Clinicopathological data of CCA patients
Seventy cases of CCA tissue micro array (TMA) liver sections, from male (62%) and female (38%) patients aged between 32 and 82 years old (median = 60 years old) were included in the study. Pathological examination confirmed that all cases were intra-hepatic adenocarcinomas from bile ducts. All 70 cases were subjected to surgical resection. Chemotherapy was administered to 6 patients before and to 24 patients after surgery, while 40 patients were not subjected to any chemotherapy. During a median of 1.3 (range, 0.01–5.20) months of follow-up, 34 of the 70 patients (48.57%) died. Follow-up information were available for the 36 patients surviving up to 100 months. The clinicopathological characteristics of CCA patients, including age, sex, tumor staging, tumor size, tumor-node-metastasis (TMN), histological grading and chemotherapy with respect to the level of LC3 and p62 expression in cancer cells and of IL-6 in CAFs are presented in Table 1.
Table 1
Clinicopathological variables of and IL-6 in CAFs, LC3 and p62 in cancer cells of 70 CCA samples in the study cohort
Factor | n | LC3 | P | P62 | P | IL-6 (in CAFs) | P |
Low | High | Low | High | Low | High |
Age (years) |
≤ 55 | 35 | 16 | 19 | 0.316 | 27 | 8 | 0.500 | 20 | 15 | 0.500 |
> 55 | 35 | 19 | 16 | | 26 | 9 | | 19 | 16 | |
Sex |
Female | 27 | 13 | 14 | 0.500 | 22 | 5 | 0.275 | 16 | 11 | 0.412 |
Male | 43 | 22 | 21 | 31 | 12 | | 23 | 20 | |
Histological types |
Non-pap | 33 | 18 | 15 | 0.316 | 26 | 7 | 0.388 | 17 | 16 | 0.335 |
Papillary | 37 | 17 | 20 | | 27 | 10 | | 22 | 15 | |
Tumor staging |
I/II | 39 | 17 | 22 | 0.168 | 29 | 10 | 0.496 | 22 | 17 | 0.544 |
III/IV | 31 | 18 | 13 | | 24 | 9 | | 17 | 14 | |
Tumor size (T stage) |
T1/T2 | 40 | 18 | 22 | 0.235 | 32 | 8 | 0.246 | 22 | 18 | 0.542 |
T3/T4 | 30 | 17 | 13 | | 8 | 9 | | 17 | 13 | |
Metastasis to lymph nodes (N stage) |
N0 | 44 | 23 | 21 | 0.342 | 33 | 11 | 0.099 | 25 | 19 | 0.412 |
N1/N2 | 26 | 12 | 14 | | 20 | 6 | | 14 | 12 | |
Metastasis to organs (M stage) |
M0 | 65 | 34 | 31 | 0.178 | 51 | 14 | 0.088 | 37 | 28 | 0.391 |
M1 | 5 | 1 | 4 | | 2 | 3 | | 2 | 3 | |
Resection status |
R0 | 49 | 24 | 25 | 0.500 | 37 | 12 | 0.604 | 26 | 23 | 0.339 |
R1 | 21 | 11 | 10 | | 16 | 5 | | 13 | 8 | |
Drug regimen after surgical resection |
No | 46 | 24 | 22 | 0.401 | 38 | 8 | 0.604 | 30 | 16 | 0.025* |
Yes | 24 | 11 | 13 | | 15 | 9 | | 9 | 15 | |
Recurrence after surgical resection |
No | 51 | 23 | 28 | 0.141 | 36 | 15 | 0.088 | 29 | 22 | 0.480 |
Yes | 19 | 12 | 7 | | 17 | 2 | | 10 | 9 | |
Correlation between IHC score and clinicopathological by Fisher’s exact probability test. |
*p value less than 0.05 was considered statistically significant. |
IL-6 expression in fibrotic stroma associates with poor prognosis in CCA patients
Immunohistochemical (IHC) staining of IL-6 and of the autophagy markers LC3 and p62 was performed and the staining intensity scored in epithelial and in stromal (namely, CAFs) cells. Faintly to negative staining of all the three proteins tested was found in normal bile duct epithelia (Fig. 1a). In contrast, CAFs and cancer cells showed positivity, yet to different extent. On the whole, the percentage of positive cells in human CCA tissues was 50% for IL-6, 24% for LC3 and 44% for p62. The IHC for IL- 6 expression in the cytoplasm was scored separately for epithelial cancer cells and in fibrotic (CAFs-enriched) area (Fig. 1). The staining ranged from absent (score 0) to strong (score 3) in CCA tissues. Twenty four of 70 cases (34%) were positive in only