RNF43 and ZNRF3 are commonly altered in serrated pathway colorectal tumorigenesis

Serrated pathway colorectal cancers (CRCs) are characterised by a BRAF mutation and half display microsatellite instability (MSI). The Wnt pathway is commonly upregulated in conventional CRC through APC mutation. By contrast, serrated cancers do not mutate APC. We investigated mutation of the ubiquitin ligases RNF43 and ZNRF3 as alternate mechanism of altering the Wnt signal in serrated colorectal neoplasia. RNF43 was mutated in 47/54(87%) BRAF mutant/MSI and 8/33(24%) BRAF mutant/microsatellite stable cancers compared to only 3/79(4%) BRAF wildtype cancers (p<0.0001). ZNRF3 was mutated in 16/54(30%) BRAF mutant/MSI and 5/33(15%) BRAF mutant/microsatellite stable compared to 0/27 BRAF wild type cancers (p=0.004). An RNF43 frameshift mutation (X659fs) occurred in 80% BRAF mutant/MSI cancers. This high rate was verified in a second series of 25/35(71%) BRAF mutant/MSI cancers. RNF43 and ZNRF3 had lower transcript expression in BRAF mutant compared to BRAF wildtype cancers and less cytoplasmic protein expression in BRAF mutant/MSI compared to other subtypes. Treatment with a porcupine inhibitor reduced RNF43/ZNRF3 mutant colony growth by 50% and synergised with a MEK inhibitor to dramatically reduce growth. This study suggests inactivation of RNF43 and ZNRF3 is important in serrated tumorigenesis and has identified a potential therapeutic strategy for this cancer subtype.


INTRODUCTION
The majority of colorectal cancers up-regulate the Wnt signalling pathway [1]. This is commonly due to mutation of the APC gene which is uncommonly mutated in BRAF mutant serrated pathway cancers [2]. Serrated neoplasia accounts for 25-30% of all colorectal cancers and these arise from sessile serrated adenomas (SSAs) or traditional serrated adenomas (TSAs) [3,4]. Approximately half of all BRAF mutant serrated pathway cancers will methylate the mismatch repair gene MLH1 and develop microsatellite instability (MSI) [5,6], whilst the remainder are microsatellite stable (MSS). A recent study reported germline mutation of the upstream Wnt inhibitor RNF43 in a minority of subjects with serrated polyposis [7,8], providing a potential alternative to APC Research Paper www.impactjournals.com/oncotarget mutation as a mechanism for altering the Wnt signal in serrated neoplasia.
RNF43 is a transmembrane E3 ubiquitin ligase that is up-regulated in response to accumulation of β-catenin following increased Wnt signalling [8,9]. Together with its functional homologue, ZNRF3, it localizes to the plasma membrane and targets Frizzled (FZD), a key receptor of the Wnt ligand, for ubiquitination and lysosomal degradation [8]. RNF43 and ZNRF3 are highly expressed in murine intestinal stem cells, and deletion of these renders cells hypersensitive to Wnt ligand secretion, resulting in intestinal adenomas with strong expression of β-catenin [8].
Release of the Wnt ligand from a signalling cell is dependent on the molecule Porcupine (PORCN). PORCN facilitates post-translational palmitoylation of the Wnt ligand which is necessary for its transport out of the cell and subsequent recognition by a FZD receptor on a receiving cell. Treatment with a Porcupine inhibitor reduces growth of RNF43 mutant pancreatic cancer cell lines, where proliferation is dependent on endogenous Wnt signalling [20]. The porcupine inhibitor LGK974 has been shown to dramatically reduce growth of pancreatic cancer cells that carry an RNF43 mutation [20].
We analysed RNF43 and ZNRF3 for mutation, copy number variation and expression in a large number of colorectal cancers stratified for molecular subtype. We hypothesized that in serrated pathway cancers where APC mutation is uncommon, inactivation of RNF43 and/ or ZNRF3 would present an alternate mechanism for activating the Wnt signal.

