1 Introduction

Cholangiocarcinoma (CCA) is a malignancy arising from the epithelium of the biliary tree. CCA is classified based on the anatomical origin as intrahepatic, perihilar or extrahepatic. Owing to late diagnosis, frequently in a stadium where metastasis has already occurred, and with limited or no treatment options, patients with CCA have a poor prognosis. Surgery is the only curative option for the minority of patients with a resectable tumor, and typically involves extensive procedures that are accompanied by considerable post-operative morbidity and mortality. Chemoresistance contributes to the low success rate of systemic chemotherapy in palliative treatment of CCA patients with unresectable tumors, metastatic spread, or those otherwise ineligible for surgery [1, 2].

We recently identified sulfatides as a class of lipids that -within the liver- are uniquely present in the bile ducts [3]. Sulfatides are derived by galactosylation of ceramide and subsequent 3’-sulfation of the carbohydrate moiety (Supplemental Fig. 1) [1, panels I-IVCD).

Fig. 1
figure 1

Sulfatides are present in tumor tissue of patients with intrahepatic cholangiocarcinoma. Mass spectrometry imaging was used to study the spatial distribution of sulfatides in control liver sections (panel I) and tumor tissue of patients with intrahepatic cholangiocarcinoma (iCCA, panel II), colorectal liver metastasis (CRLM, panel III) and hepatocellular carcinoma (HCC, panel IV). Imaged sections were stained with H&E (panel A and E), and adjacent sections were stained for the bile duct marker cytokeratin 19 (panel B and F). Molecular images were obtained by high-throughput screening of the entire section on a RapiFlex system (panel C and D), with a region of interest (black lined area in panel A and E) selected for further high resolution imaging on a Solarix FT-ICR (panel G and H). [C] Distribution of m/z 885.55 ± 0.1 (phosphoinositide [38:4], blue) as parenchymal marker, and m/z 876.62 ± 0.1 (ST [41:1], orange) as marker for bile ducts. [D] Distribution of m/z 885.55 ± 0.1 (phosphoinositide [38:4], blue) as parenchymal marker and m/z 892.63 ± 0.1 (ST-OH [41:1], yellow) as marker for the bile ducts. [G] Distribution of m/z 876.6240 ± 0.001 (ST [41:1], orange). [H] Distribution of m/z 892.6189 ± 0.001 (ST-OH [41:1], yellow). Representative images are shown

Full size image

3.2 Total sulfatide composition is largely similar in iCCA and control tissue

Whereas the sulfatide-positive area was evidently larger in iCCA tissue (Fig. 1), the median summed sulfatide intensity per pixel was similar in comparison with bile ducts (47878 [15764–267010] vs. 31776 [16430–143506] in iCCA and controls, resp.; P = 1.00) (Fig. 2A). In contrast, median total sulfatide signal was notably higher (18–47 fold) in CRLM in comparison with control (P = 0.027) and HCC tissue (P < 0.001), with a trend to elevation relative to iCCA tissue (P = 0.068). Median total sulfatide intensity in HCC was comparable to control and iCCA. Note that there is substantial within-group variation in total sulfatide intensity, which is smallest in the HCC group.

Fig. 2
figure 2

Total sulfatide abundance is elevated in tumor tissue of patients with colorectal liver metastasis. Summed sulfatide intensities per pixel from control regions (bile ducts, green) and tumor regions of intrahepatic cholangiocarcinoma (iCCA, orange), colorectal liver metastasis (CRLM, blue) and hepatocellular carcinoma (HCC, red) for all sulfatides (panel A), hydroxylated species (panel B) and non-hydroxylated variants (panel C), as well as the ratio between the latter two (panel D). Dots represent individual samples, with the median intensity depicted by a solid line. Note that all y-axis scales are logarithmic. Statistical significance for between-group comparisons is depicted. * denotes P < 0.05, *** denotes P < 0.001

