Sulfatase-2 from Cancer Associated Fibroblasts: An Environmental Target for Hepatocellular Carcinoma?

Introduction Heparin sulphate proteoglycans in the liver tumour microenvironment (TME) are key regulators of cell signalling, modulated by sulfatase-2 (SULF2). SULF2 overexpression occurs in hepatocellular carcinoma (HCC). Our aims were to define the nature and impact of SULF2 in the HCC TME. Methods In liver biopsies from 60 patients with HCC, expression and localization of SULF2 were analysed associated with clinical parameters and outcome. Functional and mechanistic impacts were assessed with immunohistochemistry (IHC), in silico using The Cancer Genome Atlas (TGCA), in primary isolated cancer activated fibroblasts, in monocultures, in 3D spheroids, and in an independent cohort of 20 patients referred for sorafenib. IHC targets included αSMA, glypican-3, β-catenin, RelA-P-ser536, CD4, CD8, CD66b, CD45, CD68, and CD163. SULF2 impact of peripheral blood mononuclear cells was assessed by migration assays, with characterization of immune cell phenotype using fluorescent activated cell sorting. Results We report that while SULF2 was expressed in tumour cells in 15% (9/60) of cases, associated with advanced tumour stage and type 2 diabetes, SULF2 was more commonly expressed in cancer-associated fibroblasts (CAFs) (52%) and independently associated with shorter survival (7.2 vs. 29.2 months, p = 0.003). Stromal SULF2 modulated glypican-3/β-catenin signalling in vitro, although in vivo associations suggested additional mechanisms underlying the CAF-SULF2 impact on prognosis. Stromal SULF2 was released by CAFS isolated from human HCC. It was induced by TGFβ1, promoted HCC proliferation and sorafenib resistance, with CAF-SULF2 linked to TGFβ1 and immune exhaustion in TGCA HCC patients. Autocrine activation of PDGFRβ/STAT3 signalling was evident in stromal cells, with the release of the potent monocyte/macrophage chemoattractant CCL2 in vitro. In human PBMCs, SULF2 preferentially induced the migration of macrophage precursors (monocytes), inducing a phenotypic change consistent with immune exhaustion. In human HCC tissues, CAF-SULF2 was associated with increased macrophage recruitment, with tumouroid studies showing stromal-derived SULF2-induced paracrine activation of the IKKβ/NF-κB pathway, tumour cell proliferation, invasion, and sorafenib resistance. Conclusion SULF2 derived from CAFs modulates glypican-3/β-catenin signalling but also the HCC immune TME, associated with tumour progression and therapy resistance via activation of the TAK1/IKKβ/NF-κB pathway. It is an attractive target for combination therapies for patients with HCC.


Introduction
Liver cancer is the fifth most common cancer and the second cause of cancer-related mortality worldwide [1]. Hepatocellular carcinoma (HCC) accounts for 80-90% of cases, typically developing in the presence of chronic liver disease (CLD) caused by hepatotropic viruses, alcohol excess, or obesity [2]. DNA damage and molecular aberrations in hepatocytes contribute to cancer development, but tumour heterogeneity and failures to identify genetic drivers common to large numbers of patients have hampered targeted therapeutic advances. The tumour microenvironment (TME), comprising endothelial, mesenchymal, and immune cells, secreted extracellular matrix proteins and growth factors, can also influence tumour biology and progression [3]. Therefore, common TME modulators may be worthy therapeutic candidates. Indeed, targeting the immune cell checkpoints has recently heralded a major advance in the treatment of HCC, although as yet, not all patients respond [4]. Additional TME candidates, to target alone or in combination, are eagerly awaited.
Mature active sulfatase-2 (SULF2), which can be membrane bound or secreted, is an endo-sulfatase that removes sulphate groups at the 6-O position on sugars in heparansulfate (HS) chains of HS proteoglycans (HSPGs) [5]. In a number of cancers, SULF2 upregulation is associated with more advanced disease and poorer patient outcome [6][7][8]. Conversely, supressed SULF2 expression has been reported to predict higher sensitivity to topoisomerase-I inhibition in patients with lung cancer [9]. Previous microarray gene expression studies have identified elevated SULF2 mRNA levels in 79 (57%) of 139 HCCs, as well as 8 (73%) of 11 HCC cell lines [10]. Patient studies were in surgical candidates, where elevated SULF2 was associated with modulation of glypican-3 and fibroblast growth factor signalling, a higher rate of recurrence and shorter survival [10]. Expression in patients with more advanced disease is not well studied. The cellular source of SULF2 in patients with HCC, its targets, its impact on the HCC TME, and patient outcome remain to be comprehensively characterized.