fibrotic area (Fig. 1a-iii), 7 cases (10%) were positive in cancer epithelial cells (Fig. 1a-iv), and 39 cases (56%) were positive in both cancer epithelial cells and fibrotic areas (Fig. 1a-v). No significant differences between staining patterns and clinicopathological features, such as age, sex, presence of lymph node metastasis or distant metastasis, grading, as well as tumor location and size were observed (Table 1). To be noted, Fisher’s exact test indicated a significant inverse correlation of IL-6 positive staining in cancer cell and fibrotic areas and the drug regimen status (Table 1, P < 0.025). Further, and most importantly, survival analysis demonstrated that high IL-6 in CAFs-containing fibrotic area was significantly associated with a shorter 5-year overall survival, in univariate (Fig. 1c; I = 0.024) and multivariate analyses (Table 2; HR = 0.562; CI = 0.338–0.934; P = 0.026).
Table 2
Multivariate Cox regression model for disease-free survival including IL-6 in CAFs, LC3 and p62 in cancer cells
Variable (no. of patients) | No. of patients who died (5-year survival cut-off) | Hazard ratio (HR) | 95% Confidence interval (CI) | P-value |
LC3 (in cancer) IHC score |
Low | 35 | 1 | | |
High | 35 | 0.401 | 0.236–0.681 | 0.001** |
p62 (in cancer) IHC score |
Low | 53 | 1 | | |
High | 17 | 0.699 | 0.400-1.221 | 0.208 |
IL-6 (in CAFs) IHC score |
Low | 39 | 1 | | |
High | 31 | 2.004 | 1.138–3.527 | 0.016* |
Age (years) | | | | |
≤ 55 | 35 | 1 | | |
> 55 | 35 | 0.505 | 0.308–0.828 | 0.252 |
Sex | | | | |
Female | 27 | 1 | | |
Male | 43 | 0.844 | 0.507–1.407 | 0.516 |
Histological types | | | | |
Non-papillary | 33 | 1 | | |
Papillary | 37 | 1.174 | 0.725–1.903 | 0.514 |
Tumor staging | | | | |
I | 11 | 1.336 | 0.380–2.129 | 0.900 |
II | 19 | 1.000 | 0.339–0.839 | 0.511 |
III | 19 | 4.670 | 1.698–12.33 | 0.414 |
IV | 21 | 2.070 | 0.330–0.869 | 0.013* |
Multivariate analysis by Cox proportional hazard regression. |
CI 95% indicates 95% confidence interval. |
*p value less than 0.05 was considered statistically significant. |
High expression of LC3 along with low expression of p62 in CCA cells correlate with better prognosis
Given the potential involvement of autophagy in CCA progression [24–26, 32–34], we sought to assess the IHC expression profile of the autophagy proteins LC3 (a marker of autophagosome) and p62/SQSTM1 (a marker of the autophagy cargo) in CCA TMAs (Fig. 2). The high IHC score of LC3 (Fig. 2a) showed a positive association with longer overall survival of CCA patients, in univariate (Fig. 2b; green line; P = 0.001) and multivariate analysis (Table 2; HR = 0.401; CI = 0.236–0.681; P = 0.001). Of note, no significant associations between p62 IHC staining and clinical outcome was found (Fig. 2c). Consistently, the prognosis was better in patients bearing a CCA with high expression of LC3 along with low (green line) or high (violet line) p62 expression, compared to patients bearing a CCA with low expression of LC3 (Fig. 2d). The Spearman’s correlation test was performed to examine the relationship between LC3 and p62 expressions in epithelial cancer cells (Fig. 2e). The pattern of high LC3 but low p62 showed a positive correlation in CCA tissues (Fig. 2e; rho = 0.518; P = 0.000), supporting the view that high LC3 was reflecting effective autophagy degradation of the cargo. Remarkably, the combined pattern of high LC3 and low p62 showed a significant correlation with the best overall survival, in univariate (Fig. 2d; green line; P = 0.001) and multivariate analysis (HR = 2.344; CI = 1.222–4.496; P = 0.01).