Full size image

In general, in each group hydroxylated sulfatide species appeared more abundant than non-hydroxylated variants. The median sum of both hydroxylated and non-hydroxylated sulfatides, as well as their ratio, was similar in iCCA relative to controls (Fig. 2BCD). The increased total sulfatide intensity in CRLM was due to elevations of both non-hydroxylated (median tenfold, P = 0.028) and hydroxylated (median 26 fold, P = 0.017) sulfatide species, with the latter having the largest impact in absolute terms. The ratio of hydroxylated over non-hydroxylated sulfatide species was not distinct in CRLM and control tissue. No compositional changes were apparent in HCC with regard to (non)-hydroxylated species when compared with controls.

We observed identical patterns when looking at (un)saturated sulfatide species. In general, in each group unsaturated sulfatides were more abundant than saturated variants (Fig. 3). Median summed intensity of both unsaturated and saturated sulfatide species was similar in iCCA and controls (P = 1.00 and P = 0.067, resp., Fig. 3ABC). In CRLM, the elevated sulfatide levels were due to both rises in unsaturated (median 25 fold; P < 0.019) and saturated species (median 23 fold; P = 0.006). The ratio of unsaturated over saturated sulfatide species was indistinguishable in the three tumor groups and the control. There is considerable variation in this ratio, especially in the control and iCCA groups.

Fig. 3
figure 3

The ratio of unsaturated over saturated sulfatide species is indistinguishable in control tissue and tumor groups. Summed sulfatide intensities for unsaturated (panel A) and saturated species (panel B) and their ratio (panel C), as well as for odd-chain (panel D) and even-chain sulfatides (panel E) and their ratio (panel F) are depicted for the respective groups. Dots represent individual samples, with the median depicted by a solid line. Note that all y-axis scales are logarithmic. Statistical significance for between-group comparisons is depicted. * denotes P < 0.05, ** denotes P < 0.01

Full size image

A final distinction was made between sulfatide species with even and odd acyl chain substitutions. In general, even-chain sulfatide appeared slightly more abundant in each group (Fig. 3DE). Increased levels of sulfatides in CRLM comprised both elevations of even-chain species (median 39 fold increase vs. control; P = 0.018) and odd-chain species (median 11 fold increase vs. control; P = 0.007). Relative to controls, the ratio of odd-chain to even-chain sulfatides was lower in CRLM (-3.2 fold, P = 0.009), while elevated in HCC (+ 3.8 fold, P = 0.043) (Fig. 3F).

We choose to not analyze between-group differences at the level of individual sulfatide species. It was noted though that sulfatides with C18 N-acyl chains (e.g. ST-OH [36:0]) were absent in all groups. This may reflect absent or low hepatic expression of ceramide synthases with C18 acyl length specificity (i.e. CerS1 and CerS4). Sulfatides with acyl chain lengths between C38-C44 comprised the dominant species in all groups (Supplemental Fig. 2), in line with the substrate specificity (C20-C26) of the major hepatic ceramide synthase CerS2 [21].

3.3 Specimen subgroups have distinct sulfatide patterns

Unsupervised hierarchical clustering showed an indicative grou** of the various specimens in distinct (sub)groups (Fig. 4). Overall sulfatide profiles differed between the various tumor types (see also Supplemental Table 4 and Supplemental Fig. 3). As particularly striking example, ST-OH [42:1] was highly abundant in tumor areas of CRLM or iCCA, but was hardly observed in HCC tumor areas (Fig. 4, Supplemental Table 4).

