IFN-β inhibits growth of PH5CH and A549 cells.

Among type I IFNs, IFN-β reportedly exerts the strongest antiproliferative effect due to its high receptor affinity [16, 25,26,27]. To quantify antiproliferative effects with high temporal resolution, we monitored cell growth in a live-cell imaging approach on the IncuCyte platform. Both A549 alveolar carcinoma as well as immortalised non-neoplastic PH5CH hepatocytic cells [45] exhibited typical exponential growth. Confirming the antiproliferative effect of IFN-β, both cell lines tested grew significantly slower when treated with 1000 IU/ml of IFN-β (Fig. 1a, b left panels). We found the growth rate of A549 cells to be reduced to 83% upon IFN-β treatment, and to even 63% for PH5CH cells (Fig. 1a, b right panels).

We hypothesized that this severe proliferative disadvantage of cells exposed to IFN-β may constitute a substantial selective pressure in quickly proliferating tissue such as malignancies, and, hence, shape the evolution of tumours. To simulate this evolutionary effect in vitro, we cocultured wildtype A549 cells (labelled by cytosolic GFP expression) with “IFN-blind” A549 cells, lacking type I and III IFN receptors (IFNAR1/IRNLR1 double KO, DKO) (labelled by nuclear mCherry). By continuous exposure to 5000 IU/ml IFN-β and frequent passaging, we exposed the cells to a strong selective pressure. After 18 days of passaging, the fraction of IFN-insensitive cells increased eightfold in the IFN-treated culture as compared to the fraction in cocultures kept in IFN-free medium (Fig. 1c, d). This confirmed our primary assumption that reduced sensitivity to IFN-β could constitute a proliferative advantage in an IFN-rich microenvironment.

Selection of IFN-β-resistant hepatocytes by long-term passaging

Within a population of cells, the response to IFN is heterogeneous and a fraction of cells remains unresponsive [55]. However, there is not much known about the molecular underpinnings of this effect, in particular it is unclear whether it is limited to random transcriptional noise [55] or whether there is a contribution of more stable effects, such as epigenetics. In the latter case, continuous exposure to IFN may favour the selection of cells with reduced sensitivity to IFN-mediated growth inhibition and may therefore contribute to oncogenic processes. To investigate this possibility, we subjected a population of non-malignant PH5CH cells to stringent selection pressure by IFN-β and frequent passaging for six weeks (Fig. 2a). To control for passaging related effects, one well was cultured in the absence of IFN. After 6 weeks of selection, cells were kept for 1 week in IFN-free medium to re-equilibrate in order to omit effects based on random transcriptional noise or transient upregulation of feedback regulators of IFN signalling. We then assessed growth rates of the cells in the absence or presence of different doses of IFN-β. Cells that underwent six weeks of IFN-pressure (PH5CHIFN−β) were substantially less affected by IFN-β-treatment than mock-selected cells (PH5CHmock), indicating successful selection of a less IFN-sensitive population (Fig. 2b, see Additional file 2: Fig. S2a for shorter selection periods). At the highest IFN dose PH5CHIFN−β proliferated almost twice as fast (0.022 vs 0.012 h−1) as their PH5CHmock counterparts (Fig. 2c). Notably, the same effect was present in each of the three populations that independently underwent six weeks of IFN-selection (Additional file 2: Fig. S2b, c). Moreover, we observed a slight increase of the basal proliferation for PH5CHIFN−β over PH5CHmock cells. Although this effect was small, it occurred in all three populations (Additional file 2: Fig. S2c). We further assessed the stability of the observed phenotype of decreased IFN-sensitivity after three more weeks of normal, IFN-free cultivation. Indeed, PH5CHIFN−β maintained their relative IFN-resistance compared to PH5CHmock albeit to a somewhat reduced extent (Additional file 2: Fig. S2d, e). In general, the reduced IFN-sensitivity was less characterised by a shift of the dose–response (IC50) but rather by an overall reduced growth inhibition across all concentrations, including the highest ones (10,000 IU/ml) (Fig. 2b, Additional file 2: Fig. S2d, b). Importantly, PH5CHmock behaved very comparable to a freshly thawed aliquot of PH5CH cells that have not undergone the 6-week selection process, underscoring that indeed the presence of IFN constituted the major selective pressure (Additional file 2: Fig. S2d, e).

