Endogenous retroviruses (ERVs) can confer benefits to their host but present a threat to genome integrity if not regulated correctly. Here we identify the SWI/SNF-like remodeler SMARCAD1 as a key factor in the control of ERVs in embryonic stem cells. SMARCAD1 is enriched at ERV subfamilies class I and II, particularly at active intracisternal A-type particles (IAPs), where it preserves repressive histone methylation marks. Depletion of SMARCAD1 results in de-repression of IAPs and adjacent genes. Recruitment of SMARCAD1 to ERVs is dependent on KAP1, a central component of the silencing machinery. SMARCAD1 and KAP1 occupancy at ERVs is co-dependent and requires the ATPase function of SMARCAD1. Our findings uncover a role for the enzymatic activity of SMARCAD1 in cooperating with KAP1 to silence ERVs. This reveals ATP-dependent chromatin remodeling as an integral step in retrotransposon regulation in stem cells and advances our understanding of the mechanisms driving heterochromatin establishment.

Chromatin in embryonic stem cells (ESCs) differs markedly from that in somatic cells, with ESCs exhibiting a more open chromatin configuration. Accordingly, ATP-dependent chromatin remodeling complexes are important regulators of ESC homeostasis. Depletion of the remodeler SMARCAD1, an ATPase of the SNF2 family, has been shown to affect stem cell state, but the mechanistic explanation for this effect is unknown. Here, we set out to gain further insights into the function of SMARCAD1 in mouse ESCs. We identified KRAB-associated protein 1 (KAP1) as the stoichiometric binding partner of SMARCAD1 in ESCs. We found that this interaction occurs on chromatin and that SMARCAD1 binds to different classes of KAP1 target genes, including zinc finger protein (ZFP) and imprinted genes. We also found that the RING B-box coiled-coil (RBCC) domain in KAP1 and the proximal coupling of ubiquitin conjugation to ER degradation (CUE) domain in SMARCAD1 mediate their direct interaction. Of note, retention of SMARCAD1 in the nucleus depended on KAP1 in both mouse ESCs and human somatic cells. Mutations in the CUE1 domain of SMARCAD1 perturbed the binding to KAP1 and Accordingly, an intact CUE1 domain was required for tethering this remodeler to the nucleus. Moreover, mutation of the CUE1 domain compromised SMARCAD1 binding to KAP1 target genes. Taken together, our results reveal a mechanism that localizes SMARCAD1 to genomic sites through the interaction of SMARCAD1's CUE1 motif with KAP1.

Malignant melanoma is an aggressive form of skin cancer. Recently, drug therapy of advanced disease has been revolutionized by new agents. More therapeutic options, coupled with the desire to extend treatment to the adjuvant setting mean that prognostic biomarkers that can be assayed from formalin-fixed paraffin-embedded clinical would be valuable. microRNAs have potential to fill this need. We analyzed 377 microRNAs in 79 primary melanomas and 32 metastases using a split sample discovery strategy. From a discovery analysis using 40 thick primary melanomas (20 cases with metastasis and 20 controls without metastasis at 5 years), microRNA expression was measured by quantitative RT-PCR (QRT-PCR). MiR-10b emerged as a candidate prognostic microRNA. This was confirmed in an independent validation set of thick primary melanomas (20 cases with metastasis and 19 controls without metastasis at 5 years). In the combined discovery and validation cohorts (n = 79), miR-10b expression showed a 3.7-fold increase in expression between cases and controls (P = 0.005) and showed a trend of increasing expression between primary melanomas and their matched metastases (Po 0.001). In situ hybridization showed expression was in melanoma cells and correlated with expression measured by QRT-PCR (P = 0.0005). We used the combined discovery and validation samples to verify the prognostic value of additional candidate microRNAs identified from other studies, and proceeded to analyze miR-200b. We demonstrated that miR-10b and miR-200b showed independent prognostic value (P = 0.002 and 0.047, respectively) in multivariable analysis alongside known clinico-pathological prognostic features (eg, Breslow thickness) using a Cox proportional hazards regression model. Furthermore, the addition of these microRNAs to the clinico-pathological features led to an improved regression model with better identification of aggressive thick melanomas. Taken together, these data suggest that miR-10b is a new prognostic microRNA for melanoma and that there could be a place for microRNA analysis in stratifying melanoma for therapy.