Clinicopathological data of cancer cohorts
Cancers displayed typical clinicopathological features when stratified for molecular subtype as previously described [21][22][23]. Patients with BRAF mutant/MSI cancers presented at significantly older ages compared to both MSS subtypes. BRAF mutant cancers were predominantly present in the proximal colons of females and displayed the methylator phenotype more commonly than BRAF wild type cancers (Table 1). The prevalence of other mutations is shown in Table  2 and Figure 1A. MSI cancers tended to show mutations in repeat tracts. In addition to the G659 (G7 deletion), recurrent frameshift mutations were present at R117 (C6 repeat tract) in exon 3 and affected 6/54 (11.1%) of BRAF mutant/MSI cancers. Other recurrent mutations in BRAF mutant/MSI cancers were X441fs (n=3), X370fs (n=2) and G363D (n=2). Eight BRAF mutant/MSS cancers (8/33, 24.2%) had RNF43 mutations including 2 point mutations and 6 frameshift mutations that were located throughout the gene. BRAF wild type cancers had a minimal mutation rate of 3/79 (3.8%) and all were missense mutations in exon 9 (Table 2, Figure 1A).
Ten colorectal cancer cell lines were analysed for presence of an RNF43 and ZNRF3 mutation. RNF43 mutations were found in five cell lines ( Table S2B).
At the ZNRF3 locus, 22q12.1, there was a high frequency of deletion events in MSS cancers (BRAF mutant/MSS: 16/33, 48.5%; BRAF wild type: 10/18, 55.6%). In addition to frequent loss of heterozygosity of ZNRF3 occurring in the BRAF mutant/MSS cohort, one of these cancers harboured a homozygous deletion spanning approximately 170kb and including the C terminal half of the ZNRF3 locus. Again BRAF mutant/MSI cancers rarely showed deletion events at this locus (1/30, 3.3%).  Figure 3A and 3B). Cytoplasmic RNF43 staining was significantly less common in cancers harbouring the X659fs mutation compared to that in wild type cancers (RNF43 cytoplasmic positive cancers: 11/27, 41% vs 35/49, 71% respectively, p=0.01). No difference in nuclear RNF43 expression was found between the X659fs mutant and wild type cancers (Table 4B).

RNF43 cytoplasmic expression is reduced in BRAF mutant/MSI cancers
β-catenin immunohistochemical staining was assessed, with nuclear accumulation indicative of elevated Wnt signal. Overall, β-catenin was more likely to be

Response to Porcupine inhibitor, LGK974, for RNF43 and ZNRF3 mutant cell lines
Cell lines were seeded at varying densities according to growth rate to promote long term colony formation. Treatment with the porcupine inhibitor, LGK974, at various concentrations (at 5uM, 10uM and 20uM) was replenished every 48 hours for approximately 2 weeks and colonies were subsequently fixed, stained and counted by two independent examiners. In all RNF43 and ZNRF3 mutant cell lines, colony formation was inhibited in a dosedependent manner of up to 53% with LGK974 treatment compared to growth seen in the DMSO treated control cells ( Figure 4A; Supplementary Figure S1A). HCT116 had the least growth inhibition with LGK974 treatment, whilst RKO and LS174T demonstrated the most. There was no overall decrease of growth with LGK974 treatment observed in any of the RNF43 / ZNRF3 wild type cell lines ( Figure 4A).

Combination of Porcupine and MEK inhibitors accentuate growth inhibition in RKO cells
The RKO cell line most closely resembles primary cancers of the serrated pathway as it is BRAF mutant, MSI and CIMP high [24,25]. RKO harbours a homozygous RNF43 X659 frameshift mutation and two heterozygous ZNRF3 mutations (X249fs, G481W). Treatment with 10μM LGK974 decreased colony growth by approximately 30% (Figure 4A and 4B), and treatment with the MEK inhibitor, PD0325901, reduced colony growth by 64% at 0.05uM and 84% at 0.1uM ( Figure 4B). The combination of LGK974 and PD0325901 synergised to further reduce RKO's cell colony growth ( Figure 4B), and this was in line with markedly reduced phosphorylation of ERK1/2 and downstream c-MYC (Supplementary Figure S2). These results demonstrate the potential effectiveness of this novel combination therapy in treating RNF43 / BRAF mutant serrated pathway cancers. (Images of cells treated with 0.002uM -0.5uM with and without LGK974 application are shown in Supplementary Figure S1B).