Fig. 2

In vitro selection of IFN-β-insensitive cell populations upon long-term treatment. a Schematic overview over selection process. b After long-term (6 weeks) IFN-β treatment and IFN-free recovery (1 week), cells were stimulated with IFN-β and growth rates were determined relative to mock-treated control (included for simplicity at x = 100). c Absolute growth rates (duplications per hour) of PH5CHIFN−β and PH5CHmock unstimulated and stimulated with 8000 IU/ml IFN-β. PH5CHIFN−β corresponds to three independently passaged long-term selected cell populations, PH5CHmock to three technical replicates of one population. Data displayed as mean ± SD

Taken together, our results indicate that even among a genetically homogeneous population of cells stable phenotypes exist that mediate increased resistance against the growth inhibitory effects of IFN signalling. Such cells exhibit a proliferative advantage when IFN is continuously present in the medium, and hence can be enriched by selection.

Characterisation of a panel of ten hepatoma cell lines

Although tumour cells exhibit dysregulated proliferation, it is known that in many cases they remain susceptible to the growth inhibitory effects of IFN [16, 56, 57]. Particularly for tumours developing in the context of chronic inflammation, e.g., in chronic HCV infection, continuous presence of IFNs in the tumour microenvironment [39] may present a dominant selective pressure. We and others [44] therefore hypothesise that many advanced tumours have undergone selection to overcome the cytostatic effects of IFNs.

In order to test this hypothesis, we assessed a panel of rather well-characterised tumour cell lines regarding their growth response to IFN-β. We chose liver-derived cancer lines because of their high likelihood to have developed in the context of chronic virus infection. Eight stemmed from hepatocellular carcinoma (HCC) and two from juvenile hepatoblastoma [58] (Table 1). As a control, we used PH5CH cells, as they reportedly are non-neoplastic and have been immortalised in vitro. Previous HBV infection can be detected in some of the cell lines by presence of viral DNA integrates [59, 60], and in HuH1 and PLC we detected ongoing HBsAg secretion (Table 1, Additional file 3: Fig. S3a). HCV was unknown at the time of establishment of most of the cell lines and due to the lack of DNA integration cannot be detected retrospectively. Despite presence of viral components in some lines, there was no basal production of IFN-β in any of them (Additional file 3: Fig. S3b).

Table 1 Characteristics of studied cell linesAssessment of malignity-associated characteristics

Dedifferentiation is a hallmark of advanced cancer. A plethora of cellular markers allows assessing the differentiation state of hepatocytes and several studies have characterised the ten selected hepatoma cell lines [63, 64, 67, 68, 72]. Although nomenclature varies, there is an overall consensus discriminating the cell lines into better differentiated “epithelial-like” and poorly differentiated “fibroblast-like” (Table 1). This grouping fits well with the histological origin of the tumours: the tissue of HLE and HLF was described as “undifferentiated hepatocellular carcinoma” [70] and that of SNU182 and SNU387 as “poorly differentiated” HCC [71], with grades III and IV representing “embryonal-cell types” based on the Edmondson-Steiner’s classification [73].

Malignancies are defined as “diseases in which abnormal cells divide without control and can invade nearby tissues” [74]. In addition to the published information (Table 1), we therefore assessed three different parameters of the selected cell lines as a measure of functional malignity: cell migration, invasion and growth rate.

Regarding motility, we measured the cells’ ability to migrate into a scratch wound. Figure 3a shows exemplary images at fixed time points (0 and 24 h post scratching) for selected cell lines. PLC and HuH1 barely moved beyond the initial scratch margins, whereas the scratch was almost completely closed by HLE and SNU387 cells. We evaluated migration rates for all cell lines in live-cell imaging (Fig. 3b). Indeed, three of the four poorly differentiated cell lines, HLE, HLF and SNU387, were among the fastest moving cells. All epithelial-like cell lines exhibited a markedly reduced migration rate with PLC being the least motile cells. In general, we found a good correlation between reported differentiation status and migration. Only two cell lines were breaking this scheme–SNU182, described to be dedifferentiated, showed a migratory behaviour similar to the epithelial-like cell lines and PH5CH, believed to be non-malignant [45], exhibited high migration comparable to the dedifferentiated HLE and HLF cell lines (Fig. 3b). As PH5CH were immortalised by transduction with a viral oncogene, they might have acquired motility characteristics untypical for hepatocytes. As a more authentic and primary-like hepatocyte system, we therefore also tested hepatocyte-like cells (HLCs) that we differentiated from human derived induced pluripotent stem cells (iPSC) [47, 75]. In a transwell migration assay, we compared their migratory capacity to one of the epithelial-like hepatoma cell lines, HuH7. Indeed, terminally differentiated HLCs were hardly able to migrate through the porous matrix of the transwell, while HuH7, comparable to the scratch wound assay, showed robust motility (Fig. 3c). Remarkably, only one of the few iPSC-derived cells that migrated through the membrane, expressed the HLC differentiation marker albumin (Additional file 3: Fig. S3c), while the majority of cells that were able to migrate appeared to be fibroblast-like, a cell-type which is likely spontaneously co-differentiated along the HLC differentiation protocol [47, 75]. This confirmed that normal, non-malignant hepatocytes have no significant migratory potential, much akin to PLC.