Correlation between IL-6, LC3 and p62 expression and with clinicopathologic features of CCA patients
At this point, it was mandatory to check whether inflammation (marked as IL-6 production in fibrotic stroma) and autophagy (marked as LC3 up-expression and p62 down-expression in epithelial cancer cells) were correlated and how the various combinations would correlate with clinical prognosis. With regard to the IHC protein expression scores, of the 70 cases, 18 cases (25.7%) were classified as low IL-6, low LC3 and low p62 (L/L/L); 11 cases (15.7%) were classified as low IL-6, high LC3 and low p62 (L/H/L); 1 case (1.4%) was classified as low IL-6, low LC3 and high p62 (L/L/H); 9 cases (12.9%) were classified as low IL-6, high LC3 and low p62 (L/H/H); whereas 14 cases (20.0%) were classified as high IL-6, low LC3 and low p62 (H/L/L); 9 cases (12.9%) were classified as high IL-6, high LC3 and low p62 (H/H/L); 2 cases (2.9%) were classified as high IL-6, low LC3 and high p62 (H/L/H); and, finally, 6 cases (8.6%) were classified as high IL-6, high LC3 and high p62 (H/H/H). Assuming the autophagy flux proceeds to completion when LC3 is up-expressed along with p62 down-expressed, the autophagy flux was clearly effective in 20 cases, of which 11 presented with low and 9 presented with high expression of IL-6. These numbers do not allow to draw any convincing correlation between the level of IL-6 in the stroma and the level of autophagy in cancer cells.
Next, we calculated the overall survival for the patients classified according to the above combinations. The Kaplan–Meier curves are shown in Fig. 3. It was found that the pattern of L/H/L, representing a low inflammatory stroma (low IL-6 staining) and an efficient autophagy flux in cancer cells (high LC3 and low p62) was significantly associated with the best prognostic clinical outcome (Fig. 3, green line; P = 0.007). In multivariate analysis, this scoring pattern was an independent and significant variable that predicted a favorable prognosis. The hazard ratio [HR] for death based on this variable was 2.535 (95% confidence interval [CI] 1.122–5.727; P = 0.025, Table 3). To be noted, the pattern with low IL-6 in the stroma and high LC3/high p62 in cancer cells (violet line; L/H/H; HR: 1.659, 95% CI: 0.813–3.384) also showed a good prognosis when compared to the other patterns having high IL-6 and/or low LC3 expression. Thus, disregarding the expression of p62, the combination of low (stromal) IL-6 with high (cancer) LC3 seems to provide the patients with a higher survival probability (HR: 0.317, 95% CI: 0.147–0.687 for L/H/L and HR: 1.659, 95% CI: 0.813–3.384 for L/H/H; see also Table 3), consistent with the data shown in Fig. 2d.
