IntroductionNeuroblastoma is a developmental cancer of early childhood, arising from sympathoadrenal precursors of the neural crest1. Tumour cells are highly heterogenous and this heterogeneity contributes to the high relapse rate and poor survival of high-risk patients, representing a stubborn clinical challenge. Although the common oncogenic drivers found in adult human cancers are infrequent upon presentation in neuroblastoma1,2, several other drivers are known including MYCN gene amplification and activations of the tyrosine kinase ALK, or tyrosine phosphatase PTPN111. Nevertheless, 75% of tumours have no clear oncogenic drivers1,3.In our search for potentially new drivers with oncoprotein-like behaviour, we have examined the roles of HMMR, also known as RHAMM and CD1684, in neuroblastoma cells. HMMR may be of interest in a neural tumour because it has been implicated in neurite extension processes in neuronal cell lines, including in a hybrid neuroblastoma/glioma line NG108-155, and HMMR loss-of-function generates neurodevelopmental defects in vertebrate embryos6. HMMR also has oncogenic roles in several other human cancer systems, sustaining cell proliferation, survival and migration in cells derived from cancers of brain, lung, ovary, prostate, head and neck and breast7,8,9,10. HMMR is a cell surface hyaluronic acid (HA) receptor11. HA and its catabolized products can promote cell proliferation and survival, motility and metastasis in tumour cells12, interacting with cells through an HMMR/CD44 complex and signaling through ERK, AKT, SRC, Rho GTPases and FAK4,8,11,12,13,14. Interestingly, HHMR also acts in the nucleus, binding to microtubules and centrosomes and regulating mitotic spindles and chromosomal stability through interactions with DYNLL1 complexes, CHICA and BRCA14,15. HMMR also localises TPX2 to centrosomes, maintaining spindle pole assembly16,17. It operates through the release of TPX2 after HMMR degradation by BRCA1, leading to Aurora kinase A (AURKA) activation and BRCA1 phosphorylation16.HMMR is strongly expressed in neuroblastomas and we hypothesised that it may have pro-oncogenic potential corresponding to that seen in other cancer models. Using in silico analysis we examined the relationship between high HMMR expression and prognostic outcomes in neuroblastoma tumours. At a cellular level we targeted HMMR for inactivation using CRISPR/Cas9 in the KELLY neuroblastoma cell line. Our analyses show that HMMR is indeed a promoter of several parameters of tumour cell behaviour and that these are variably affected by HA ligands. Lastly, we used phosphoproteomics to begin to explore the potential biochemical roles of HMMR in neuroblastoma cells, confirming that it modulates ERK signaling and potentially also MTOR and DNA damage response (DDR) pathways.ResultsHMMR as an independent prognostic marker in neuroblastomaAlthough high HMMR expression has been associated with cancer progression18, this is yet to be investigated for neuroblastomas. HMMR expression was elevated in neuroblastomas compared to normal tissues, benign ganglioblastomas and neural crest-derived tumour pheochromocytoma (Fig. 1A and Supplementary Fig. S1). Moreover, HMMR is ranked in the top 1–5% overexpressed genes among those in the HA signaling axis, HA binding molecules and those associated with cell motility (Supplementary Fig. S1). This supports a possible role of the HMMR gene in the establishment or progression of neuroblastoma.Fig. 1HMMR expression and neuroblastoma clinicopathological features. (A) The expression of HMMR in normal and neuroblastoma tissues examined by R2 genomics platform. Normal tissues (blue) are divided in 9 groups and pheocytochromocytomas/paragangliomas (green) in 2 groups. Neuroblastic tumours (red) are divided in 4 groups, from benign ganglioneuromas, ganglioblastomas to aggressive metastatic neuroblastomas. The authors and sample numbers are also shown. Grade staging (B) and overall survival (C) are depicted for the Kocak and SEQC datasets. (D) t-SNEA maps analysis on the Kocak dataset performed in R2, for MYCN- amplified or non-amplified group (left panel), MYCN expression (middle panel) and HMMR expression (right panel).Full size imageTo further explore the role of HMMR in neuroblastomas, we examined the association between HMMR expression and tumour staging. Higher HMMR expression associated strongly with increased INSS tumour grade (Fig. 1B). We also found that elevated HMMR expression correlated significantly with poor overall survival (OS) in patient datasets analysed in the R2 platform (Fig. 1C). In t-SNEA maps, elevated HMMR expression partially overlaps with MYCN-amplified patient groups (AMP), but shows a somewhat differential expression pattern between patient data points, and it is also expanded to the non-AMP group (Fig. 1D). A Cox univariate and multivariable logistic regression analysis was performed on the SEQC dataset, demonstrating that HMMR expression, but not other HA-related pathway genes, is an independent risk factor for neuroblastoma patients (Table 1). To clarify the potential biological functions of HMMR we examined the genes that show positive correlations with HMMR expression in 4 tumour datasets (Supplementary Fig. S2). With this 2581 gene set, the HMMR co-expression signature again correlated with poor OS in neuroblastoma (Supplementary Fig. S2).Table 1 Univariate and multivariable Cox regression of the prognostic covariates in patients with NB (n = 490, SEQC patient dataset from R2 Genomics). See Methods for full details.Full size tableHMMR promotes neuroblastoma cell proliferationEquipped with these prognostics data, we next directly examined the cellular function of HMMR in neuroblastoma cells. Using CRISPR/Cas9 in KELLY cells (strong HMMR expressors) we created out-of-frame mutations in a region encoding the HMMR N-terminus (Fig. 2A). Three HMMR knock-out (KO) subclones were identified: KA5 (1 bp homozygous insertion), KA14 and KA16 (1 bp homozygous deletions; these are not proven to be independent subclones) (Supplementary Fig. S3). A further subclone, KC17, arose from the CRISPR/Cas9 selection process, but HMMR was not targeted with indels; we have not tested for other off-target alterations. The initial rationale for including this KC17 in the study was thus as a form of control for the CRISPR/Cas9 treatment process. It was confirmed that HMMR protein was absent from KO subclones KA5, KA14 and KA16, but was retained in KC17 and parental KELLY (Fig. 2B). The subcloned lines were all morphologically similar to parental KELLY (Supplementary Fig. S3). HMMR depletion in KA5, KA14 and KA16 inhibited their proliferative expansion compared to KELLY and KC17, agreeing with a similar role in other cancer cell types (Fig. 2C)7,8. Moreover, low density growth assays showed a strong reduction in colony forming ability in cells lacking HMMR (Fig. 2D), indicating a loss of clonogenic capacity.Fig. 2HMMR depletion suppresses proliferation and clonogenicity. (A) Schematic of the HMMR protein, with interaction domains shown for microtubules, HA, CHICA and Calmodulin, plus the carboxy-terminal bZIP region52. The guide RNA target site used in CRISPR/Cas9 is shown. (B) Immunoblot analysis of HMMR in parental KELLY cells and the clones generated by CRISPR/Cas9. (C) Cell proliferation assay using resazurin, normalized to growth of KELLY cells (n = 4–7). (D) clonogenic assay quantified from 6-well plate assays (example plate image shown, crystal violet staining) (n = 4). In C and D data are expressed as a mean ± SD; *p