Macrophage migration inhibitory factor engages PI3K/Akt signalling and is a prognostic factor in metastatic melanoma

Human melanoma cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Lonza) supplemented with 5% foetal bovine serum (FBS; Sigma-Aldrich) at 37°C in a humidified atmosphere of 5% CO₂. Me1007 and MM200 were established from primary melanomas [28], MelCV, MelRMu and MelFH were from lymph node metastases [29] and MelRM was derived from a bowel metastasis [29]. Melanoma cell lines with the prefix Mel were isolated from fresh surgical biopsies from patients attending the Sydney and Newcastle Melanoma Units. Where indicated, cell number and viability were estimated using an ADAM-MC Automatic Cell Counter (Digital Bio). The assay employs the propidium iodide (PI) method comparing suspensions of PI-stained intact cells (measuring non-viable cells) against PI-stained permeabilised cells (measuring total cells). Cell suspensions were measured in triplicate for each time point.

Cells were lysed using NDE lysis Buffer (1% Nonidet P-40, 0.4% sodium deoxycholate, 66 mM EDTA, 10 mM Tris–HCl, pH 7.4) supplemented with protease and phosphatase inhibitors (Complete protease inhibitor mixture and PhosSTOP, respectively; Roche Applied Science). Protein concentrations were quantitated using BCA assay (Pierce) before electrophoresis on SDS-PAGE gels. Western blotting detection using ECL was performed as previously described [30] with bands visualized using a cooled charge-coupled device camera system (Fuji-LAS-4000, Fujifilm Life Science Systems). Primary antibodies used were: MIF (MAB289; R&D Systems); pAKT (pSer473) and total AKT (9271 and 9272, respectively; Cell Signaling Technologies); cyclin D1 and CDK4 (sc-20044 and sc-23896, respectively; Santa Cruz Biotechnology); p27 (610242; BD Transduction Laboratories™) and GAPDH (sc-25778; Santa Cruz). Secondary antibodies used were horseradish peroxidise (HRP)-conjugated anti-mouse or anti-rabbit (1706516 and 1706515 respectively; BioRad Laboratories).

Cells were seeded into 6-well plates at 10⁵ cells per well and allowed to reach 30-40% confluency before transfection. Synthetic siRNA duplexes were purchased from Shanghai GenePharma (PRC). Targeting sequences and validation experiments are shown in Additional file 1: Figure S1. Cells were transfected at indicated concentrations with siRNA duplexes using Lipofectamine™ RNAiMAX (Invitrogen, #13778) according to manufacturer’s instructions. Efficiency of gene knockdown was assessed by Western blotting.

DNA content analyses including quantitation of apoptotic (sub-G1) cells were performed using the propidium iodide (PI) staining method as described elsewhere [29]. The Click-iT™ EdU flow cytometry assay (Invitrogen, #C35002) was also used to determine the percentage of cells in S-phase. Briefly, three days after transfection with siRNA, cells were pulsed with 5-ethynyl-2′-deoxyuridine (EdU; 10 μM for 3 hours) before processing the cells according to manufacturer’s instructions. Receptor expression studies were performed using indirect immunostaining as previously described [31]. All flow cytometry was performed using a FACS Calibur II instrument with analyses conducted with either the Cell Quest software package v4 (Becton-Dickinson) or FlowJo v10.

The ability of cells to grow under anchorage-independent conditions was measured by a soft agar colony formation assay. Briefly, 6 well plates were under-coated with 1 mL of 0.6% low melting point agar (MetaPhor®) in DMEM. Cells were harvested and 1×10⁴ cells resuspended in 1 mL of 0.3% agar/DMEM/10% FBS and the cell suspension was poured on the bottom agar layer. Plates were then incubated for 3–4 weeks before staining with 0.005% (w/v) crystal violet to visualise colonies. Bright field photomicrographs from random fields were collected using an Axiocam MRm camera fitted to Axiovert 200 inverted microscope (Zeiss) and these used to count colony frequencies. The size of colonies was estimated using the Axiovision software package (v4.8.1, Zeiss).

Publically available microarray gene expression data sets were sourced from the NCBI gene expression omnibus (GEO) database (http://​www.​ncbi.​nlm.​nih.​gov/​geo/​) and normalized data used to determine the relative levels of genes of interest using methods previously described [32]. Where indicated, MIF expression was correlated with patient outcome whereby the primary end points for survival analyses were disease-specific survival (measured from the date of diagnosis to disease-specific death, or otherwise censored at the time of the last follow-up or at non-disease-related death). Time to disease-specific death was plotted using Kaplan-Meier survival curves.