Table 3
Multivariate Cox regression model for disease-free survival including IL-6 in CAFs, LC3 and p62 in cancer cells (combined model)
Variable (no. of patients) (IL-6/LC3/p62) | No. of patients who died (5-year survival cut-off) | Hazard ratio (HR) | 95% Confidence interval (CI) | P-value |
Low/Low/Low | | | | |
No | 52 | 1 | | |
Yes | 18 | 0.577 | 0.312–1.069 | 0.081 |
Low/High/Low | | | | |
No | 59 | 1 | | |
Yes | 11 | 0.317 | 0.147–0.687 | 0.004** |
Low/Low/High | | | | |
No | 69 | 1 | | |
Yes | 1 | 0.322 | 0.043–2.405 | 0.269 |
Low/High/High | | | | |
No | 61 | 1 | | |
Yes | 9 | 1.659 | 0.813–3.384 | 0.164 |
High/Low/Low | | | | |
No | 56 | 1 | | |
Yes | 14 | 1.901 | 1.028–3.513 | 0.040* |
High /High/Low | | | | |
No | 61 | 1 | | |
Yes | 9 | 0.612 | 0.298–1.256 | 0.181 |
High /Low/High | | | | |
No | 68 | 1 | | |
Yes | 2 | 0.944 | 0.229–3.889 | 0.937 |
High /High/High | | | | |
No | 64 | 1 | | |
Yes | 6 | 1.067 | 0.460–2.478 | 0.880 |
Multivariate analysis by Cox proportional hazard regression. |
CI 95% indicates 95% confidence interval. |
*p value less than 0.05 was considered statistically significant. |
The role of adjuvant chemotherapy and of inflammatory and autophagy markers expression in patient survival
Next, we assessed how and whether postoperative chemotherapy had affected the patient survival depending on the inflammatory and autophagy levels in the cancer. Thirty patients received chemotherapy (6 before and 24 after surgery), which included gemcitabine for 8 patients (26.7%; 12.29% of total), cisplatin for 7 patients (23.3%; 10.61% of total) and 5-FU for 15 patients (50%; 21.33% of total). In an attempt to clarify the respective role of chemotherapy versus CCA inflammation/autophagy status in the clinical outcome, we estimated the overall survival in the patients with favorable status (based on above analysis), i.e., with low IL-6, high LC3 and low p62, that were subjected or not to chemotherapy (see Fig. S1). Additionally, to see whether the chemotherapy per se affected the clinical outcome, we have estimated the overall survival also for the other patients. The Kaplan-Meier curves are shown in Fig. 4. It is clearly apparent that patients who could not benefit of chemotherapy had the poorest outcome (Fig. 4a). Strikingly, the patients bearing a CCA with low IL-6 in fibroblasts and high LC3 and low p62 pattern in cancer cells (L/H/L) were the ones most benefiting the adjuvant chemotherapy. This outcome was statistically significant (Fig. 4a, purple line; L/H/L; P < 0.01). The univariate and multivariate analyses were performed to evaluate the prognostic factors affecting overall survival for the low IL-6, high LC3 and low p62 pattern with drug-based therapy (Table 4). Multivariate analysis indicated that the status of drug-treated plus low IL-6 plus high LC3 plus low p62 was an independent factor associated with longer overall survival. Overall, we found a positive correlation between drug-treated and low IL-6 (in fibroblasts) plus high LC3 plus low p62 in cancer cells (Table 5; Pearson r = 0.898, P < 0.01). Based on the previous observations, we assumed that high expression of LC3 could impact the chemoresponsivity regardless of the level of p62 expression
Table 4
Multivariate Cox regression model for disease-free survival including drug receivable, IL-6 in CAFs, LC3 and p62 in cancer cells (combined model)
Variable Drug regimen/IL-6/LC3/p62 | Hazard ratio (HR) | 95% Confidence interval (CI) | P-value |
No/Others | 1 0.595 | 0.082-4.300 | 0.607 |
Yes/Others | 1 0.154 | 0.019–1.267 | 0.082 |
No/Low/High/Low | 1 0.581 | 0.080–4.227 | 0.592 |
Yes/ Low/High/Low | 1 0.095 | 0.012–0.766 | 0.027* |
Multivariate analysis by Cox proportional hazard regression. |
CI 95% indicates 95% confidence interval. |
*p value less than 0.05 was considered statistically significant. |
Table 5
Pearson correlation coefficients between IHC scores of t drug receivable, IL-6 in CAFs, LC3 and p62 in cancer cells (combined model) components in human CCA tissues
| Drug regimen | Low IL-6/High LC3/Low p62 |
Drug regimen Pearson Correlation Sig. (2-tailed) N | 1 70 | .898** .000 70 |
Low IL-6/High LC3/Low p62 Pearson Correlation Sig. (2-tailed) N | .898** .000 70 | 1 70 |
*p value less than 0.05 was considered statistically significant. |
To test this hypothesis, we have estimated the OS in the cohort of 30 patients subjected to chemotherapy considering the group bearing a high LC3-expressing CCA along with low IL-6 in CAFs (regardless of p62 expression) (n = 13) vs the other combinations (n = 17). Again, the former group showed a better OS (Fig. 4b). Finally, we compared the OS of this group (13 patients) vs the group of patients bearing a tumor with high stromal IL-6 and low cancer cell LC3 (n = 7) and the group of patients bearing a tumor with other combinations (n = 10), and again the former group showed a better survival (Fig. 4c).