Fig. 4
figure 4

Specimen subgroups have distinct sulfatide patterns. Unsupervised hierarchical clustering of all analyzed samples was performed based on the intensities of constituent sulfatides. The clustergram depicts the scaled intensities of all detected sulfatides, sorted per group. Boxed regions (red coloring) are discussed in the text

Full size image

In case of the controls, a scattered image was obtained as some specimens show strong signals of ST-OH [39:2], whilst others displayed elevated levels of ST-OH [34:1] and ST-OH [42:1], which are dominant species in the CRLM and iCCA groups as well. The iCCA group showed inter-tumor heterogeneity, and the samples clustered in two (overlap**) groups. One subgroup showed elevated levels of ST [37:0] [10 out of 17], whilst the other group had elevated levels of ST-OH [42:2] [10 out of 17]. ST [37:0] was also observed being upregulated in a portion of HCC tumors. The CRLM tumors showed a remarkable homogeneity amongst the 10 examined specimens. Nearly all samples contained upregulated levels of four hydroxylated sulfatide species, viz. ST-OH [34:1], ST-OH [40:1], ST-OH [41:1] and ST-OH [42:1]. The abundance of these compounds in iCCA and controls was much more heterogenous.

A number of sulfatides were clearly upregulated in the HCC specimens, viz. ST [33:0] (6 out of 10) and ST [39:1] (5 out of 10), as well as ST-OH [37:1] (7 out of 10) and ST-OH [30:1] (4 out of 10). ST [37:0] was clearly enhanced in some HCC samples, though also upregulated in part of the iCCA tumors. ST-OH [39:2] was observed in both HCC tumors and the majority of control samples.

To evaluate intra-tumor heterogeneity, we analyzed iCCA specimens from 6 patients where a second specimen of (a different part of) the tumor was available. Intensities of (non)hydroxylated sulfatide species in the tumor area were similar (Supplemental Fig. 4). Likewise, spatial distribution of sulfatide species was grossly comparable.

3.4 The ratio of unsaturated to saturated sulfatides is related to shorter disease-free survival in iCCA

We conducted correlation analysis to evaluate whether the sulfatide pattern is linked to tumor biology and/or clinical outcomes in patients with iCCA. To this end, (disease-free) survival, tumor recurrence, tumor grade and tumor size/extension and lymph node involvement (T and N components in TNM staging system) were correlated to the total intensities of sulfatide classes and ratios derived thereof.

The tumor differentiation state (G grade), tumor stage and lymph node status (TNM classification) were not related to total sulfatide intensity or compositional features (Supplemental Figs. 5–7).

Tumor recurrence within the follow-up period of 2 years, was noted in 10 out of 15 patients who underwent actual liver resection. All operated patients received adjuvant chemotherapy (capecitabine or gemcitabine plus cisplatin). Note that two patients received an exploratory laparotomy but were not resectable at first surgery (Supplemental Table 1). A significant correlation was found between disease-free survival and saturation index (P = 0.002) (Fig. 5A). Patients with a high ratio of unsaturated to saturated sulfatides showed earlier tumor recurrence, with a median time to recurrence of 10 and 20 months in high and low ratio groups, respectively. Disease-free survival was not linked with other sulfatide compositional features (Supplemental Table 5).

Fig. 5
figure 5

A high ratio of unsaturated over saturated sulfatides in iCCA tumor area is associated with earlier tumor recurrence. Disease-free survival (P = 0.002) (A) and overall survival (P = 0.15) (B) in relation to the saturation ratio (unsaturated over saturated sulfatides) in patients with iCCA. Red is high ratio, blue is low ratio, groups separated on 50% quantile

Full size image

There was no in-hospital mortality or mortality within 90 days, in iCCA patients undergoing actual liver resection. Cancer-related death within 2 years of operation occurred in 8 out of 17 patients, with demise of two other patients unrelated to their malignancy. The saturation ratio was associated with overall survival in Cox proportional hazard analysis (P = 0.041, Supplemental Table 5), but not in Kaplan–Meier analysis (Fig. 5B). No other correlations were apparent for sulfatide characteristics and overall or disease-specific survival (Supplemental Table 5).

4 Discussion

We aimed to determine if sulfatides are present in tumors originating from the intrahepatic bile duct, and whether sulfatide compositional features are related with clinical outcomes. Here, we demonstrate for the first time that sulfatides are present in tumor tissue of patients with iCCA. Although the sulfatide-positive area was evidently larger in iCCA tissue, summed sulfatide intensity per pixel was similar in comparison with bile ducts. Likewise, global composition of sulfatides in iCCA was not distinct from control tissue. Strikingly though, a high ratio of unsaturated to saturated sulfatides was associated with reduced disease-free survival in patients with iCCA.