We designed four individual siRNA oligonucleotide duplexes targeting MIF and determined the ability of each to down-regulate cellular protein levels of MIF. Western blotting analysis identified two sequences (MIF-21 and MIF-25) that were effective in reducing MIF levels in melanoma cell lines (Additional file 1: Figure S1A and B). During the course of this work, another study targeting MIF in lung cancer cells also demonstrated efficient knockdown of MIF using duplexes identical to the MIF-25 targeting sequence [33].

To determine the effects of MIF knockdown on melanoma cell growth, cell number and viability were measured each day over a 5 day period. Comparison of MelCV and Me1007 melanoma cells transfected with MIF-25 siRNAs confirmed a substantial reduction in the total MIF protein measured in cell lysates relative to negative control (NC) siRNA treatment (Figure 1A and B, respectively). For both cell lines, the total number of cells began to decrease after 3 days of MIF knockdown (Figure 1C and D) and this was accompanied by significant reductions in cell viability (Figure 1E and F). Both active siRNA duplexes (MIF-21 and MIF-25) promoted equivalent biological responses (Additional file 1: Figure S1C and data not shown) indicating that depletion of endogenous MIF can significantly compromise the proliferative capacity and viability of melanoma cells in culture.

To address the mechanism whereby MIF depletion was associated with reduced cell growth and viability, DNA contents were measured by flow cytometry. Representative profiles of MelCV and Me1007 cells are shown after treatment with control NC siRNA or siRNA against MIF (Figure 2A and B). As shown, MIF depletion resulted in an increased number of cells in the G0/1-phase in both cell lines (i.e. from 57% to 87% and 61% to 71% for MelCV and Me1007 cells, respectively). For MelCV cells there was a reduction in the number of cells recorded in the S- and G2/M phases after MIF depletion (Figure 2A) while in Me1007 cells the major effect appeared to be a reduction in the percentage of cells in S-phase (Figure 2B).In addition to the changes in cell cycle parameters, we also assessed whether cells were undergoing increased rates of apoptosis as suggested by the decreased viability observed in Figure 1. An estimate of apoptosis was determined from the DNA content analysis as the number of cells appearing in the sub-G0/1 region, i.e. cells with DNA content of less than 2n. This analysis showed that the basal level of apoptosis in control cultures increased ~2-3 fold when the MelCV or Me1007 were treated with siRNA against MIF for 3 days (Figure 2C and D). Taken together with the results of Figure 1, these data suggest that reduced cell growth occurred as a result of cells accumulating in the G0/1 phase and that the progressive decline in cell viability was caused in part by increased rates of apoptosis.To better define the effects of MIF knock down in melanoma cell lines, particularly their decreased proliferative capacity, the ability of cells to enter the S-phase of the cell cycle was measured using the Click-iT™ EdU flow cytometry assay. Click-iT analysis of MelCV and Me1007 cells treated with MIF siRNA showed a clear reduction in cells entering S-phase (Figure 3A and B). The results from 5 independent experiments show that inhibition of MIF expression significantly reduces the percentage of cells in S phase compared to negative control siRNA transfection for both the MelCV and Me1007 melanoma cell lines (Figure 3C and D, respectively). MIF depletion also significantly reduced the number of cells entering S-phase in four of six melanoma cell lines examined (Figure 3E) suggesting the proliferative capacity of the majority of melanomas have some degree of reliance on MIF expression.

In addition to altered cell proliferation, the ability of cells to undergo anchorage-independent cell growth is another hallmark of cancer. Loss of MIF expression in MelCV and Me1007 melanoma cell lines resulted in significantly less colonies in both cell lines compared to controls (Figure 4A–D). Moreover, the colonies formed after MIF knockdown were also significantly smaller than controls (Figure 4E and F). Taken together, these results provide further evidence that MIF expression regulates both cell cycle entry and the clonogeneic capacity of melanoma cells in vitro.

As part of our studies we also sought to determine whether the individual responses of cell lines to MIF could be explained by expression of the known cellular receptors for MIF that comprise CD74 and its co-receptor CD44, along with the chemokine receptors CXCR2 and CXCR4. Analysis by both Western blotting and flow cytometry showed that all six melanoma lines expressed both CD44 and CXCR4 while none expressed CXCR2 (Additional file 1: Figures S2 and S3). All lines varied in expression of CD74 but there was no clear correlation between expression levels and sensitivity of individual cell lines to MIF depletion. Interestingly, there was also no correlation between the V600E BRAF status of each line and sensitivity to MIF depletion since the V600E BRAF positive cell lines MelCV and MelRMu were amongst the most sensitive to MIF depletion (refer Discussion).