DISCUSSION
The E3 ubiquitin ligases RNF43 and ZNRF3 normally cooperate to downregulate the Wnt signal by targeting the Frizzled receptor for degradation. This study provides a comprehensive analysis of RNF43 and ZNRF3 in cohorts of colorectal cancers stratified by BRAF mutation and MSI status. The high rate of mutation and structural variation in BRAF mutant cancers suggests alteration of these genes is important for progression of serrated pathway cancers, where Wnt activation by APC mutation is uncommon [2].  [26] demonstrated R-spondin independent growth in murine colonic organoids. Here we report 71-80% of MSI cancers from two independent series had an x659 frameshift mutation in the distal region of the gene. This is predicted to result in a premature stop codon and was also reported by Giannakis et al to occur in 64% of MSI cancers [18]. The high mutation rate suggests this mutations confers a positive selective advantage [13]. This was confirmed by Yan et al where biallelic RNF43 X659fs mutation, along with targeting of ZNRF3, in a murine colon organoid model conferred R-spondin independence [27].
Serrated neoplasia may originate in sessile serrated adenomas (SSA) or traditional serrated adenomas (SSA), the former being the precursor of MSI cancers and the latter preceding the development of BRAF mutant / MSS cancers [28,29]. Sekine et al [30] reported RNF43 mutations in 6% SSA and 24% TSA, compared to our findings of 87% in BRAF mutant MSI and 24% BRAF mutant MSS cancers. This suggests that RNF43 may assist progression of SSA to MSI cancers, whereas RNF43 mutation may be involved at an earlier stage of TSA development. ZNRF3 mutations were not observed in SSAs and in only 1% of TSAs [30] compared to 30% BRAF mutant/MSI cancers and 15% BRAF mutant /MSS cancers, suggesting that this less common mutation is involved in progression to cancer rather than precursor development. BRAF mutant/MSS cancers also commonly showed deletion of the RNF43 locus at 17q22, which may contribute to gene silencing and compensate for the lower RNF43 / ZNRF3 mutation rate in this cancer subgroup.
Cytoplasmic RNF43 expression was least frequently observed in the BRAF mutant/MSI cancers. Because RNF43 abrogates the Wnt signal at the level of the Frizzled receptor, it may be expected to be localised predominantly in the cytoplasm or membrane rather than the nucleus. Whether its expression is transient and is degraded soon after functioning, or whether it is maintained within the cell for a period of time is unknown. RNF43 has been reportedly expressed in both the cytoplasm and nucleus of gastric cancer cells [14], the cytoplasm of gastric cancer and glioma cells [31,32] as well as in the nuclear membrane and endoplasmic reticulum of cervical cancer cells [33]. A recent study described nuclear localization of RNF43 where it functioned to inhibit the Wnt pathway through its sequestering of TCF4, a transcription factor activated by beta-catenin target gene, to the nuclear membrane which prevented TCF4's mediated transcription of genes involved in downstream Wnt signalling [34]. RNF43 may have various time and cell type specific roles each requiring a unique subcellular localization. In the current study, nuclear localisation of β-catenin was used as a surrogate for elevated Wnt signalling. This was observed in the majority of BRAF wildtype cancers where elevated cytoplasmic RNF43 staining was also observed, likely as a futile attempt to dampen the Wnt signal by RNF43-mediated degradation of the Frizzled receptor. BRAF wildtype cancers also showed a significant increase in level of transcript expression. MSI cancers were more likely to maintain a normal staining pattern for β-catenin and this was correlated with a lower incidence of abnormal cytoplasmic staining for RNF43, most likely due to the high RNF43 mutation rate in this subgroup.  The RNF43 and ZNRF3 mutant colorectal cancer cell lines were all MSI, and all harbour a mutation within the EGFR-MAPK signalling pathway. Treatment of these cells lines with the Porcupine inhibitor, LGK974, resulted in consistently and significantly reduced cellular proliferation compared to wild type cell lines. The RNF43 wild type cancer cell lines all harbour either APC or β-catenin mutations which would be predicted to render them insensitive to upstream inhibition of the Wnt signal, as was observed. The approximate 50% growth reduction for the RNF43 and ZNRF3 mutant cell lines was not as dramatic as was observed for an RNF43 mutant organoid model, where the colonoid was sensitive to the similar porcupine inhibitor, IWP2 [35]. This difference may reflect a different magnitude of reliance on endogenous Wnt signalling compared to the colorectal cell lines. Organoid models are more representative of in vivo cancer growth due to their three dimensional structure and shorter time in culture. The epigenetic instability of BRAF mutant cancers in particular can induce DNA methylation and silencing of other Wnt pathway molecules that may alter the sensitivity of the cells to an upstream porcupine inhibitor. The novel combination of Porcupine and MEK inhibition synergistically inhibited RKO cell growth by more than 90%, which suggests this may be a promising therapeutic strategy to treat BRAF / RNF43 / ZNRF3 mutant serrated pathway cancers.
Overall this study has identified that the E3 ubiquitin ligases, RNF43 and ZNRF3, are frequently mutated in BRAF mutant cancers of the serrated pathway, particularly those that are MSI. Heterozygous loss of the RNF43 and ZNRF3 loci were identified in BRAF mutant/MSS cancers. BRAF mutant cancers did not upregulate RNF43 or ZNRF3 expression at the transcript or protein level, unlike the BRAF wild type cancers that attempt to mitigate the Wnt signal. Functionally, it was shown that RNF43 as well as ZNRF3 mutant colorectal cancers may be Wnt ligand dependent and that inhibition of the Wnt signal by targeting Porcupine may facilitate decreased growth of a proportion of MSI serrated pathway cancers. Furthermore, the novel combination of Porcupine and MEK inhibition resulted in considerable growth reduction, which indicates this may be a promising treatment approach for patients with serrated pathway cancers harbouring RNF43 and/or ZNRF3 mutations.