In line with our earlier observation [3], sulfatides were specifically present in the bile ducts but were not detected in the parenchyma in tumor-distal control tissue (Fig. 1). In iCCA the entire tumor area was sulfatide-positive, with a total of 44 distinct sulfatide species detected in widely varying intensities (Fig. 4, Supplemental Tables 2–4). Considerable inter-tumor variability in total sulfatide intensity was observed in iCCA (Fig. 2A). Results from a second tumor region (available for 6 patients) showed no distinct changes in sulfatide intensities between the two replicated regions (Supplemental Fig. 4), demonstrating the robustness of the method and indicating that inter-tumor variation was not due to sampling location. Despite reported links with aggressiveness of tumors, no correlations between total sulfatide intensity with overall and disease-free survival (Supplemental Table 5) or tumor biology (i.e. T and N component in TNM stage, Supplemental Figs. 5–7) were apparent in iCCA. When looking at distinct classes of sulfatide species, relative abundance of unsaturated over saturated species was found to be associated with shorter disease-free survival (Fig. 5A). The molecular events linking sulfatide saturation index with earlier iCCA tumor recurrence are unclear yet.

It is tempting to speculate that in iCCA tumors with an elevated ratio of unsaturated over saturated sulfatides, enzymes coding for desaturases (e.g. the Δ9-stearoyl desaturase 1, SCD1) are elevated, thus, promoting incorporation of unsaturated acyl chains in ceramide and eventually showing up in sulfatides. Increased expression of the lipogenic enzyme SCD1 has been observed in cancer cells originating from a variety of tissues, and this has been linked with enhanced tumorigenic properties [22, 23]. Inhibition of SCD1 was shown to inhibit cell proliferation in among others lung cancer cell lines and in cell lines and xenografts of lipogenic-subtype pancreatic tumors [22, 24]. SCD1 inhibition is considered as therapeutic target in the treatment of cancer, and preclinical studies are ongoing. Inter-tumor variation in expression of the dihydroceramide Δ4-desaturases DEGS1 and DEGS2 that introduce a double bond in the sphingolipid backbone to generate ceramide, may also contribute to heterogeneity in the sulfatide saturation index in the iCCA group. Detection of saturated sulfatides (e.g. ST [34:0], with d18:0 as tentative sphingoid backbone and a C16:0 N-acyl chain) suggests that dihydroceramides can be further metabolized to more complex sphingolipids. DEGS inhibition results in diminished proliferation of various cell types in vitro [25]. It is thus conceivable that the balance between unsaturated and saturated sulfatides impacts on cell growth. It remains to be determined whether SCD1/DEGS inhibition is of benefit to patients with iCCA, which are typically diagnosed in an advanced stage of the disease.

The presence of sulfatides in liver tumor tissue was not restricted to iCCA, but also observed in tumors originating from liver parenchymal cells (HCC) or the colon (CRLM). In the latter group, total sulfatide intensities were notably higher (up to 47-fold) relative to control and HCC tumor tissue. All structural classes (i.e. sulfatides containing (non)hydroxylated, (un)saturated or even/odd chain N-acyl groups) contributed to elevated sulfatide levels in CRLM. Elevated expression of sulfatides was previously reported for primary colon carcinoma and cell lines derived thereof, and this was associated with increased metastatic potential [26, 27]. Apparently, sulfatide overexpression was maintained upon metastatic spread to the liver. During the follow-up period, tumor recurrence occurred in 6 out of 10 patients with CRLM, with an all cause-mortality of 50% in this group. Cox proportional hazard analysis revealed no significant associations between sulfatide compositional features and overall or disease-free survival in the CRLM group (Supplemental Table 6).

