Colon cancer has been proposed to be sustained by a small subpopulation of stem-like cells with unique properties allowing them to survive conventional therapies and drive tumor recurrence. Identification of targetable signaling pathways contributing to malignant stem-like cell maintenance may therefore translate into new therapeutic strategies to overcome drug resistance. Here we demonstrated that MEK5/ERK5 signaling activation is associated with stem-like malignant phenotypes. Conversely, using a panel of cell line-derived three-dimensional models, we showed that ERK5 inhibition markedly suppresses the molecular and functional features of colon cancer stem-like cells. Particularly, pharmacological inhibition of ERK5 using XMD8-92 reduced the rate of primary and secondary sphere formation, the expression of pluripotency transcription factors SOX2, NANOG, and OCT4, and the proportion of tumor cells with increased ALDH activity. Notably, this was further associated with increased sensitivity to 5-fluorouracil-based chemotherapy. Mechanistically, ERK5 inhibition resulted in decreased IL-8 expression and NF-κB transcriptional activity, suggesting a possible ERK5/NF-κB/IL-8 signaling axis regulating stem-like cell malignancy. Taken together, our results provide proof of principle that ERK5-targeted inhibition may be a promising therapeutic approach to eliminate drug-resistant cancer stem-like cells and improve colon cancer treatment.
The identification of stem-like cells within tumors has reshaped our understanding of cancer development, introducing an additional layer of complexity to the concept of intratumoral heterogeneity1. The existence of cancer stem cells (CSCs) was demonstrated in several solid tumors, including colon cancer2,3,4. Importantly, CSC populations are characterized by their remarkable potential to perpetuate themselves through self-renewal, while retaining the ability to differentiate into the full repertoire of neoplastic cells forming the heterogeneous tumor mass5. Owing to their highly tumorigenic and adaptable phenotype, colon CSCs are currently recognized as the only subset of neoplastic cells holding attributes for tumor initiation, sustained growth, and metastasis formation6. Moreover, colon CSCs show increased resistance to conventional antitumor regimens7,8,9,10,11, arising as particularly well-suited feeders of tumor regrowth and relapse after initial response to chemotherapy6. Adding to the clinical implications of the CSC concept, expression of stemness-associated signatures is associated with worse clinical outcomes in colon cancer patients12,13,14. Elucidation of the molecular players regulating stem-like cell maintenance in colon cancer may therefore translate into new therapeutic strategies to overcome drug resistance and avoid tumor recurrence.
Malignant stem-like cells reproduce many of the signaling programs employed during embryonic development and tissue homeostasis15. The extracellular signal-regulated kinase 5 (ERK5 or BMK1) is a non-redundant member of the mitogen-activated protein kinase (MAPK) family that operates within an exclusive MAPK kinase 5 (MEK5)-ERK5 axis to control cell proliferation, survival, differentiation, and motility16. Targeted deletion of Mek5 and Erk5 in mice provided the first evidence for their essential role in development, leading to embryonic lethality at mid-gestation due to defective endothelial cell function and cardiovascular formation17,18,19,20. In addition, MEK5/ERK5 signaling has been implicated in the regulation of neurogenic21,22,23,24, myogenic25,26, and hematopoietic27,28,29 differentiation and lineage commitment. Mechanistically, ERK5 was proposed to act independently to maintain naive pluripotency and control cell fate decisions in mouse embryonic stem cells, suggesting multiple critical functions for this kinase during differentiation30.
In the intestine, activation of ERK5 is triggered as a bypass route to rescue epithelial cell turnover upon Erk1/2 ablation31; however, the physiological relevance of this cascade in the gastrointestinal tract remains to be elucidated32. On the other hand, substantial attention has been given to the link between aberrant MEK5/ERK5 signaling and the pathogenesis of colon cancer33,34,35,36. Dysregulation of both MEK5 and ERK5 in human tumor samples is associated with more aggressive and metastatic stages of the disease33,34,35, and poorer survival rates34,35,36. Moreover, evidence from different experimental models showed that ERK5-mediated signaling promotes tumor development, metastasis, and chemoresistance37, recapitulating the aforementioned features of colon CSCs6. However, thus far, no relationship has been established between colon cancer stem-like phenotypes and MEK5/ERK5 signaling.