Having established that MIF exerts effects on the cell cycle entry and the clonogenic capacity of melanoma cells, we sought to determine which pathways are activated downstream of MIF. Signalling through the PI3K/Akt pathway is well established to play an important role in melanoma progression [34‐36] but to date no link has been made between MIF expression and regulation of the PI3K/Akt pathway in this setting. Therefore, we examined the effects of MIF knockdown on the expression of key Akt-signalling components in melanoma cells. With respect to the proliferative capacity of melanoma cells, it was observed from Figure 3 that four melanoma cell lines tested were sensitive to MIF depletion (MelCV, Me1007, MelRMu and MelFH) and another two (MelRM and MM200) were comparatively resistant. All six melanoma cell lines were subjected to treatment with MIF siRNA with knockdown of MIF protein after three days of transfection confirmed relative to controls using Western blotting (Figure 5A). Analysis of Akt phosphorylation status in cell lysates indicated a strong reduction in MelCV, Me1007 and MelRMu cells (~40-70% of controls) as a consequence of MIF knockdown with a lesser reduction observed in MelFH, MM200 and MelRM cell lines (~20% of controls; Figure 5A).We then sought to determine whether there was a direct correlation between the relative effects of MIF knockdown on cell proliferation (inhibition of S phase; Figure 3E) and the relative levels of Akt activation for each cell line. There was a demonstrable positive correlation where cell lines most sensitive to MIF depletion also had the greatest change in Akt activity and vice versa (Figure 5B). Further analysis of the downstream cell cycle modulators known to be influenced by Akt signalling was also undertaken. CDK4 and cyclin D1 involved in G1/S transition also showed some level of inhibition across the six cell lines, whereas the expression of cyclin-dependent kinase inhibitor, p27, was relatively increased in most of the cell lines following MIF depletion. On balance these results support the notion that Akt-signalling is down-regulated in response to MIF knockdown with the degree of sensitivity to MIF depletion commensurate with the inhibitory effects observed on the Akt pathway.

After establishing that MIF expression is important for the maintenance of melanoma cells in vitro, we investigated whether MIF expression levels were also elevated and/or associated with clinical outcomes in melanoma. Firstly, we independently performed in silico analyses of expression microarray data comparing the relative transcript levels of MIF in staged melanoma against normal skin and naevi (samples of normal skin, benign naevi, atypical naevi, melanoma in situ, vertical growth phase (VGP) and metastatic growth phase (MGP) melanoma, melanoma-positive lymph nodes (LN); deposited as GEO dataset GSE4587 [37]). The expression levels of MIF were determined as normalized intensity values (GeneSpring 7.1, Silicon Genetics) for each sample (Figure 6A). The level and pattern of MIF expression show a general increase in MIF levels associated with disease progression. Dividing the samples into two groups, “early stage” (normal skin, benign and atypical naevi, melanoma in situ) and “advanced stage” (VGP and MGP melanomas, LN) demonstrated a statistically significant increase in MIF expression in “advanced-stage” tissue samples compared to the “early-stage” group (Figure 6B). To substantiate this finding, further analyses were conducted on the data set generated by Xu et al.[38] consisting of eighty-three fresh biopsies from melanoma patients (profiled using the Affymetrix U133A microarray platform; GEO accession number GSE8401). The distribution of MIF expression for primary melanoma (n = 31) and metastatic melanoma (n = 52) are shown in Figure 6C and D, respectively, where MIF mRNA levels appear relatively increased in metastatic samples. Analysis of average levels in each group showed a statistically higher level of MIF in metastatic melanoma compared to primary tumour samples (Figure 6E). Collectively these findings support the notion that MIF expression is up-regulated during melanoma progression.

We then sought to establish whether levels of MIF expression in melanoma were predictive of outcome by exploiting expression microarray data associated with clinical outcomes. Since the levels of MIF differed between primary and metastatic melanoma, analyses were conducted on each classification where tumour biopsies were initially segregated into quartiles of MIF expression ranging from low to high values. Kaplan-Meir plots of MIF levels against survival appeared to show no predictive significance of MIF levels in primary melanoma tumours (Figure 7A) whereas a clear trend was evident for metastatic samples where the first and second quartiles segregated from the third and fourth quartiles (Figure 7B). Further analysis of the data using a 50% cut-off showed that high levels of MIF in metastatic disease conferred significantly poorer outcome compared to those tumours expressing lower levels of MIF mRNA (univariate Cox regression; hazard ratio = 2.946; 95% confidence interval 1.440-6.029; p = 0.003; Figure 7C). The same analyses conducted on the primary melanoma data showed no significant relationship between MIF expression in and survival (hazard ratio = 1.091; 95% confidence interval 0.312-3.809; p = 0.892). As further validation, the same analysis of an independent dataset of metastatic melanoma tissues (GSE19234; [39]) also indicated significantly worse outcomes for patients whose tumours expressed higher levels of MIF (univariate Cox regression; hazard ratio = 4.600; 95% confidence interval 1.6-12.9; p = 0.004; Figure 7D). Thus in patients where metastasis had already occurred, those cases with tumours displaying the highest levels of MIF progressed faster.