Patients
The retrospective case series included 60 patients with HCC presenting to the Newcastle upon Tyne Hospitals NHS Foundation Trust between 2000 and 2010, with surplus formalin-fixed, paraffin-embedded (FFPE) tissues available for research, as well as n = 20 later patients treated with sorafenib. Patients with histologically benign disease or cholangiocarcinoma were excluded. Patient demographics and clinicopathological information included age, gender, underlying liver disease aetiology, liver function, Edmondson-Steiner tumour grade [11] modified for biopsy reporting by experienced pathologists, tumour node metastases and combined Barcelona Clinic for Liver Cancer (BCLC) stages [12], treatments administered, and patient survival from the time of biopsy. In an additional cohort of 20 patients with advanced disease referred for sorafenib, responders were those who had stable disease on imaging for at least 3 months, while non-responders were those whose treatment was discontinued due to poor tolerance or disease progression on imaging at 3 months. Finally, primary culture cancer activated fibroblasts (CAFs) were derived from human HCC tissues in 2021 (online suppl. Methods; for all online suppl. material, see www.karger.com/doi/10.1159/000525375).
Biopsies from the cohort of sorafenib-treated patients were additionally stained for CD45 (pan immune marker), CD4, CD8 (T cell markers), CD66b (neutrophil marker), and CD68 and CD163 (macrophage markers). The IHC slides were analysed digitally using Aperio Imagescope, with a positive pixel algorithm expressed as a ratio of total negative + positive pixels.
Cell Culture LX-2 liver myofibroblast cells were a gift of Scott Freidman (Mount Sinai, USA) and were cultured in DMEM high glucose media supplemented with 1% penicillin and streptomycin, L-glutamine, and 2% FBS. These cells were validated with 100% match to a profile for LX-2 CVCL_5792, comparing 8 sort tandem repeat loci in the Cellosaurus Database, by NorthGene Limited, Newcastle, UK. Huh7 and Hep3B HCC cells were purchased from the American Type Culture Collection (ATCC), using stocks frozen within 6 passages from purchase. COS-7 fibroblasts were a gift from Ralf Weiskirchen (Aachen, Germany). These were confirmed to be of African Green Monkey origin by mitochondrial DNA PCR using species-specific primers, with sort tandem repeat analysis showing an 87.5% match (7/8 loci) with the previously reported CV1/COS-7 profile [13]. COS-7 cells were cultured in DMEM high glucose media supplemented with 1% penicillin and streptomycin, L-glutamine and 10% FCS. All cells were cultured at 37°C in an atmosphere of 5% CO 2 . LX-2 cells were stimulated with 10 ng TGFβ for 48 h unless stated otherwise. Huh7 HCC cells and COS-7 cells were transduced with mission  TRC2 shRNA lentiviral particles targeting SULF2 (TRCN0000364518; Sigma-Aldrich, USA) or TRC2-pLKO-puro nontargeting (NT) with hexadimethrine bromide (shRNA sequences in online suppl. Table 1). Cells were selected using puromycin.

3D Tumour Spheroid Hanging Droplets
Mixed cell-type spheroids; 1,500 cells from Huh7 or Hep3B HCC cells were combined 1:1 with control or SULF2 knockdown (KD) COS-7 cells in 20 µL media per sphere on the lid of a 10 cm 3 dish (n = 10 per group). The lid was inverted and cells were gravity suspended in hanging droplets for 3 days with 10 mL sterile PBS in the dish beneath to maintain a humidified environment. The change in spheroid volume from days 3-8 was quantified. Tumour cell spheroids; 3,000 Huh7 or Hep3B cells were suspended in hanging droplets to form spheres. Spheroids were treated with stromal conditioned media (CM) for 3 days. Spheroid volume was calculated by measuring the area and applying the formula V mm3 = 0.09403 × ((A pixel × 0.28)/1,000) 1.5 .
Details of in vitro cell culture, 2D and 3D models, SULF2 KD, CCL2 and SULF2 ELISA, primary CAF isolation, migration assays, fluorescent activated cells sorting, RNA extraction, Western blotting, quantitative RT-PCR, primer sequences and antibodies are provided in the supplementary information and online supplementary Tables 1-3.