It is conceivable that enhanced expression of biosynthetic enzymes (e.g. GAL3ST1) contributes to elevated sulfatide levels in CRLM. It should be noted though that no direct comparison could be made with normal colon and primary colon cancer tissue, and the different tissue origin may account for the elevated levels in comparison with control and primary liver tumor tissue (i.e. HCC). Previously, overexpression of GAL3ST1 was demonstrated in the context of renal clear cell carcinoma and ovarian cancer [12, 28, 29]. Given that both hydroxylated and unsaturated sulfatide species had substantial quantitative contributions to elevated sulfatide content in CRLM, overexpression of the fatty acid 2-hydroxylase FA2H and fatty acid/sphingolipid desaturases is likely in CRLM.

Although we did not observe correlations between sulfatide hydroxylation parameters and survival in CRLM patients, elevated levels of hydroxylated sulfatides may impact on tumor biology [30,31,32,33]. Empirical evidence for such action is thus far limited and circumstantial, and the literature indicates that effects may be tumor dependent. Hydroxylation of glycosphingolipids was associated with a highly tumorigenic and invasive phenotype of urothelial cell lines [34]. Moreover, elevated levels of hydroxylated glycosphingolipids were associated with chemoresistance of ovarian carcinoma cell lines. On the other hand, FA2H inhibition resulted in reduced growth and enhanced cisplatin sensitivity of gastric cancer cells [35]. Likewise, a tumor suppressive role has been proposed for FA2H in breast cancer [30]. Note that in these latter instances, changes were not directly linked to hydroxylated glycosphingolipids, and could have been effects of hydroxylated fatty acid in free form or after incorporation in lipids other than glycosphingolipids. Likewise, in the context of colorectal cancer, fatty acid 2-hydroxylation was reported to inhibit tumorigenic properties of colorectal cancer cells [36]. Therapeutic potential of FA2H and/or SCD1/DEGS modulation in CRLM remains to be determined.

The third liver tumor in which we observed sulfatides was HCC. Interestingly, the sulfatide species predominating in control bile ducts and iCCA and CRLM tumor tissue, viz. ST-OH [42:1], was virtually absent in HCC tissue (Fig. 4). Instead, the odd-chain species ST-OH [37:1] was most abundant in HCC (Fig. 4, Supplemental Table 4). Of note, odd-chain sulfatides could be detected in all groups, but their relative abundance (Fig. 3) was increased in HCC. This is reflected in the clustergram (Fig. 4), where HCC appears to have an odd-chain sulfatide signature. Larger studies are required to test specificity and clinical relevance of this putative signature, e.g. in differential diagnosis of HCC and iCCA which can be problematic upon clinical imaging or in histopathological diagnosis [37, 38]. It is tempting to speculate that peroxisomal alpha-oxidation, which is prominent in peroxisome-dense hepatocytes, results in generation of odd-chain fatty acids that can be incorporated into the cellular lipid pools, including ceramides and sulfatides. During follow-up, tumors recurred in 4 out of 10 patients with HCC, and all cause-mortality in this group was 20%. Cox proportional hazard analysis revealed no significant associations between sulfatide compositional features and overall or disease-free survival in these patients (Supplemental Table 7). In an earlier study high expression of sulfatides (based on immunoreactivity) in HCC tumor tissue and HCC-derived cell lines was linked with metastatic behavior [39]. Furthermore, enhanced synthesis of sulfatides through overexpression of the sulfotransferase GAL3ST1 increased metastatic potential of HCC cell lines [39]. Further exploration of a tumorigenic link with aspects of sulfatide metabolism is also warranted for HCC.