Fig. 3

Functional assessment of the malignity of ten liver cancer cell lines. a + b Scratch wound migration assay. a Representative brightfield microscopy images; black lines: initial wound, red signal: nuclei. b Left: Cell confluence [%] in wound area over time. Right: Determination of migration rates (Δconfluence [%]/h) in the period 0–24 h by linear regression. c Migration of HLC and HuH7 through uncoated transwell membranes, Hoechst staining after 24 h. Left: Quantification. Right: Representative images (10 × magnification). d + e Transwell migration/invasion assay. d Representative membranes (matrigel coated, control: uncoated), stained with crystal violet 24 h after seeding. e Invasiveness, determined as the ratio of total cell count penetrating a matrigel-coated membrane vs. the corresponding uncoated membrane. f Proliferation rates (untreated). Panel (b) displays mean ± SEM, all other graphs mean ± SD of three independent repetitions

While migration is essential for metastasis, another prerequisite of metastatic cells is their capacity to penetrate the basal membrane [76]. This feature, invasiveness, requires the expression and activation of extracellular proteases [77]. Hence, as a second measure of malignity, we assessed the cell lines’ invasiveness by quantifying their migration across a matrigel-coated transwell membrane in a chemotactic gradient. Reassuringly, despite their unexpected high motility, non-malignant PH5CH cells were only inefficiently able to cross the matrigel layer, showing the lowest quantifiable invasiveness in our assay (Fig. 3d, e). Furthermore, all four dedifferentiated (i.e. potentially highly malignant) cells lines, HLE, HLF, SNU387 and SNU182, were efficiently invading and transmigrating the membrane, underscoring their advanced cancerous stage (Fig. 3e). For SNU182, their comparably low migratory behaviour was reproduced in this assay, but their relative ability to traverse the matrigel layer indicates high invasiveness (compare Additional file 3: Fig. S3d, e). Furthermore, the high migration rate of HuH7 was confirmed and they exhibited a robust invasive capacity (see also Additional file 3: Fig. S3d, e). For the remaining epithelial-like cells, HuH6 and HuH1 did show invading cells, however migration was below our cut-off for quantification, and HepG2, Hep3B and PLC hardly had any successfully transmigrating cells.

Unrestricted cell growth is the defining feature of tumours, and clearly one of the prime selective advantages of individual cells during tumour evolution. We therefore determined proliferation rates of our cell lines as a third parameter in live-cell imaging. Interestingly, the well-differentiated epithelial-like cell lines exhibited moderate growth rates that were rather comparable to each other, with Hep3B marking the lower end (0.018 h−1) and HuH7 standing out at 0.025 h−1 (Fig. 3f). In contrast, proliferation rates differed vastly between the dedifferentiated fibroblast-like cell lines (Fig. 3f). Whereas SNU182 exhibited the slowest proliferation among all tested cell lines at 0.012 h−1, SNU387 grew comparably to the well-differentiated cluster by 0.021 h−1. Strikingly, HLE and HLF that already demonstrated the most pronounced migration and invasion, also proliferated at markedly higher rates by 0.033 h−1 and 0.028 h−1, respectively. Similar to migration, PH5CH also stood out with regard to proliferation, featuring the second-fastest proliferation rate among all tested cell lines at 0.091 h−1. Again, this is likely related to the expression of the large T-antigen.

In conclusion, despite malignity being a vaguely defined and multi-factorial feature, in our functional assays classical hallmarks of advanced (“aggressive”) tumour cells, such as cell migration, invasiveness and overall proliferation rate correlated well with the reported differentiation status of the cell lines (Table 2) [63]. Notable exceptions are HuH7, which proliferated and migrated at markedly higher rates than other epithelial-like cells, and SNU182, which–despite being classified as fibroblast-like–exhibited very slow proliferation and low motility. Furthermore, non-malignant, artificially immortalised PH5CH were among the most quickly proliferating cells and showed a high propensity to migrate. Hence, they cannot be regarded a proper reference for non-transformed hepatocytes in all respects.

