NEW YORK – Researchers at UCSF have discovered genomic variants associated with spontaneous preterm birth (sPTB), as well as therapeutic compounds capable of influencing them.
In a study recently published in the journal Science Advances, Jingjing Li and his colleagues identified 99 genes strongly correlated with sPTB and nine compounds that show a potentially therapeutic effect on them.
Spontaneous preterm birth occurs when the thick smooth-muscle layer of the uterus, called the myometrium, prematurely shifts from a quiescent to a contractile state. The progesterone receptor normally regulates this quiescent state, blocking labor by suppressing proteins that trigger contractions, along with labor-related inflammatory factors.
Through a large-scale, machine learning-driven genomic and epigenomic analysis of the pregnant myometrium, Li and his colleagues identified 315 genes related largely to muscle contractility and inflammatory responses, most of which had not previously been described in the sPTB context.
"This is pretty exciting," Li said, "because for a long time people in the field always said, 'OK, preterm birth, that's all inflammatory responses.' But now, we see more direct evidence that it's the dysregulation of muscle contractility in the uterus."
Of those 315 genes initially flagged by researchers, the differential expression of 99 genes correlated most strongly with the onset of labor. A subset of 69 of those genes were largely associated with muscle contractility down-regulated and 30 genes were largely associated with inflammation up-regulated at labor onset. Importantly, the UCSF team analyzed and compared results from both European and African American cohorts in an attempt to make their findings more relevant to a broader overall population.
The UCSF researchers "take the next step in our understanding of sPTB by using an innovative machine learning model (DEEP+) to integrate uterus-specific epigenomes with existing sPTB [genome-wide association study] frameworks and providing deeper mechanistic insights into sPTB," Viral Jain, an assistant professor of pediatric neonatology at the University of Alabama, said in an email. Jain wasn't involved in the UCSF-led research.
DEEP+ predicts tissue-specific chromatin accessibility from the assay for transposase-accessible chromatin using sequencing (ATAC-seq –– a method for determining chromatin accessibility across the genome) data. The algorithm quantifies the likely effect that each genomic mutation has on predicted chromatin accessibility changes from a reference to an alternative allele, with higher scores indicating more significant changes in chromatin architecture resulting from regulatory mutations.
Li maintains that compared to more conventional association-based approaches, such as quantitative trait locus analysis, DEEP+ scores more directly quantify the mechanistic effects of genomic mutations.
"[We] look at each mutation's variants and ask if different alleles will compromise the regulatory activity, based on our deep learning approach," Li said. "Then, because we have the genetic data for each allele, we know how likely [that] allele is enriched in the patient population compared with a control cohort. We integrate all the information and can rank them based on possible functional consequences [of] changing gene regulation."
The next logical question after identifying the genes of interest, Li said, was to ask if any of them might serve as therapeutic targets for preventing sPTB.
There are currently few therapeutic options for preventing preterm labor. Prophylactic treatment with progestin, such as Covis' Makena (hydroxyprogesterone caproate injection), has been the only available treatment to date. Progestin therapy is shown to have heterogenous results, however, and it is not fully understood why. The US Food and Drug Administration, for instance, withdrew its approval for the drug last year, after a clinical trial failed to show that it reduced the risk of preterm birth and improved the health of the babies born to mothers who were treated with Makena.
Li and his colleagues conducted a study of 48 pregnant women with a history of sPTB. All participants received 250 mg of Makena beginning at 16 weeks until 36 weeks of gestation. Among these women, 28 were considered responders to the therapy and 20 were considered non-responders. Whole-genome sequencing revealed that a significant enrichment of regulatory mutations in muscle contraction-related genes among the non-responders compared to responders, suggesting that some of the heterogeneity seen in progestin response could be attributed to regulatory mutations affecting myometrial muscle contractility genes.
"This helps us develop a framework for future patients," Li said. "It's possible [that] we can first screen their genomes, see how likely they will respond, and only give medication to those people with confidence of being responders."
Because of the dearth of medications approved specifically for sPTB, however, Li and his team explored the possibility of repurposing drugs approved for other indications.
"A lot of them were cancer drugs," Li said. "No one really thought about how they would interact with regulating muscle contractility."
By computationally screening a library of over 4,000 FDA-approved drugs, clinical trial drugs, and preclinical compounds from the Broad Drug Repurposing Hub, the researchers identified nine small molecules predicted to have the greatest impact in attenuating the activity of these genes.
The team tested the effect of each drug on modulating muscle contractility in primary human uterine smooth muscle cells, comparing quiescent cell states to those mimicking labor-like phenotypes. Of the nine compounds, five –– bisindolylmaleimide IX, RKI-1447, 4,5,6,7-tetrabromobenzotriazole, LY294002, and URMC-099 –– each reduced myometrial contractions, whereas Pfizer's Bosulif (bosutinib), LY294002, and SB-203580 induced contractions, suggesting therapeutic potential to treat preterm labor.
"This is a really big thing," Li said. "There's no medication at all on the market right now, and [these findings] give us a lot of hope."
"The validation of [the UCSF team's] genomic result in the in vitro experiment and the strong correlation between the European and African American cohorts add to the strength of the study," Jain said.
Still, Jain cautioned that because the data were derived from term non-contracting uterus and characterized with genomic loci in sPTB, they don't necessarily identify genomic changes that may have started much earlier in pregnancy and led to preterm birth.
Nonetheless, Li said that the study gives him confidence that the effectiveness of progestin treatment should be reconsidered as a viable option targeted to those patients most likely to respond.
The method put forth in this study, Li added, opens up possible new ways of evaluating the effectiveness of investigational therapies undergoing clinical trials. "When we perform clinical trials," he said, "we essentially evaluate effectiveness based on population averages." The potential benefits of a drug on a population subset, Li argued, could get lost in that averaging.
The genes discovered by the UCSF team add to a pool of other potential prognostic biomarkers of preterm birth. Sera Prognostics, for example, has been investigating insulin-like growth factors binding protein 4 (IBP4) and sex-hormone binding globulin (SHBG) in an ongoing clinical study assessing the efficacy of the company's blood-based PreTRM Test in improving sPTB outcomes.
UCSF has filed a patent application covering the methods described in the recently published study, but Li said that it is too early to have any serious discussion about how to apply that patent going forward. "We haven't really gone to that level yet," he said.
Li and his team are now testing the nine drugs they've identified as potential sPTB therapies to confirm they can reprogram uterine smooth muscle contractility. They are also working with clinical collaborators at Stanford to retrospectively identify patient samples that previously had received progesterone therapy. "We hope to further expand the patient cohort," Li said, "so our model performance can be better estimated from a larger patient cohort."