Expression of autophagy markers and clinical outcome in CCA-bearing patients from TCGA database
Above data refer to a cohort of patients living in a specific geographic area (province of North-Eastern part in Thailand) where liver fluke O. viverrini infection is a recognized cause of CCA. To see whether our observation could be extended to CCA cases from other countries, and likely with a different pathogenesis, we have interrogated the TCGA database. Thirty-four cases are reported in the database for which are available, with some exceptions, data on the mRNA expression and Copy Number Variation (CNV) of IL-6 and of the autophagy genes LC3, p62/SQSTM1, and BECN1, along with information on Overall Survival (OS). BECN1 is the first identified mammalian autophagy gene, and it is a haploinsufficient tumor suppressor coding for the BECLIN1 protein [35]. Details of this cohort of patients are provided in Table 1S. The oncoprint showing the alterations in BECN1 and MAP-LC3B gene expression is shown in Fig. S2a. Data on mRNA expression of LC3 were available for 33 CCA, of which 4 with high expression (2 patients underwent chemotherapy) and 29 with low expression (only 7 patients underwent chemotherapy). Though not statistically significant because of the small numbers, the trend shows that the patients bearing a CCA highly expressing LC3 have a better prognosis seen as OS (Fig. 5, panels a and b). Consistently, better prognosis was observed in patients bearing CCA with MAP-LC3B gene amplification (n = 5) compared to patients bearing CCA with diploid CNV (n = 22) or with shallow (monoallelic) deletion (n = 6) (not shown). Data on BECN1 mRNA expression was available for 34 patients (Fig. S2b). Quite surprisingly, the 29 patients bearing a CCA expressing low level of BECN1 showed a better OS (not significant; P = 0.23) than the 5 patients bearing a CCA expressing high level of BECN1 (Fig. S2c). It should be considered, however, that the latter patients were not subjected to chemotherapy whereas in the group of low BECN1 expressors 9 patients were subjected to chemotherapy. Also, to be noted, the 5 CCA with high BECN1 expressed low level of LC3 (Fig. S2d). Interestingly, 4 patients with low BECN1 and high LC3 tumor (2 of them underwent chemotherapy and 2 did not) showed a much better prognosis (not significant; P = 0.37) than 5 patients with high BECN1 and low LC3 tumor (Fig. S2e). Data on p62/SQSTM1 gene alteration and OS were available for 33 patients (Fig. S3). Though not statistically significant (P = 0.99) because of the small numbers, we found that 29 patients bearing CCA with low level of p62 (only 8 underwent chemotherapy) showed a much better prognosis than 4 patients bearing CCA with high level of p62 (no one underwent chemotherapy) (Fig. S3). Assuming that low accumulation of p62 in cancer cells is indicative of increased autophagy flux, these data are consistent with the above data on LC3. Data on IL-6 gene alteration and OS were available for 34 patients (Fig. S4), of which the three bearing CCA with high expression showed the worse prognosis (not significant; P = 0.71). Unfortunately, there are no data available in the TCGA dataset for the level of IL-6 specifically expressed in the stroma.