Table 2 Summary of the cell lines’ functional malignant characteristicsComparison of IFN-sensitivity across cancer cell lines

According to our hypothesis, advanced tumours that developed in the context of an IFN-rich environment may be more resistant to growth inhibition by IFN. We, therefore, assessed the antiproliferative impact of increasing doses (0–8000 IU/ml) of IFN-β onto our panel of liver cancer cell lines. Indeed, IFN-β inhibited proliferation for all cell lines, however, to a substantially varying extent (Fig. 4a, Additional file 4: Fig. S4). While PH5CH showed the previously observed inhibition of roughly 50% at high doses, three of the well-differentiated hepatoma cell lines were affected even markedly stronger: HuH6, HuH7 and PLC, the latter of which exhibited an almost complete growth arrest at 8000 IU/ml IFN-β (Fig. 4a). On the contrary, the two fibroblast-like cells lines scoring highest in our malignity assays above (Table 2), HLE and HLF, proved profoundly resistant to IFN-β, with proliferation rates reduced by a mere 10–15% even at the highest dose (Fig. 4a). The different response profiles of PLC on one side and HLE and HLF on the other side of the non-malignant PH5CH is drastic (Fig. 4b). The effect of IFN-β on the remaining cell lines was more intermediate, with Hep3B showing least of an impact despite being a well-differentiated hepatoma cell line, and both SNU387 and SNU182 exhibiting a slightly stronger growth inhibition than PH5CH despite being fibroblast-like and, at least SNU387, higher malignity in our assays (Fig. 4a). Nonetheless, when plotting proliferation rates at 8000 IU/ml IFN-β over the respective capacity to migrate or invade, there appears to be an imperfect but statistically significant trend of more pronounced IFN-resistance in cells of higher malignity (p = 0.024, Pearson r = 0.70 for migration, Fig. 4c; p = 0.017, Pearson r = 0.94 for invasion, Additional file 3: Fig. S3f). This supports our hypothesis that IFN-resistance might be a feature more frequently found in advanced tumours that have undergone longer and/or stronger selection. Importantly, we do not propose that reduced sensitivity of cells towards IFN is causative or mechanistically linked to an increase in cell motility or invasiveness, but both might be consequences of more progressed tumour evolution. This is, for example, also supported by our observation that long-term selected PH5CHIFN−β did exhibit reduced IFN-sensitivity (Fig. 2b, c), but did not exhibit an increased migratory phenotype (Additional file 2: Fig. S2f, g).

Fig. 4

Antiproliferative effects of IFN-β on hepatoma cells. a Hepatoma cells were mock treated or stimulated with increasing doses of IFN-β (1–8000 IU/ml) and cell numbers were monitored for 96 h (imaged every 6 h) by live-cell imaging. Shown are growth rates normalised to untreated controls, for a medium (222 IU/ml) and the highest (8000 IU/ml) dose of IFN-β; see Additional file 4: Fig. S4 for all doses. b Full titration of HLE and HLF (dedifferentiated, high malignity), PLC (well differentiated, low malignity) and the control cell line PH5CH. c Relative growth rates at 8000 IU/ml IFN-β plotted against respective migration rates as determined before (Fig. 3b); plot against invasiveness see Additional file 3: Fig. S3f. Linear regression was performed and Pearson correlation was calculated. All graphs display mean (± SD in a and b) of three independent repetitions

In summary, we found largely differing degrees of sensitivity to the antiproliferative effects of IFN-β across the studied panel of liver cancer cells. The fact that all investigated cell lines are of hepatocytic origin indicated that the observed differences cannot be due to cell type-specificity. Rather, they likely reflect the large heterogeneity typical for tumour cells, from which favourable phenotypes are selected and becoming dominant during tumour evolution. In fact, we found indications that particularly advanced tumours may develop resistance against growth inhibiting effects of IFN.