Statistics
Statistical analyses were conducted using SPSS version 23 or GraphPad Prism version 7.00. For comparisons between groups, continuous data was assessed with Mann-Whitney or Kruskall-Wallis tests. Categorical data was analysed using χ 2 tests. The principal documented endpoint in the clinical cases studied was overall survival, recorded as months from diagnosis until January 31, 2019. Differences in cumulative survival were determined using the Kaplan-Meier method and a Log-Rank test. The Cox proportional hazards-regression model was used to identify parameters associated with survival.

Results
Elevated SULF2 in the HCC TME Was Associated with Poorer Survival SULF2 tissue expression patterns were determined immunohistochemically in 60 human HCC diagnostic biopsies. The cohort median age was 69 years with a male to female ratio of 5 to 1. Overall, 49% (29/60) had type 2 diabetes, with 73% of patients having either alcohol-related liver diseases, nonalcoholic fatty liver diseases, or no known CLD. Cirrhosis was absent in 49% (29/60), reflecting the need for biopsy rather than radiological diagnosis in the absence of established cirrhosis. This selected cohort was of patients deemed fit for therapy and the majority had preserved liver function and BCLC A-C disease, with an overall median survival of 20.3 months. Demographic and clinicopathological features are summarized in Table 1.
In non-neoplastic and neoplastic tissues, SULF2 was detected in endothelial cells that served as an internal positive IHC control. SULF2 was detected at low levels in hepatic arterioles, benign bile duct epithelium, occasional nonparenchymal cells within the portal tracts and sinusoids, as well as on the canalicular surfaces of hepatocytes (Fig. 1a). In tumour tissues, SULF2 was overexpressed in 58% (35/60) of HCC cases, corroborating previously reported gene expression data. Whole tumour tissue mRNA expression, however, does not capture the predominant cell type responsible for expression. In our biopsy cases, the tumour cells were the source of elevated SULF2 in only 15% (9/60), where SULF2 was detected in the tumour cell cytoplasm and/or membrane (Fig. 1b). The more common cell-type expressing elevated SULF2 were the CAFs-identified αSMA-positive myofibroblast-like cells, with SULF2 expressed in 52% (31/60), in either a focal intense or widespread form ( Fig. 1c; online suppl. Fig.  1A). This was corroborated by interrogation of the publicly available (TCGA) human liver cancer dataset on the cBioPortal website (http://www.cbioportal.org/), where SULF2 expression in HCC correlated strongly in a highly significant fashion with the myofibroblast markers vimentin, αSMA (ACTA2), COL1A1, COL1A2, and TIMP2 (online suppl. Fig. 1B). Notably, SULF2 expression was less common and intense in αSMA-positive myofibroblasts in the portal tracts or bridging fibrous septa of the non-neoplastic liver (online suppl. Fig. 1C), in keeping with the TME promoting SULF2 expression.
Clinical features of the nine cases in which SULF2 expression was present versus absent in tumour cells are shown in Table 1. SULF2-positive HCCs were typically larger and associated with extrahepatic disease, consequently with a more advanced tumour node metastasis stage. The alpha-fetoprotein (AFP) level in this small group was also significantly higher (p = 0.03). Median survival was reduced, although not significantly so. Elevated SULF2 expression in CAFs demonstrated no specific associations with tumour grade or stage, but survival was similarly reduced in the presence of CAFs-SULF2 upregulation (12.2 months vs. 35.0 months).
Survival analyses excluding surgically treated patients with relatively early-stage HCC (10 treated surgically) highlighted the difference in those with more advanced disease. In the 50 patients receiving locoregional, medical or supportive treatments (Table 1; Fig. 1d), SULF2 elevation in CAFs was associated with a median survival of 7.2 months, compared to 29.2 months in those without CAF-SULF2 (p = 0.003; Kaplan-Meier). Univariate analyses to identify factors associated with survival are shown in Table 2, as are those factors with a p value <0.05 when entered into a multivariate Cox regression analysis. Serum TNM, tumour node metastases; PST, performance status. Continuous data are presented as mean±standard error unless otherwise stated, with statistical comparisons using a Mann-Whitney test. Categorical data were compared using a χ 2 test. Survival was assessed by the Kaplan-Meier method. "Other" included small numbers with Hepatitis C (n = 4); haemochromatosis (n = 4), cryptogenic cirrhosis (n = 4); Hepatitis B (n = 1); autoimmune hepatitis (n = 2); and α-1-antitrypsin deficiency (n = 1). AFP, albumin level, and performance status, in addition to SULF2 in CAFs, were independently associated with poorer survival (Table 2).