In the present study, we show that MEK5/ERK5 signaling contributes to sustained stemness in colon cancer, at least in part, through the activation of a downstream NF-κB/IL-8 axis. More importantly, we provide evidence that pharmacological inhibition of ERK5 may be a promising therapeutic approach to eliminate malignant stem-like cells, avoid chemotherapy resistance, and improve colon cancer treatment.
MEK5/ERK5 signaling activation correlates with colon cancer stem-like cell phenotypes
Three-dimensional sphere models are widely used to selectively promote the growth of tumor cell populations with stem-like properties38,39, representing a functional system for the in vitro discovery of new signaling pathways regulating self-renewal and differentiation in CSCs. In the present study, we used a panel of established human colon cancer cell lines to generate sphere cultures. For this purpose, cells were grown in non-adherent conditions, using serum-free medium supplemented with growth factors. Under this experimental setting, only malignant cells with stem cell features are expected to survive and proliferate, giving rise to free-floating multicellular spheres, also known as tumorspheres38,39. After 1 week, HCT116, HT29, SW480, and SW620 cells were shown to efficiently form tumorspheres (Supplementary Figure S1a), which is in agreement with previous observations40,41,42. Additionally, the expression levels of genes involved in intestinal cell differentiation, including BMP4, CDX2, AQP3, and ADA, were significantly decreased in tumorsphere cultures, as compared with their adherent counterparts (p < 0.05) (Supplementary Figure S1b). On the other hand, the expression profile of the stemness-associated transcripts SOX2, NANOG, OCT4, and BMI1 was mostly enriched (p < 0.05), further confirming that sphere-forming populations were enriched for undifferentiated cells.
To determine whether MEK5/ERK5 signaling may be a relevant player in colon cancer stem-like cells, we first analyzed the activation status of these kinases in tumorsphere and matched adherent cultures. Immunoblot analysis showed that, except for HCT116-derived tumorspheres, colon cancer cells grown as spheres had significantly higher levels of MEK5 phosphorylation, compared with monolayer-cultured cells (p < 0.05) (Fig. 1a, upper panel). Further, ERK5 phosphorylation was shown to be consistently increased in tumorsphere cultures across all cell lines tested (p < 0.01) (Fig. 1a, lower panel), validating that MEK5/ERK5 signaling is overactivated in neoplastic populations enriched for stem-like cells. In turn, forced activation of ERK5 by ectopic expression of a constitutively active mutant of MEK5 (CA-MEK5) in SW480 adherent cultures (Fig. 1b) was associated with lower expression of genes involved in differentiation, and higher levels of stem cell markers, relative to empty vector control cells (p < 0.05) (Fig. 1c). Changes in NANOG, OCT4, and SOX2 were confirmed at the protein level (Fig. 1d). Together, these findings demonstrate that MEK5/ERK5 activation correlates with a shift toward an undifferentiated state in colon cancer cells, suggesting that colon cancer stem-like populations may be dependent on ERK5-mediated signaling.
ERK5 inhibition suppresses colon cancer stem-like cell properties
To address the functional role of MEK5/ERK5 signaling in colon cancer stem-like cells, HCT116, HT29, SW480, and SW620 cells were plated as tumorspheres, and grown in the presence of XMD8-92, a small-molecule inhibitor of ERK543 (Fig. 2a, b). Self-renewal was then measured according to second-generation sphere formation without any additional treatment. Consistent with our hypothesis, XMD8-92 significantly reduced the frequency of primary and secondary tumorsphere formation (p < 0.05) (Fig. 2c). This was further associated with the disruption of sphere morphology and size (p < 0.05) (Fig. 2b, d), with minimal effects on cell viability (Supplementary Figure S2), suggesting that besides self-renewal, ERK5 inhibition also impairs the proliferative potential of stem-like malignant cells. Worthy of note, sphere growth was conducted at clonal density (0.25-0.5 cells/μL) to avoid cell aggregation and sphere fusion. Single-cell assays confirmed the clonal origin of tumorspheres44, as well as the ability of XMD8-92 to inhibit self-renewal and the rate of sphere formation in both HCT116 and SW620 cells (p < 0.05) (Fig. 2e). Finally, to verify the contribution of ERK5 to tumorsphere formation, ERK5 expression was specifically silenced by RNA interference in non-adherent HCT116 cultures (Fig. 2f, g). Interestingly, knockdown of ERK5 led to a marked decrease in the number (p < 0.01) and size of HCT116-derived spheres (p < 0.001) (Fig. 2h), phenocopying the effects of XMD8-92 treatment. These results demonstrate that ERK5 inhibition depletes the population of sphere-initiating, self-renewing cells in colon cancer cultures.
