Congratulations... amazing!

We would especially like to thank all of our colleagues who collaborated with us on this, including @balynzaro.bsky.social and Albertas Navickas, for their contributions to this work. And our funders, including @arcinstitute.org.

Finally, we used double-knockdown for in vivo lung colonization assays to confirm the expected epistatic interaction between RBMS3 and TXNIP.

We confirmed that TXNIP is correlated with RBMS3, and similarly its expression is associated with better outcomes. Then we looked at TXNIP expression in a cohort of 96 breast cancer samples in-house, stratified by stage, that confirmed reduced expression as disease progresses.

To figure out which, if any, of the targets of RBMS3 that we identified are responsible for this phenotype, we performed an in vivo CRISPR screen. Among the targets tested, we found that silencing TXNIP increases metastatic lung colonization.

To turn this clinical association into causation, we took advantage of metastasis assays in xenografted mice. We showed that silencing and over-expressing RBMS3 modulates metastasis in vivo in two independent models.

Interestingly, this association was most notable in the basal and Claudin-low subtypes of breast cancer, which are known to be aggressive subtypes.

We consistently observed that reduced expression of RBMS3 is associated with poor clinical outcomes in breast cancer patients, in multiple datasets.

To gain insight into the functional consequences and phenotypic effects of this regulon, we looked at pathways and complexes that may be enriched among the RBMS3 target. We saw a significant enrichment for TGFb and VEGF signaling that are associated with metastatic progression in breast cancer.

Finally, we demonstrated the requirement and sufficiency of RBMS3 binding sites for this RBSM3-mediated transcript stabilization. We cloned binding sites from 13 high confidence mRNA targets into a reporter assay, which showed RBMS3-dependent increase in reporter expression.

Replacing AUA-carrying mRNAs with these empirical RBMS3 CLIP targets confirmed that RBMS3 direct binding drives transcript stabilization.

To show that this effect results from a direct interaction between RBMS3 and its target mRNAs, we performed CLIP-seq (UV crosslinking followed by immuno-precipitation and sequencing). We observed pervasive 3’UTR binding, and also a strong enrichment for the AUA element we had identified.

Silencing RBMS3, as expected, resulted in a significant decrease in the stability and expression of the mRNAs that contain AUA-elements in their 3’UTRs.

Among these, we chose RBMS3 for further characterization, simply because it was the highest ranking RBP for correlated expression with its putative regulon—identified based on the AUA cis-regulatory element, which was determined by analysis of GreyHound’s sequence importance scores.