Patient demography and sample selection
Cancer and matched normal samples were obtained either as fresh frozen tissue from patients undergoing surgery at the Royal Brisbane and Women's Hospital, or as formalin-fixed paraffin embedded (FFPE) tissue from Envoi Specialist Pathologists, Brisbane, Australia. This study was approved by the Royal Brisbane and Women's Hospital and Bancroft Human Research Ethics Committees.
Clinicopathological data including gender, age at diagnosis and anatomical site of cancer (with proximal defined as proximal to the splenic flexure) were collected. DNA from fresh cancer and matched normal tissue was extracted using AllPrep DNA mini kit (Qiagen, Dusseldorf, Germany). DNA from the FFPE cancers was extracted by the Chelex-100 method (Bio-Rad Laboratories, CA, USA). The presence of MSI had been previously analysed for the RBWH cancer samples using the National Cancer Institute's 5 marker panel [21,36,37]. Cancers from Envoi Specialist Pathologists were evaluated for immunohistochemical loss of MLH1 mismatch repair protein expression as a surrogate for MSI. Presence of the BRAF V600E (a1796t) mutation, p53 mutation (over exons [4][5][6][7][8] and KRAS mutation (over codons 2 and 3) had been previously investigated for the fresh RBWH samples [21] whilst BRAF V600E (a1796t) and KRAS (codons 2 and 3) mutations were analysed for the formalin fixed Envoi samples as previously described [21,[38][39][40].

RNF43 and ZNRF3 mutation analysis
Sanger sequencing was analysed across the coding region of RNF43 (performed by Macrogen Incorporated, South Korea) in 87 BRAF mutant (54 MSI and 33 MSS) and 79 BRAF wild type cancers from the RBWH series. ZNRF3 was sequenced in all 87 BRAF mutant and an unselected subset of 27 BRAF wild type cancers. The second series of FFPE cancers from Envoi Specialist Pathologists consisting of 48 BRAF mutant (36 MSI and 12 MSS) and 29 BRAF wild type cancers were sequenced to further validate the frequency of the X659fs mutation (sequencing primers: F 5'GTCCAGGCCTCCTATTCCTC; R 5' CTGGTAGCAGCCTCTTGTCC). Only mutations predicted to elicit a functional response (frameshift, missense, nonsense and splice mutations) were included in analyses, and mutations found intronically, in untranslated regions or silent mutations were excluded from further investigation.

