E3 ubiquitin ligase HECTD2 mediates melanoma progression and immune evasion

The ubiquitin-proteasome system maintains protein homoeostasis, underpins the cell cycle, and is dysregulated in cancer. However, the role of individual E3 ubiquitin ligases, which mediate the final step in ubiquitin-mediated proteolysis, remains incompletely understood. Identified through screening for cancer-specific endogenous retroviral transcripts, we show that the little-studied E3 ubiquitin ligase HECTD2 exerts dominant control of tumour progression in melanoma. HECTD2 cell autonomously drives the proliferation of human and murine melanoma cells by accelerating the cell cycle. HECTD2 additionally regulates cancer cell production of immune mediators, initiating multiple immune suppressive pathways, which include the cyclooxygenase 2 (COX2) pathway. Accordingly, higher HECTD2 expression is associated with weaker anti-tumour immunity and unfavourable outcome of PD-1 blockade in human melanoma and counteracts immunity against a model tumour antigen in murine melanoma. This central, multifaceted role of HECTD2 in cancer cell-autonomous proliferation and in immune evasion may provide a single target for a multipronged therapy of melanoma.

HECTD2 or its mutant were generated by transduction with VSVg-pseudotyped viral particles harbouring Flag-tagged HECTD2 or HECTD2 C743A . Cells were FACS-purified by GFP expression twice, 72 hours and one week post transduction, on the Avalon, Propel Labs/Bio-Rad. Braf V600E HECTD2-deficient cells were generated by transfection with the px459 plasmids containing guide RNAs 268 or 271 and 382 or 383 and the repair template which was generated via PCR using the pUC57 plasmid described above with the RANGER DNA polymerase (Meridian Bioscence) and the following primers: forward 5'-GATCTGTCCATCTGTAACTTAT-3', reverse 5'-GTGTGTCATTGCTGTCCAAAT-3'. Before transfection the amplicon was purified with the QIAquick PCR purification kit (Qiagen). After 48 hours post transfection the cells underwent 1 µg/ml puromycin selection for three days before FACS sorting the cells for GFP expression on the Avalon. HCmel31 and HCmel31.Hectd2 cells expressing FB29 were generated by transduction with VSVg-pseudotyped viral particles harbouring FB29, and after 72 hours and one week post transduction were FACS sorted for envelope expression using the F-MuLV Env gp70-specific mAb 720 on the Avalon. All murine cell lines were verified for their origin and were mycoplasma free. IGR-1, HEK293T and U937 human cancer cell lines underwent an authentication process which included species identification and STR profiling performed at the Cell Services facility at the Francis Crick Institute. Certain cell lines were additionally treated for the indicated times with 200 µM of the HECTD2 inhibitor BC-1382 (Sigma-Aldrich). Cell growth was additionally assessed using an AlamarBlue (Invitrogen)-based assay, according to manufacturer's instructions. AlamarBlue fluorescence was read with a TECAN Spark plate reader (TECAN) and was plotted as arbitrary fluorescence units. All the transplantable cancer lines were passaged for a maximum 18-24 times or kept for a maximum of 8 weeks in culture. To monitor the levels of inflammatory mediators in melanoma upon HECTD2 overexpression, one million melanoma cells were plated for 24-48 hours in 6-well plates in 2 ml of culture medium supplemented or not with 10 µg/ml LPS, 0.5 µM LPA, 20 ng/ml TNF-α, or 100 U/ml IFN-γ. 48 hours post treatments the cells were collected and the levels of canonical and non-canonical NF-κB targets were measured via RT-qPCR, using the following primers:

Gene synthesis, cloning and mutagenesis
Gene encoding flag-tagged murine or human HECTD2 were synthesised and cloned into pRV-IRES-GFP vector upstream of the internal ribosome entry site (IRES) sequence. The C743A mutation was introduced the murine gene using QuickChange site-directed mutagenesis kit (Agilent). The F-MLV clone FB29 envelope gene was synthesised and cloned into pRV-IRES-GFP vector. All gene synthesis,

Live cell imaging
Live cell imaging and quantitative analysis of cell growth and morphology were performed label-free on the Livecyte system (Phasefocus). Ten thousand melanoma cells in 0.5-2 ml medium were seeded per well in a 24-well plate 18 hours prior to imaging. Four regions of interest were selected for each well using 10× magnification and imaged with 5 minutes intervals. Images were acquired using a 10× Plan N objective and Livecyte Acquire software. Single-cell tracking, segmentation and analysis were performed using Livecyte Analyse software.

Protein modelling
The model of the putative HECT domain of murine HECTD2 (NP_001156943. were determined by calliper measurements of two axes and calculated using the formula: π × a × b, where a = half of length and b = half of width of the tumours. These measurements were confirmed by additional calliper measurements and weighing of the resected tumour mass at the end of the observation period. Investigators were not blinded to the different groups.

Gene and protein functional annotation
Pathway analyses were performed using g:Profiler web server [4] (https://biit.cs.ut.ee/gprofiler) with genes ordered by the degree of differential expression. P values were estimated by hypergeometric distribution tests and adjusted by multiple testing correction using the g:SCS (set counts and sizes) algorithm, integral to the g:Profiler server.

Analysis of gene expression and essentiality in tumour cells lines was performed on the Cancer
Dependency Map [5] (DepMap) portal website (https://depmap.org/portal).

Proteomics analyses
Five replicate cultures of at least 10 6 cells each of HCmel31 or HCmel31.Hectd2 c1 cells were processed for label-free proteomic analyses. Cell pellets were resuspended with pipetting, first in 50 µl ice-cold PBS, then a 2× lysis buffer was added (also 50 µl).

RNA-seq data analysis
RNA-seq reads were obtained from TCGA, CCLE, and the two we cohorts of melanoma patients that were treated with PD-1 blocking antibodies [6,7] (Accession numbers GSE91061 and GSE78220).
Reads were aligned to the GENCODE assembly (basic, version 30) or our custom transcript assembly, as previously described [8]. Transcripts per million (TPM) calculations were carried out for all transcripts with a custom Bash pipeline using GNU parallel [9] and Salmon [10]  library was used to plot the distances. Sample clusters were defined using the cutree() function. All downstream differential expression analyses and visualisation were carried out using Qlucore Omics Explorer 3.3 (Qlucore, Lund, Sweden).

DNA-seq data analysis
We used alleleCounter Median VAFG for the remaining 341 samples was 0.96, and counts for A, C and T did not exceed 4. Of note, coverage at this location was globally only half of the mean read coverage reported for normal samples in PCAWG (39X), likely due to a very high GC content around this SNP (77% in a +/-125bp window), with a flagged (RepeatMasker) region due to high G/C repeats adjacent to this locus.