Glypican-3 in the HCC TME
Glypican-3 is a morphogen HSPG and both its expression and 6-O sulfation, which promote canonical Wnt signalling, have been reported as SULF2 regulated in PST, performance status; HR, hazards ratio. Factors associated with survival in 50 patients for whom surgical treatment was not an option were assessed by univariate analysis. Factors with a p value <0.01 were entered into a multivariate Cox regression analysis. Two distinct multivariate analyses are shown, the first considering stromal SULF2 and the second (shaded in grey) showing a similar analysis, but in which SULF2 overexpression in either HCC cells or stromal cells was classed as "present." Significance and HR with upper and lower 95% confidence intervals are shown.   Table 4. Data mean ± SEM; n = 7 to 10 spheroids per condition. **p < 0.01; ***p = 0.001; ****p < 0.001. "P" denotes number of passages for CAFs.

Modelling the Impact of Stromal SULF2
To model the impact of stromal SULF2 on tumour cell growth, we used COS-7 cells (African green monkey kidney embryonic fibroblasts expressing high endogenous SULF2, with 98.5% homology to human SULF2; https:// www.ensembl.org BLAST [16]), alongside manipulating SULF2 expression in human liver LX-2 myofibroblast cells, which have low levels of endogenous SULF2 (Fig. 2a, b). A stable SULF2 KD was generated in the SULF2-rich COS-7 cells, also confirmed by Western blotting and ELISA (Fig. 2a). SULF2 was transiently overexpressed in LX-2 cells and empty vector-transfected cells as a control (Fig. 2b). Secretion of SULF2 in the media was confirmed by Western blotting (Fig. 2b). The HCC TME, in which we report frequent upregulation of CAF-SULF2, is characterized by activation of the TGFβ1 signalling pathway [17] known to induce SULF2 in renal disease [18]. Indeed, SULF2 protein levels were elevated in LX-2 cells challenged in vitro with exogenous TGFβ1 (Fig. 2b), with a strong and a highly significant correlation in gene expression between the two confirmed in the cancer genome atlas (TCGA) HCC dataset (Fig. 2b). Primary CAFs from 4 human HCC with IHC evidence of stromal SULF2 expression (Fig. 2ci + cii), expressed endogenous SULF2 and aSMA, confirmed by dual labelling immunofluorescence (Fig. 2ciii). SULF2 expression levels have been previously characterized in HCC cell lines [10], with Hep3B cells producing very low levels versus higher expression in Huh7 cells. We compared the impact of SULF2-CM on these low and high expressors in further experiments.

Stromal SULF2 Promotes the Growth, Migration, and Invasion of HCC Cells in vitro
A 3D-HCC spheroid model was adopted to assess the impact of stromal SULF2 on HCC growth. The growth rate of Hep3B spheroids cultured in CM collected from TGFβ1-stimulated LX-2 cells or cells over-expressing SULF2 was significantly increased compared to spheroids exposed to CM from control unstimulated or empty vector-transfected LX-2 cells, with low SULF2 (Fig. 2d, e). Similarly, Hep3B spheroid growth was significantly increased when cultured in CM from control SULF2-expressing COS-7 cells but not SULF2 KD COS-7 cells (Fig. 2f), confirming an HCC mitogenic role for stromal SULF2.
In a 2D transwell culture, COS-7 cells increased the metabolic activity and cellular proliferation of both Huh7 and Hep3B cells, while tumour cells alone or co-cultured with SULF2 KD COS-7 cells did not (online suppl. Fig.  3A, B). Fibroblast-derived SULF2 also markedly increased the migration and invasion of the SULF2 null Hep3B cell line compared to controls (online suppl. Fig.  3C, D).