Differences in IFN-signalling in hepatoma cells of high vs. low malignity

The differences in the antiproliferative effect of IFN-β exposure across the tested cell lines could reflect general differences in the cell lines’ antiviral response and IFN system. Interestingly, however, we observed no major differences in replication of vesicular stomatitis virus (VSV) and the induction of an antiviral response (expression of the ISG IFIT1), with no correlation to the observed proliferative phenotypes (Additional file 5: Fig. S5a, b). Nonetheless, as antiproliferative phenotypes were strikingly different, we moved on to more closely investigate the cells’ IFN signalling pathway (scheme, see Fig. 5a), particularly with regard to its dynamics. The first major step in signal transduction downstream of the IFN-receptor (IFNAR) complex is phosphorylation of STAT1 and STAT2. For the further experiments, we focussed on the most lowly (HLE, HLF) and highly sensitive cells (PLC), as well as the non-malignant control PH5CH exhibiting an intermediate phenotype (Fig. 4). We stimulated them with a moderate dose of IFN-β (130 IU/ml) and assessed STAT1 and STAT2 phosphorylation in a time course up to three hours (Fig. 5). We normalised the phospho-specific signal to the total level of the respective STAT; of note, the overall expression of both STATs, being ISGs themselves, did not change significantly over the course of the experiment. Interestingly, while the dedifferentiated cell lines HLE and HLF were generally responsive to IFN treatment (see also Additional file 5: Fig. S5a, b), phosphorylation of STAT1 and STAT2 was significantly higher in PLC and the PH5CH control (Fig. 5b, c). This striking difference in the degree of phosphorylation was not reflected in the kinetics of phosphorylation. In both HLE/HLF and PLC (as well as PH5CH) STAT phosphorylation occurred already at 15 min and reached peak levels at 60–90 min (Fig. 5d, e). For STAT2, phosphorylation in PLC appeared to be slightly faster than in the other cell lines (Fig. 5e). Taken together, despite general functionality of the IFN signalling cascade in both HLE/HLF and PLC, we observed a substantially lower degree of STAT phosphorylation upon IFN treatment of HLE and HLF cells, mirroring their significantly reduced antiproliferative response.

Fig. 5

STAT1/2 phosphorylation upon stimulation with 130 IU/ml IFN-β. a Representative immunoblots of HLE, HLF, PLC and the non-malignant PH5CH control. Samples were adjusted for total protein content; calnexin levels vary between cell lines. Illustration on the right shows an overview of the IFN signalling pathway indicating all molecules assessed in the figure. b + c Quantified levels of phosphorylated STAT (pSTAT) relative to corresponding mean total STAT. Area under curve (AUC) used to quantify response over full time course. d + e Same data as in b + c, pSTAT normalised to its peak values (100%) to allow for better comparison of kinetics. All graphs display mean ± SD of three independent repetitions. Statistical analysis was performed with one-way ANOVA followed by Tukey’s multiple comparison test

We next investigated the impact of reduced STAT phosphorylation on the induction kinetics of ISGs. Therefore, we measured the expression of three ISGs upon IFN-β treatment in a time course over 24 h: (1) IFIT1, one of the strongest induced type I ISGs; (2) USP18, a major negative feedback regulator of type I IFN signalling that has previously been implicated with antiproliferative effects; and (3) IRF1, a major ISG of the type II IFN response mediating growth inhibiting effects of IFNγ [78]. IRF1 is also induced downstream of the type I IFN receptor (IFNAR) by STAT1-STAT1 homodimers [79]. Importantly, it is a so called tuneable ISG, meaning that its induction is less switch-like but increases gradually with duration and intensity of the IFN signal [80], reminiscent of what has been described for the antiproliferative effects [24]. We furthermore included HuH1 in the experiments as a representative of well-differentiated malign (as opposed to non-malign PH5CH) cell lines with intermediate growth inhibition by IFN-β (Fig. 4a).

As already observed upon VSV infection (Additional file 5: Fig. S5b), all five tested cell lines showed comparable induction of IFIT1 and, strikingly, expression kinetics were virtually identical (Fig. 6a). Peak levels were only slightly lower in HLE and HLF (reaching significance only versus HuH1; Fig. 6a, right panel). In case of the negative regulator USP18, induction occurred earlier in HLE and HLF (Fig. 6b), resulting in slightly but reproducibly higher mRNA expression at 4 h (Fig. 6d). Peak induction, however, was comparable between all five cell lines (Fig. 6b, right panel). This faster USP18 induction might contribute to a quicker and stronger dampening of IFN responses in HLE and HLF, however it has been reported to hardly affect IFN-β signalling [81]. Lastly, for IRF1 we found a very steep and early induction with a peak already at 2 h post treatment for HLE and HLF, but also HuH1 and PH5CH, whereas PLC reached a peak only by 4 h (Fig. 6c, d). IRF1 expression was very transient and exhibited a rather rapid decay in HLE, HLF and PH5CH, while it was somewhat more sustained in HuH1 and particularly in PLC. Together with higher peak expression levels (Fig. 6c, right panel), this led to appreciably higher IRF1 expression at 6 h in HuH1 and PLC (Fig. 6d). The higher and particularly longer lasting expression of this transcription factor might contribute to the strong antiproliferative impact of IFN on PLC. Nonetheless, we were overall surprised to see rather minimal differences in the induction kinetics of these three different classes of ISGs. This suggested the strong difference in the cell lines’ proliferative response to IFN-β might in fact not be due to canonical antiviral transcriptional responses.