IL-6 secreted by CAFs inhibits autophagy and reduces the chemosensitivity of CCA cells
To explain the above data, we hypothesized that IL-6 released by CAFs negatively affected the chemosensitivity of neighboring CCA cells via inhibiting the autophagy stress-response to the drug. To this end, we have specifically inhibited the production of IL-6 by transfecting the CAFs with an appropriate si-RNA. The data shown in Fig. 6 prove that the si-RNA transfection effectively down-regulated the expression and secretion in the medium of IL-6 by CAFs, without altering their myofibroblast-like phenotype (as monitored by α-SMA expression).
Next, we tested whether and how the conditioned media from control or si-RNA-transfected CAFs would affect the autophagy regulation and the chemosensitivity of CCA cells. To this end, we employed the KKU-213 cell line that was shown to be very aggressive in previous study [30]. As a representative of chemotherapeutics, we chose 5-FU because it is the one mostly used for the adjuvant therapy in our cohort of patients. We monitored the cell growth of KKU-213 cells incubated for up to 96 h in medium from control or si-IL-6-transfected CAFs and exposed to 5-FU. SRB staining, which reflects the protein mass in the culture, indicated that cell growth was stimulated by CAFs-conditioned (scramble) medium while it was inhibited by conditioned medium derived from si-IL6-transfected CAFs (Fig. 7a). More importantly, the growth was greatly inhibited by 5-FU, and even more when the treatment was performed in the cells incubated with the CAFs-conditioned medium lacking IL-6 (Fig. 7a). The clonogenic assay confirmed that 5-FU could inhibit the proliferation of KKU-213 cells more effectively when incubated in the medium of CAFs avoid of IL-6 (Fig. 7b and 7d). Cytofluorometer analysis further proved that this effect was not merely due to a block of cell proliferation and instead was due to induction of cell death, very likely apoptosis based on the hypodiploid subG1 peak (Fig. 7c and 7e). To be noted, when treated with 5-FU the percentage of the subG1 population (referable to apoptotic cells) in control medium was of approx. 33% while in CAFs-derived medium was of approx. 23% (i.e., one-third less) and in IL-6 deficient CAFs-derived medium was of approx. 57% (i.e., it nearly doubled). Finally, we asked whether the CAFs-derived effects on chemo-responsivity to 5-FU was linked to autophagy in CCA cells. We further investigated on the link between the CAFs secretions (from scramble si-RNA- or siRNA IL-6-transfected) and the autophagy-dependent response to 5-FU. To this end, we looked at the expression of the autophagy proteins LC3 and p62, of the pro-apoptotic protein BAX and of the ant-apoptotic protein BCL-2. In drug-untreated cultures, compared to the protein expression in cells cultivated in standard medium, the IL-6 rich conditioned CAFs-medium limited the activation of autophagy (reduced conversion of LC3-I into LC3-II; accumulation of p62) and reduced the expression of BAX, whereas the IL-6 deficient CAFs medium (upon si-IL-6 transfection) greatly stimulated the autophagy flux, reduced the expression of BCL-2 and restored the expression of BAX (Fig. 8a-b). On treatment with 5-FU, autophagy was stimulated along with increased expression of BAX in the cells cultivated in standard medium, consistent with induction of a toxic stress response and onset of apoptosis (Fig. 8a-b). Remarkably, this response to 5-FU was largely impaired in the cells incubated with IL-6-rich CAFs medium (from scramble siRNA transfected) while it was enhanced when the cells were treated in the CAFs medium lacking IL-6 (Fig. 8a-b). To definitively assess the functional link between autophagy and apoptosis in response to 5-FU we performed a double-immunostaining for LC3 (to mark the cells with ongoing autophagy) and for BAX (to mark the cells undergoing apoptosis). The images in Fig. 8c show that 5-FU can induce BAX-mediated apoptosis only in the cells cultivated in standard medium or in IL-6 deficient CAFs-medium while it is not effective in cells cultivated in IL-6-rich CAFs-medium. The double-staining also demonstrates that apoptosis (BAX-positive) ensued in the same cells in which autophagy (LC3-positive) was induced (Fig. 8c).