To our knowledge, this is the first report on spatial sulfatide distribution in hepatobiliary tumors. The expression of sulfatides in the three studied tumor types raises questions on their involvement in tumorigenesis and tumor progression. Sulfatides have been recognized as adhesive molecules in the outer leaflet of the plasma membrane, that bind to specific target proteins [4, 12]. For instance, sulfatides serve as native ligands for P-selectin on the cell surface of platelets, thus, contributing to metastasis of breast cancer cells [5, 40]. Sulfatides are furthermore components of lipid rafts, which are dynamic signaling hubs that control cell proliferation and differentiation [41, 42]. It is conceivable that hydroxylation and unsaturation(s) in the N-acyl chain of sulfatides, affects its packing in lipid rafts and/or lipid raft dynamics/function. In the context of clear cell renal carcinoma, enhanced sulfatide-mediated platelet binding protected tumor cells from cytotoxic assaults by Natural Killer cells. Accordingly, sulfatides may contribute to evasion of a tumor-directed immune response. Increased expression of GAL3ST1 was associated with decreased survival in primary clear cell renal carcinoma.

Whereas ceramides are generally considered to induce apoptosis, its galactosylated derivative βGalCer has anti-apoptotic effects in breast cancer cells [12, 40]. Likewise, the lactosylated derivatives (i.e. LacCer) appear to act as pro-survival signal in the context of fluke-associated CCA, where accumulation of hydroxylated LacCer[d18:1-h23:0] was associated with a drastic reduction in patient survival [16]. With regard to sulfatides, available evidence indicates that anti- or pro-apoptotic actions of sulfatides are context specific. It has been reported that sulfatides act as pro-apoptotic molecules, making cancer cells more prone to environmental stressors such as hypoxia and anticancer drugs [43, 44]. Sulfatides within tumors may promote apoptotic cell removal and alter the phenotype of tumor-associated macrophages [45, 46]. On the other hand, enhanced synthesis of sulfatides through overexpression of GAL3ST1 results in increased metastatic potential of tumor cells, indicating diversion of apoptotic to pro-survival signaling [5]. Further research is required to address the role of sulfatides in hepatobiliary tumors, and a potential therapeutic role of modulating sulfatide metabolism.

Our study has several limitations. Firstly, the iCCA samples were obtained from patients operated 2–3 years ago, and follow-up after discharge from hospital was limited to 2 years. Information on longer-term survival is not available for these and the CRLM/HCC patients. Secondly, the tumor samples were in 15 out of 17 cases obtained from resectable patients, and this could result in bias to a more favorable outcome and/or tumor biology. Only patients with non-locally advanced tumors (typically less aggressive tumors with less metastatic potential) are considered for surgical treatment. Overall survival increases from less than one year for inoperable patients, to a median survival of 27 months for patients with iCCA undergoing surgery (5-year survival between 20 and 40%). Thirdly, the sample size was relatively small and unbalanced. This is especially reflected in the tumor grades, with only a single patient having a grade 1 tumor and higher-grade tumors being limited to five patients each, and three tumors classified as grade 2–3 (Supplemental Table 1). With the achieved sample sizes it is realistic that the study is underpowered to detect differences in survival outcomes in relation to sulfatide compositional features. The statistical findings should therefore be regarded as indicators for further investigations, since at least an effect on disease-free survival was observed in patients with iCCA. Validation of the initial findings in independent patient cohorts of adequate size, will be necessary. Given the distinct presence of sulfatides in the tumor but not the tumor-distal area, a potential implication of our findings is the (real-time) monitoring of sulfatides during resection, to assess tumor-free margins during surgery in patients with hepatobiliary tumors as is currently being validated for other tumor types such as e.g. breast cancer and thyroid cancer [47,48,49].

In conclusion, expression of sulfatides was observed in tumor tissue of patients with iCCA, HCC and CRLM. In patients with iCCA, a high ratio of unsaturated to saturated sulfatide species was associated with reduced disease-free survival. In CRLM, markedly elevated levels of sulfatides were apparent, along with increased fractions of hydroxylated and unsaturated sulfatide species. HCC was marked by absence of sulfatide species common to bile ducts and iCCA and CRLM tissue, and relative abundance of odd-chain sulfatides. Follow-up studies are required to confirm initial links between sulfatides and survival in iCCA .