Thread by @SVanharanta

We are very excited to share our new paper out today @Nature, in which we demonstrate that common kidney #cancer causing genetic alterations need transcriptional lineage factors for oncogene activation. Details in the thread👇 nature.com/articles/s4158…

Driver mutations are unevenly distributed across cancers, suggesting that their effects are tissue and cancer type specific, but how this specificity arises remains poorly understood.

An example of this is the tumour suppressor VHL which is inactivated in most clear cell renal cell carcinomas (ccRCCs), but not in other common cancer types.

Building on the genetic data and the idea that cell identity is defined by specific transcriptional programs, we set out to test whether cancer driver mutations need lineage transcription factors for oncogenic activity, a mechanism that could explain their tissue specificity.

In the context of ccRCC this meant testing whether a renal lineage specific transcriptional program is needed for the tumorigenic activity of the transcription factor HIF2A, a well-documented oncogenic mediator downstream of VHL loss in ccRCC.

We first used CRISPR/Cas9 screening to identify PAX8 and HNF1B as specific transcriptional dependencies in ccRCC. In parallel, we found by mass spec that PAX8 co-localizes with HIF2A in nuclear complexes. ChIP-seq confirmed PAX8/HIF2A co-localization on chromatin.

To test the functional relevance of these observations, we then identified pro-tumorigenic gene regulatory elements bound by HIF2A using an in vivo CRISPRi screen. The strongest hit was a CCND1 upstream enhancer, which we named E11:69419. E11:69419 bound both PAX8 and HIF2A!

CCND1 is an oncogene. E11:69419 regulated CCND1 expression and its activity was dependent on specific PAX8 and HIF2A binding sites in reporter assays. Strikingly, PAX8 binding at E11:69419 was inhibited by a RCC protective rs7948643 allele previously identified by GWAS analysis.

VHL mutations are frequent in ccRCC, but not in papillary RCC. However, papillary RCCs also express PAX8. As E11:69419 depended on both PAX8 and HIF2A for CCND1 expression, our results suggested that rs7948643 should be associated with ccRCC risk but not papillary RCC risk.

To test this we teamed up with Mark Purdue, @mitchiela and @NCIChanock to perform RCC subtype-specific re-analysis of a large RCC GWAS data set. As predicted, rs7948643 genotype was strongly associated with ccRCC but not papillary RCC!

We also found that expression of the oncogene MYC from ccRCC metastasis-associated 8q amplifications requires PAX8 and the downstream mediator HNF1B, which activates specific distal MYC enhancers in normal and malignant cells of renal origin.

So, different ccRCC-associated genetic alterations depend on renal lineage factors for oncogenic activity, which could explain their ccRCC specificity. Similar mechanisms could be applicable to other cancer types and driver genes as well.

The project was spearheaded by the amazing @saroor_patel with critical contributions from many members of our team and fantastic collaborators Jay Carroll, @grantissimus @shamith @NCIChanock @pauloagr @ShokoHirosue @LudoWesolowski @EvaPapachristou @VeroCaraf @AnnaDyas et al.

Thank you @BarbMarte for facilitating the publication process and the referees for helping us improve our work.

Finally, my team has just moved to @HelsinkiUniMed and we are recruiting, please get in touch if you are interested in joining us (a job ad will be released very soon).