cohort was enriched with subjects at risk for PTB either because of obstetric history or race, and the identified markers may not be applicable to the ‘general’ population. Subsequent cohort studies are needed to resolve this concern.

A large body of work confirms labor and cervical softening/shortening are associated with inflammation⁴¹. Yet, the molecular pathways leading to inflammation differ in term and preterm labors¹⁷. The association of PTB with the increased expression levels of the four PCF RNAs begins by 16 weeks and suggests a relationship with implantation or aberrant placental maturation, and that the inflammation characteristic of labor at any gestational age is a terminal event generated along multiple pathways. Our first corollary proposes the earlier in disease development a marker is present, the more likely the marker is to be part of the underlying disease and not a byproduct of secondary events. The association of the differentially expressed PCF RNA panel in this initial validation study suggests sPTB≦ 32 weeks is a placental disease potentially detectable by 16 weeks. If true, sPTB≦32 weeks may be amenable to early screening and ultimately to therapeutic intervention.

We note the work of Ngo³⁹ et al above. In that same paper, they measured 140 preselected PCF RNAs weekly in 31 Danish women from <12 weeks until delivery. Multiple PCF RNAs were differentially expressed across gestation, enabling them to predict with good accuracy the number of weeks between sampling and delivery. They then applied the Danish model to two USA groups of women at high risk for sPTB. The first group was described above: 15 Black women from Pennsylvania sampled 24–30 weeks, 8 delivered prematurely. The second group consisted of 23 Black women from Alabama sampled at 24 weeks, 5 delivering prematurely (30.6 ± 2.4 weeks) and 18 at term (38.7 ± 0.5 weeks). In support of our hypothesis on the origin of sPTB≦32 weeks, the Danish model correctly predicted the term deliveries but not the PTBs, suggesting the mechanisms of labor differ. Ngo et al then performed RNA-seq on the Pennsylvania samples. Thirty-eight RNAs were differentially expressed and a panel of 7 selected RNAs tested in the Alabama samples in a case control model and obtained a regression derived AUC 81%. As their PTB samples were obtained within 7 days of delivery, their markers may reflect elements of advanced disease and explain why none of their identified RNAs overlapped with the RNAs identified in the current study.