To investigate the molecular basis underlying the differential frequencies of tumorsphere formation, the expression of core pluripotency transcription factors was next examined. ERK5 inhibition by XMD8-92 resulted in a significant downregulation of SOX2, OCT4, and NANOG in all cellular models under sphere-forming conditions, as assessed by quantitative reverse transcription polymerase chain reaction (RT-PCR) (p < 0.05) (Fig. 3a). These results were further confirmed by immunoblot analysis in HCT116-derived tumorspheres (Fig. 3b). Similarly, flow cytometry analysis of aldehyde dehydrogenase (ALDH) activity, a well-characterized marker of colon CSC subpopulations45, demonstrated a decrease in the proportion of ALDH-positive cells upon XMD8-92 treatment (p < 0.05) (Fig. 3c). Taken together, the aforementioned data demonstrate that ERK5 signaling inhibition suppresses malignant stem-like phenotypes and function, and support the notion that MEK5/ERK5 is required for sustained stemness in colon cancer cells.
ERK5 pharmacological inhibition sensitizes colon cancer stem-like cells to chemotherapy
Colon CSCs have been demonstrated to be highly resistant to standard-of-care chemotherapy7,8,11, and combination strategies leading to the suppression of these therapy refractory cells may ultimately translate into improved treatment efficacy and patient outcome. To evaluate the effect of ERK5 inhibition on cancer stem-like cell response to 5-fluorouracil (5-FU)-based chemotherapy, fully formed HCT116-derived tumorspheres were treated with FOLFOX (5-FU plus oxaliplatin) or FOLFIRI (5-FU plus irinotecan), alone or in combination with XMD8-92 (Fig. 4a). Remarkably, XMD8-92-treated tumorspheres showed enhanced sensitivity toward conventional FOLFOX and FOLFIRI treatment, as evidenced by an increase in cell death, compared to chemotherapy alone (p < 0.01) (Fig. 4b). Consistent with these observations, the combination of FOLFOX or FOLFIRI with XMD8-92 was further associated with increased caspase-3/7 activity (p < 0.01) (Fig. 4c), PARP cleavage (p < 0.05) (Fig. 4d, left panel), and XIAP degradation (p < 0.01) (Fig. 4d, right panel), demonstrating that ERK5 inhibition primes stem-like malignant populations to chemotherapy-induced apoptosis.
ERK5 inhibition suppresses interleukin-8 expression through an nuclear factor-kB-dependent mechanism
To identify mechanisms downstream of MEK5/ERK5 that might contribute to stem-like cell maintenance in colon cancer, we performed a comparative PCR array analysis of genes associated with CSC features in HCT116 tumorspheres treated with XMD8-92 or vehicle control. A total of 13 genes were found to be differentially expressed in response to ERK5 inhibition (log2-transformed fold change below −1 or above 1) (Fig. 5a, b). In line with the functional and biochemical characterization of tumorspheres, XMD8-92 treatment led to an upregulation of the differentiation factor GATA3, and a downregulation of the pluripotency factors KLF4 and MYC, and CSC markers PROM1/CD133 and PLAUR/CD87. This was further associated with a decrease in the expression of the ATP-binding cassette transporter ABCG2. On the other hand, inconsistent effects were found for proliferation and migration-related genes (KITLG, LIN28A, KLF17, and ZEB1). Array results were validated by independent quantitative RT-PCR of a selection of differentially expressed transcripts (Supplementary Figure S3).
Apart from the impact of ERK5 inhibition on CSC-associated markers, gene expression profiling also revealed NFKB1 and the nuclear factor-κB (NF-κB)-regulated CXCL8/IL-846 as being downregulated in XMD8-92-treated tumorspheres (Fig. 5a, b). Quantitative RT-PCR confirmed that treatment with XMD8-92 reduced IL-8 expression in HCT116, as well as SW480 and SW620 tumorspheres (p < 0.01) (Fig. 5c). Similar results were observed when genetically silencing ERK5 in HCT116 cells under sphere-forming conditions (p < 0.01) (Fig. 5d). These data suggest that elimination of tumor cell populations with stem-like traits through ERK5 inhibition might be a result of downstream IL-8 repression. Conversely, IL-8 mRNA levels were enriched in tumorsphere models where MEK5/ERK5 signaling was found to be induced (p < 0.05) (Fig. 5e), and in SW480 cells expressing CA-MEK5 (p < 0.05) (Fig. 5f), supporting the existence of a functional link between ERK5 activation and interleukin (IL)-8 signaling.