Transcript expression
19 BRAF mutant (13 MSI, 6 MSS), 27 BRAF wild type and 12 matched normal mucosa samples were www.impactjournals.com/oncotarget randomly chosen from the freshly collected RBWH series and analysed for transcript expression on the HumanHT-12 v4 Expression BeadChip arrays (Illumina, San Diego, CA). Total mRNA (500 ng) was reversetranscribed, amplified and biotinylated using the Illumina TotalPrep-96 RNA Amp Kit (Ambion, Austin, TX). The labelled cRNA (750 ng) was hybridized to the BeadChip arrays followed by washing, blocking, and staining with streptavidin-Cy3 according to the manufacturer's specifications. Fluorescence intensity for each probe on the array chips were measured on the iScan system and the data was extracted using GenomeStudio software (Illumina). All arrays passed quality control criteria in the GenomeStudio package as described by the manufacturer. A sample probe profile file generated in GenomeStudio without background subtraction or normalisation was used for subsequent analysis in R. The data was background corrected and quantile normalised using the limma package function. Probes with detection p-values P>0.05 in >95% of all samples were excluded from further analysis. 33

Immunohistochemical analysis
Tissue sections obtained from FFPE blocks from the Envoi series underwent antigen retrieval at low pH (pH6, Reveal decloaker; Biocare Medical, CA, USA) for 15mins at 105°C. H 2 O 2 and Sniper were used to facilitate endogenous peroxidase and protein blocks respectively. Primary antibodies were manually applied and incubated: RNF43 antibody (anti-RNF43 HPA-008079; Sigma, St Louis, MO, USA) at 1/500 dilution for 1 hour and betacatenin (anti-beta-catenin 224M16 [14] Cell Marque, California, USA) at 1/600 for 1 hour. MACH3 rabbit or mouse secondary antibody probe and polymer were applied for 15 and 30 minutes respectively (Biocare Medical, CA, USA), and DAB chromagen (Biocare Medical, CA, USA) was applied for 5-8 minutes.
Observation of samples was performed by an expert gastrointestinal pathologist (MB) and scored on a scale of 0-3 with 0 representing absent and 3 representing maximal staining. A score of 0-1 was considered negative and a score of 2-3 as positive. RNF43 staining within the cancer region of a sample section was compared to that observed within normal lymphocytes, which served as an internal control.

Cell colony assays
All cell lines were authenticated using short tandem repeat (STR) profiling in accordance with ATCC standards. Cell lines were last tested in June 2015.
Cell lines were seeded at varying densities in 6 well plates (HCT116, SW480 at 500 cells/well; DLD1, RKO, Lim1215 at 750 cells/well; SW48, Lisp1, LS174T, LoVo, HT29 at 1000 cells/well) and maintained in 2mls RPMI media as described above. 24 hours later, cells were treated in triplicate with either DMSO in control wells or the Porcuine inhibitor LGK974 (at 5uM, 10uM or 20uM in DMSO) (Cayman Chemicals, Michigan, USA) every 48 hours. At approximately 2 weeks, cells were fixed with ice-cold 100% methanol and stained with 0.5% crystal violet in 80% methanol. For LGK974 treatments, cell colonies were counted by 2 examiners and the counts were averaged. For combination treatments, RKO cells were treated at 24 hours post plating and then every 48 hours with either DMSO, 10uM LGK974, and/or the MEK inhibitor PD0325901 (Selleck Chemicals, Texus, USA) (at 0.002uM, 0.05uM, 0.1uM, 0.3uM or 0.5uM). Drug combination cell colony assays were quantitated using Aperio ImageScope v12.1.0.5019 with application of the Positive Pixel Count v9.1 algorithm.

Statistical analysis
Categorical data was analysed for significance using Fisher's exact test or Pearson's chi-squared test where appropriate. Continuous data was analysed with either a student's t-test or ANOVA as appropriate. P values <0.05 were considered significant.

CONFLICTS OF INTEREST
The authors have no conflicts of interest to declare for this study or in the production of this manuscript.