Fig. 6

Induction of ISG mRNA upon stimulation with 130 IU/ml IFN-β. a–c: Left panels: kinetics displayed as expression levels normalised to peak values. Gross outliers were excluded (IFIT1: HLE, PH5CH, PLC; IRF1: PH5CH; n = 2). Right panels: absolute expression levels determined at peak time points. d Relative expression levels at selected time points. e Type I IFN receptor mRNA expression in untreated cells. All graphs display mean ± SD of three independent repetitions (exceptions noted above). Absolute expression levels were determined independently of housekeeping genes as 230−Ct/4.2 ng of total RNA. Statistical analysis was performed with one-way ANOVA followed by Tukey’s multiple comparison test

As clear differences in terms of STAT1/2 activation were evident, we hypothesised the differential response of the cell lines might originate already at the receptor level. Hence, we analysed the mRNA expression levels of the two receptor subunits, IFNAR1 and IFNAR2 (Fig. 6e). For IFNAR1, we found a slight trend towards lower expression in HLE and HLF. For IFNAR2, this trend was substantially stronger, with PLC expressing approximately tenfold higher levels of IFNAR2 as compared to HLE. In analogy, expression was markedly higher in PH5CH and HuH1, both being much more susceptible to growth inhibition by IFN-β than HLE or HLF.

We could not clearly identify the underlying mechanism for the profoundly different antiproliferative response of the hepatoma cell lines upon IFN-β treatment. The most pronounced difference was observed for the mRNA expression levels of the IFN receptor chains, in particular IFNAR2. This translated into significant differences in STAT1/2 phosphorylation between HLE/HLF and the less malign PH5CH and PLC. Curiously, these differences were apparently overridden downstream in the signalling cascade, as induction kinetics and strength for three very different ISGs–IFIT1, USP18 and IRF1–were remarkably similar.

Differentially expressed factors in IFN-β-resistant hepatocytes

We could roughly classify the hepatoma cell lines used in this study according to their degree of malignity (Fig. 3). Nonetheless, they stem from completely different, i.e., genetically diverse individuals (exception: HLE and HLF), and their tumour evolution occurred under unknown physiological constraints and selection pressures. Hence, identifying concrete and common molecular mechanisms explaining resilience towards the antiproliferative action of IFN-β may be very challenging. We therefore made use of our in vitro selection assay, in which we subjected non-neoplastic PH5CH to continuous IFN-β stimulation (Fig. 2). As in this setting the genetic background of IFN-sensitive PH5CHmock and IFN-β-resistant PH5CHIFN−β is identical and we applied one single strong and defined selection pressure, expression differences may have a substantially higher likelihood to be causal for the observed phenotype. We therefore subjected these two populations of PH5CH to quantitative full proteome mass spectrometry. As expected, the two populations were overall highly similar (Fig. 7a). We specifically analysed central components of the IFN signalling pathway. Unfortunately, not all proteins were detected by this unbiased full proteome approach. Still, among the successfully detected proteins were key components such as the kinases JAK1 and TYK2 as well as transcription factors STAT1, STAT2 and IRF9 (not detected in all replicates). For none of them we found statistically significant differences in protein levels (Fig. 7b). Furthermore, putative negative regulators of IFN signalling, such as STAT3 [82], SOCS2, PIAS1 and PIAS4, were expressed to virtually identical levels (Fig. 7b). Nonetheless, global analysis of the proteomes did reveal statistically significant differences in expression of several proteins (Fig. 7c); none of them, however, is a canonical member of IFN signalling. We performed global pathway analyses on the differentially expressed proteins (Additional file 6: Fig. S6, Additional file 7), but future follow-up studies will be required to scrutinise the potential involvement of identified pathways in IFN signalling.