Inhibition of ERK5 has been previously shown to suppress IκB phosphorylation, preventing its degradation and subsequent NF-κB activation33. Here we investigated the relevance of the interplay between ERK5 and NF-κB signaling pathways in colon CSC. Consistent with previous observations in monolayer-cultured cells36, XMD8-92 treatment in HCT116-derived tumorspheres resulted in decreased IκB phosphorylation, and increased IκB protein levels (p < 0.01) (Fig. 6a). Moreover, using a luciferase reporter system, NF-κB transcriptional activity was found to be significantly impaired following XMD8-92 exposure (p < 0.05) (Fig. 6b), mirroring the repression of IL-8 upon EKR5 inhibition, and suggesting a possible autocrine ERK5/NF-κB/IL-8 axis driving stem-like cell malignancy. To investigate this further, NF-κB p65 and a dominant-negative-IκBα mutant (DN-IκBα) were respectively used to induce and block NF-κB activity in HCT116 cells (Fig. 6c). Overexpression of NF-κB p65 led to a marked upregulation of IL-8, compared to empty vector cells, an outcome that was largely reversed by the addition of XMD8-92 (p < 0.05) (Fig. 6d). Conversely, DN-IκBα reduced IL-8 mRNA levels and abolished the effect of XMD8-92 in the expression of this chemokine. Overall, these data demonstrate that NF-κB is involved in the regulation of IL-8 by ERK5, and provide a functional mechanism by which MEK5/ERK5 signaling contributes to the maintenance of stem-like properties in colon cancer.
In cancers such as those of the colon, tumor initiation and progression occurs through aberrant activation and/or mutation of the same molecular mechanisms that control normal stem cell dynamics6, within a network that goes beyond core pluripotency pathways, and is currently recognized to be controlled by multiple protein kinase cascades47. Exemplifying this phenomenon is the MEK5/ERK5 signaling pathway, which has been shown to participate in both development and tumorigenesis16. In this framework, we hypothesized that ERK5-mediated signaling could contribute to the maintenance of a stem-like population in colon cancer. Particularly, we demonstrate that MEK5/ERK5 activation is increased in several cell line-derived models enriched for malignant stem cells (Fig. 1); and that ERK5 inhibition using XMD8-92 suppresses both self-renewal and the expression of colon CSC-associated markers (Figs. 2 and 3). In line with our results, ERK5 has been previously identified as a critical player for sphere formation and tumor initiation in lung carcinoma cells48. Additionally, the suggested role of ERK5 in defining a CSC-like phenotype is consistent with the notion that activation of epithelial-to-mesenchymal transition (EMT) programs induces the acquisition of CSC traits that facilitate metastasis formation49. In this regard, we have previously demonstrated that MEK5/ERK5 activation is associated with upregulation of the mesenchymal marker vimentin, promoting colon cancer cell invasive and metastatic behavior in an orthotopic xenograft model33. Moreover, several other studies reported that MEK5 and ERK5 regulate EMT features, generation of circulating tumor cells, and metastatic seeding in different tumor contexts50,51,52. Still, further investigation is required to evaluate the possible influence of ERK5-mediated signaling to the molecular mechanisms underlying the connection between EMT and the CSC state.