SULF2 Blockade Reversed 3D Spheroid Growth
Hep3B cell proliferation, induced by CM from TGFβ1stimulated LX-2 cells or by CM from primary CAFS expressing endogenous SULF2, was suppressed by an anti-SULF2 antibody, rather than an IgG isotype control antibody (Fig. 3a, b). Similarly, COS-7 CM-induced Hep3B proliferation was significantly diminished by an anti-SULF2 antibody but not a control antibody (Fig. 3c). In Huh7 cells that produce SULF2, the SULF2 Ab also suppressed spheroid growth (online suppl. Fig. 4A).

CAF-SULF2 Contributes to Sorafenib Resistance
For the last decade, sorafenib has been -and remains -a standard first-line therapy for patients with advanced HCC. Sorafenib is associated with a median survival advantage of 2-3 months [19]. In 3D cultures, Hep3B growth was attenuated by sorafenib treatment when cultured in SULF2 low CM from LX-2 cells or SULF2 KD COS-7 cells (Fig. 3d). However, the effect of sorafenib was mitigated, with increased growth observed, when spheroids were grown in CM from TGFβ1-stimulated LX-2 cells or COS-7 cells (Fig. 3d; online suppl. Fig. 4B, C). Treatment with a SULF2 Ab restored sorafenib sensitivity and growth attenuation, supporting SULF2-mediated sorafenib resistance (Fig. 3d; online suppl. Fig. 4B, C).
We explored outcomes in an independent cohort of 20 patients with advanced HCC referred for sorafenib treat-   Table 4. Data mean ± SEM; 7 to 10 spheroids per condition. ****p < 0.001. Liver Cancer 2022;11:540-557 DOI: 10.1159/000525375 ment (online suppl. Table 5). Their median age was 72.5 years. The majority had ALD or NAFLD, with 50% having no evidence of cirrhosis. Fourteen were referred for first-line medical therapy, with six referred for 2nd-line treatment, having previously received arterial treatment. Eighteen proceeded with sorafenib treatment, starting at 400 mg twice daily. Seven had a treatment break or dose reduction prior to permanent discontinuation, with sorafenib discontinued either because of toxicity or disease progression. The duration of therapy was shorter in those with focal intense or widespread CAF-SULF2 expression (absent, scant, focal, focal intense, or widespread; Spearman −0.536; p = 0.022; n = 18 -Kaplan-Meier p = 0.003) (online suppl. Fig. 4D).

Glypican-3 and β-Catenin Expression Were Modulated by CAF-SULF2
Mechanistically, we first considered the impact of CAF-SULF2 on canonical Wnt/β-catenin signalling -a well-established target of SULF2. Huh7 cells overexpress SULF2 in the presence of a wild-type Wnt-β-catenin pathway [20]. shRNA SULF2 knockdown Huh7 as well as NT shRNA Huh7 cells were stimulated with the frizzled receptor ligand Wnt3a (100 ng/mL). Wnt3a stimulated the activity of a β-catenin TOPflash luciferase reporter assay in both cell lines, with significantly greater activity in NT shRNA, in keeping with partial dependence on SULF2 (Fig. 4a). Further, in vitro stimulation of Huh7 cells with SULF-CM from COS-7 cells increased GPC3 as well as β-catenin expression, which KD-SULF2-CM did not (Fig. 4b). SULF2-induced β-catenin expression was predominantly membranous. Wnt3a induced nuclear translocation of β-catenin, which was more evident in the SULF-CM-treated cells with higher levels of β-catenin (representative images Fig. 4b -enlarged panels). In our patient cohort, we assessed the relationship between SULF2, glypican-3, and β-catenin by IHC (Fig. 4c-g). While there were cases with CAF-SULF2, glypican-3 and tumour nuclear β-catenin, as well as cases reminiscent of in vitro data -with membranous accumulation of β-catenin -there were no consistent, significant associations with CAF-SULF2 (data not shown). The lack of an association between SULF2 and glypican-3 in the TGCA dataset (Spearman Correlation −0.09, p = 0.09) corroborated these data. We concluded that while CAF-SULF2 contributes to the regulation of Glypican-3/Wnt and β-catenin in HCC, supported by in vitro data, the in vivo heterogeneity was in keeping with additional mechanisms contributing to the poor prognosis attributed to CAF-SULF2 in our case series.