Apart from the role of malignant stem-like cells in tumor initiation and metastasis, the CSC concept also provides a framework to understand therapy resistance6. Evidence from patient-derived three-dimensional cultures and xenograft models indicate that colon CSCs are intrinsically drug-resistant8,9,10,11. Moreover, the proportion of cells expressing CSC-associated markers was found enriched within residual tumors of colon cancer patients receiving chemoradiotherapy7. Therefore, identification of targetable signaling pathways controlling colon cancer stem-like phenotypes will undoubtedly fuel the development of combination regimens to overcome current therapy limitations. We have recently shown that ERK5 inhibition enhances the anticancer properties of 5-FU in a murine xenograft model of colon cancer36. Here we extend these earlier studies by revealing that XMD8-92 treatment sensitizes HCT116 cancer stem-like cells to 5-FU-based chemotherapy (Fig. 4), establishing a novel strategy to eliminate drug-resistant populations generated upon phenotypic switching into the CSC state. In parallel, we also found that ERK5 inhibition in tumorspheres leads to downregulation of ABCG2 (Fig. 5 and Supplementary Figure S3), a drug-efflux pump that is responsible for acquired resistance to both 5-FU53 and irinotecan54, also contributing to CSC malignant properties55. Indeed, and in agreement with our results, the ERK5/MEF2 pathway has been proposed to regulate the expression of several ABC transporters, among which ABCG256. It is therefore possible that part of the mechanism behind the increased susceptibility of stem-like colon cancer cells to chemotherapy upon ERK5 inhibition could involve ABCG2 repression.
Finally, our data demonstrate that NF-κB-mediated IL-8 expression might be a fundamental element of CSC-like function downstream of MEK5/ERK5 signaling (Figs. 5 and 6). In colon cancer, aberrant expression of the CXC chemokine IL-8, in tumor tissues or in circulation, was shown to be associated with poor differentiation, depth of invasion, and distant metastasis57,58,59. Functionally, IL-8 signaling promotes EMT, stem cell-like traits, and chemoresistance60,61,62. Regarding CSCs, the pro-inflammatory and angiogenic activity of IL-8 is known to be essential for the establishment of a supportive microenvironment for self-renewal and stem-like cell survival63. However, according to our experimental conditions, we suggest that IL-8 expression by tumor cells might also contribute to sustained stemness through autocrine signaling. Indeed, a similar feedback loop mechanism has already been proposed for the regulation of colon cancer cell proliferation and migration60,64. Strengthening our hypothesis, while NF-κB controls IL-8 expression46, this chemokine is in turn responsible for triggering NF-κB transcriptional activity65. Consistently, NF-κB signaling has also been linked to CSC-like features in several solid tumors, including glioblastoma, breast, prostate, and non-small cell lung cancer66,67,68,69. On the other hand, pharmacological inhibition and genetic knockdown of either MEK5 or ERK5 were reported to suppress lipopolysaccharide-, IL-1β-, and tumor necrosis factor-α-induced production of IL-8 in primary human endothelial cells and monocytes70. Similarly, we demonstrate that specifically silencing ERK5 recapitulates the effects of XMD8-92-mediated inhibition of ERK5 kinase activity, depleting the population of sphere-initiating cells (Fig. 2), and the expression of IL-8 in HCT116 tumorspheres (Fig. 5). Nevertheless, we cannot fully exclude that putative XMD8-92 off-target activity may partially account for the observed phenotypic effects of this small-molecule inhibitor against colon cancer stem-like cells71,72. Moreover, although cancer cell lines are representative of the different colorectal cancer molecular subtypes, which validates their utility as tools to investigate tumor biology and drug response73,74, future studies will be necessary to evaluate the impact of ERK5 inhibition on patient-derived in vitro and in vivo models.
Taken together, our findings provide proof of principle that pharmacological inhibition of ERK5 may be an effective strategy to target self-renewing, drug-resistant colon cancer stem-like cells. Adding to the clinical relevance of this signaling pathway, aberrant MEK5/ERK5 activation also contributes to increased tumor cell proliferation and metastasis33, inducing a chemoresistant phenotype in both CSCs and non-CSCs36. The plurality of mechanisms through which ERK5 activity drives the process of tumorigenesis reinforces the therapeutic potential of blocking this cascade in colon cancer treatment. Still, heterogeneous tumor masses comprising different populations of differentiated and cancer stem-like cells are expected to be sustained by different oncogenic mechanisms. Therefore, the introduction of ERK5-targeting agents in clinical evaluation should be envisioned as part of combination regimens designed to avoid resistance and tumor recurrence, bringing together conventional cytotoxic drugs and innovative targeted therapies.