Paracrine Regulation of Tumour IKKβ/NF-κB Signalling Axis by Stromal SULF2
We went on to explore the contribution of SULF2 to paracrine crosstalk in the TME, using inhibitors of inflammation associated JNK, TAK1, and IKKβ signalling pathways. Stromal CM induced activation of JNK and STAT3 pathways in the tumour cells in the presence and absence of SULF2 (Fig. 6a), with JNK inhibition sup-pressing both SULF2-dependent and independent Hep3B spheroids growth (Fig. 6b). Conversely, SULF2-dependent growth of Hep3B spheroids was suppressed in the presence of either TAK1 or IKKβ inhibitors (Fig. 6c-e). TAK1 and IKKβ are kinases upstream of NF-κB signalling, with subsequent Western blotting confirming stromal SULF2 from COS-7 cells induced phosphorylation of the downstream target, NF-κB subunit RelA at serine 536 (RelA-P-ser536) (Fig. 6f). In the Hep3B spheroids, SULF2-induced RelA-P-ser536 persisted when cells were treated with sorafenib in the presence of control COS-7 SULF2-containing CM, but not when sorafenib was added to SULF2 KD COS-7 CM (Fig. 6e). Sorafenib inhibition of JNK and STAT3 phosphorylation was similar regardless of CM containing SULF2. These in vitro data are in keeping with NF-κB activation contributing to stromal SULF2-induced sorafenib resistance.
In our patient series, nuclear localization of RelA-P-ser536 in tumour cells was a dramatic feature in HCC biopsies with high levels of SULF2 (Fig. 6g). Nuclear Re-lA-P-ser536 positivity was scant/absent in tumour cells in the absence of SULF2. Nuclear RelA-P-ser536 in tumour cells and the presence of CAF-SULF2 were strongly correlated (Spearman's rho 0.722; p < 0.0001, Pearson χ 2 test 0.005), supporting stromal SULF2 as a regulator of a Re-lA-P-ser536 in human HCC in vivo.

Discussion/Conclusion
CAFs are a major source for growth factors, cytokines, and morphogens, and they comprise a marker of poorer prognosis in the TME [31,32]. Our study of histological SULF2 protein expression in biopsy rather than resection tissues has advanced data from previous transcriptomic datasets. In our cohort, SULF2 upregulation was in CAFs rather than in tumour cells in the majority of cases with a dramatic and detrimental impact on patient survival. SULF2 loss-of-function, gain-of-function, and SULF2 antibody studies were explored in 2D and 3D co-culture systems, confirming paracrine CAF-SULF2-induced proliferation, migration, invasion, and sorafenib resistance in HCC cells.
Mechanistically, we first explored the impact of exogenous SULF2 on glypican-3/Wnt signalling and while it is known that SULF2 modulates this pathway in cancers, the mutational status of β-catenin and other Wnt signalling aberrations were unknown in our biopsy cases and the additional contribution of SULF2 was unclear. SULF2 promotes Wnt/β-catenin signalling by 6-O desulfation of the morphogen Glyipcan-3 and is also reported to elevate glypican-3 expression. Our in vitro data showed that CM containing SULF2 promoted activation of a β-catenin reporter and increased Glypican-3 in HCC cells. In vivo, there was no clear-cut significant correlations between SULF2, glypican-3 and β-catenin immunohistochemistry. Given that there was a clear association between CAF-SULF2 and poor prognosis, we explored additional candidate mechanisms.
We considered TME signalling pathways in a wider context, showing that TGFβ-1-mediated upregulation of CAF-SULF2 activated PDGFβ/STAT3 pathway in CAFs, whilst also increasing the phosphorylation and activation of NF-κB pathway in the adjacent tumour cells, conferring therapy resistance. These findings are in keeping with growing insights into CLD and cancers. The PDGF/ PDGFR signalling pathway is of established paramount importance in these fields [33][34][35][36], with TGFβ1-mediated induction of metastasis and epithelial-mesenchymal transition also attributed to upregulation of stromal PDGFRβ [36]. STAT3, a downstream target of PDGFRβ in nonparenchymal cells [37], is also upregulated in the TGF-β1 rich profibrotic niche [38] and regarded as a hallmark of liver inflammation and malignancy [39]. Of note, both activation of STAT3 and NF-κB pathways in nontumour liver tissues have established prognostic relevance in patients with HCC [39,40]. The NF-κB pathway is also activated in many types of inflammation-associated cancers [41], often with advanced disease. It has also been associated with resistance or poor tolerability to chemotherapeutic agents in breast cancer [42], head and neck squamous cell carcinoma [43], and prostate cancer [44]. In HCC, in vitro data supporting a role for SULF2 as a mediator of sorafenib resistance has been previously reported, with SULF2 suppression or loss-of-function (SULF2N491K mutation) altering sorafenib sensitivity [45]. Our data add to this growing literature, highlighting the roles of stromal SULF2 and as an upstream regulator of TME TAK1/IKKβ/NF-κB signalling, contributing to therapy resistance. The direct mechanism remains to be elucidated. FGF/FGFR signalling, for example, is known to be modulated by SULF2 interaction with HS, with recent studies identifying FGFR1 as an upstream activator of TAK1 [46]. TGCA associations between SULF2 and FGFR1 were very highly significant (FGFR1 0.6281 p = 1. . Perhaps the most compelling aspect of this study, given the current excitement centred on the role of immunooncology therapies in HCC and the need to understand what underpins response to treatments, is the identified SULF2 association with the HCC immune class [23] and the possibility that SULF2 has a role to play in immune exhaustion. Evident in around 25% of HCC patients, the HCC immune subclass is divided into an active immune subclass and a TGFβ1 exhausted immune subclass, with the enrichment of SULF2 in this latter group is noteworthy. Rather than simply a bystander observation, our data Stromal SULF2-dependent Hep3B growth and sorafenib sensitivity were regulated by TAK1/IKKβ/NF-κB RelA-P-ser536, with nuclear p65 associated with stromal SULF2. a Western blotting of Hep3B whole cell lysates, showed that activation of JNK/ pJNK and STAT3/pSTAT3 pathways in Hep3B cells was driven by COS-7 CM independently of SULF2 and was sensitive to sorafenib treatment. b JNK inhibition limited Hep3B spheroid growth in both SULF2-dependent and independent fashion. In contrast, TAK1 (c) or IKKβ (d) inhibition limited SULF2-dependant Hep3B spheroid growth. e Similarly, Hep3B spheroid growth promoted by CM from SULF2-expressing LX-2 cells was suppressed by IKKβ inhibition. fi Activation of RelA-P-ser536 in Hep3B cells was dependent on control COS-7 CM and persistent in the presence of sorafenib, as was AKT activation. Quantification of RelA-P-ser536 (p-p65) relative to total p65 detected by Western blot is shown in fii, where data are mean ± SEM of 4 experiments. g IHC staining confirmed that SULF2 positivity in stromal cells associated with nuclear RelA-P-ser536 in adjacent tumour cells. Images are at ×20 magnification; scale bars represent 100-200 microns. Hep3B spheroid data are expressed as mean ± SEM of n = 10 experimental repeats. ns, not significant, *p = 0.05,**p < 0.01; ***p < 0.001, ****p < 0.0001. Liver Cancer 2022;11:540-557 DOI: 10.1159/000525375 support a role for SULF2 in the recruitment of tumourassociated macrophage into the cancer niche, via release of the chemoattractant CCL2. In pancreatic and colorectal cancers, αSMA positive CAFs recruit tumour-associated macrophages via secretion of IL6, CCL2, G-CSF, TGFβ1, and FAP [47,48], while activation of PDGFRβ/ STAT3 is associated with CCL2 release [49,50]. Here we have shown stromal SULF2 release following TGFβ1 stimulation, a striking correlation in vivo between SULF2 and TGFβ1, as well as TGFβ1 regulated TME signatures, SULF2 activation of stromal PDGFRβ/STAT3 and release of CCL2, with a significant association between stromal SULF2 and tumour macrophages in vivo. In vitro studies on human PBMCs have confirmed that the presence of SULF2 preferentially induces monocyte migration and induces in them a phenotype associated with immunosuppression and accelerated tumour growth.
In conclusion, we have provided tissue evidence and in vitro mechanistic data, supporting SULF2 as a potentially critical mediator of TME-driven HCC progression -worthy of further investigation as a prognostic and predictive biomarker but also as a therapeutic target alone or in combination with other therapies. DOI: 10.1159/000525375 sis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; and study supervision. Fiona Oakley -study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; statistical analysis; obtained funding; study supervision; and study supervision. Helen L. Reeves -study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; statistical analysis; obtained funding; and study supervision.

Data Availability Statement
All data generated or analysed during this study are included in this article and its online supplementary material. Further enquiries can be directed to the corresponding author.