Materials and methods
Human HCT116, HT29, SW480, and SW620 colorectal carcinoma cell lines were obtained from ECACC (Porton Down, UK), passaged for <6 months after resuscitation, and routinely tested for mycoplasma contamination using Mycoalert detection kit (Lonza, Basel, Switzerland). Cells were cultured in adherent conditions in McCoy’s 5A (HCT116), RPMI 1640 (HT29), or Dulbecco’s modified Eagle’s medium (DMEM) (SW480 and SW620), all supplemented with 10% heat-inactivated fetal bovine serum (FBS) and 1% antibiotic/antimycotic solution (all from Gibco, Thermo Fisher Scientific, Paisley, UK). For the generation of tumorspheres, cells were grown in non-adherent conditions in serum-free DMEM/F12 medium containing 2% B27 supplement, 1% N2 supplement, 1% non-essential amino acids, 1% sodium pyruvate, 1% penicillin-streptavidin, 4 μg/mL heparin, 40 ng/mL recombinant human epidermal growth factor (all from Gibco) and 20 ng/mL recombinant human basic fibroblast growth factor (Peprotech, London, UK). All cell cultures were maintained at 37 °C under a humidified atmosphere of 5% CO2.
Small molecules and chemotherapeutic agents
The ERK5 pharmacological inhibitor XMD8-92 was obtained from Selleckchem (Madrid, Spain) and prepared in dimethyl sulfoxide (DMSO; Sigma-Aldrich, MO, USA). Clinical-grade 5-FU, oxaliplatin, and irinotecan were kindly provided by Hospital São Francisco Xavier (Lisbon, Portugal), and diluted to stock concentrations in phosphate-buffered saline (Gibco, Thermo Fisher Scientific). Stock solutions were aliquoted and stored at −80 and −20 °C, respectively. All subsequent dilutions were freshly prepared in culture medium. Experiments were performed in parallel with DMSO vehicle control. Final DMSO concentration was always 0.1%.
For overexpression experiments, CA-MEK5 plasmid (pWPI-MEK5DD; S313D/T317D) was kindly provided by Dr. Robert C. Doebele (University of Colorado, CO, USA)75. Constructs for NF-κB p65 (pCMV4 p65)76 and DN-IκBα (pCMX IkB alpha M; S32A/S36A)77 were obtained from Addgene (#21966 and #12329, respectively). For small interfering RNA (siRNA)-mediated knockdown of ERK5, the MAPK7 Silencer Select was used (#s11149; Applied Biosystems, Thermo Fisher Scientific). HCT116 and SW480 cells were plated at 3 × 105 cells/well on 35 mm dishes and transfected with either 1 μg of plasmid DNA or 80 nM of siRNA using Lipofectamine 3000 (Invitrogen, Thermo Fisher Scientific), according to the manufacturer’s instructions. In both cases, cells were allowed to grow for at least 24 h before further treatment or re-plating for tumorsphere formation.
To measure tumorsphere formation, colon cancer cells were plated as single cells in 24-well ultra-low attachment plates (Corning, NY, USA) at 250–500 cells/well, and cultured in 1 mL serum-free DMEM/F12 supplemented with growth factors. After 8 days, spheres were collected, dissociated into single cells, and reseeded as above for secondary sphere formation. For each generation, the number of tumorspheres was determined under an inverted microscope. Additionally, the number of cells per tumorsphere was quantified using trypan blue exclusion assay. Alternatively, cells were sorted at a density of 1 cell/well into 96-well ultra-low attachment plates (BD FACS Aria III, BD Biosciences, CA, USA), and allowed to grow in 200 μL tumorsphere medium. The wells without cells were excluded from analysis one day after plating, and a minimum of 90 wells per condition was considered. Sphere-forming efficiency was calculated after 14 days according to the proportion of wells with tumorspheres versus initially seeded wells. In all cases, cells were allowed to adapt for 24 h and then treated with 4 μM XMD8-92 or DMSO vehicle control, except for second-generation spheres, which were grown without further treatment.
Cells with high ALDH enzymatic activity were identified using the Aldefluor assay (StemCell Technologies, Grenoble, France) according to the manufacturer’s protocol. In brief, 5,000–10,000 single cells were seeded in 5 mL tumorsphere medium using non-tissue culture-treated 55 mm dishes (Gosselin, Hazebrouck, France), cultured for 24 h, and then treated with either 4 μM XMD8-92 or DMSO vehicle control. Eight-day tumorspheres were collected, dissociated into single cells, resuspended in assay buffer containing 1.5 μM BODIPY-aminoacetaldehyde, and incubated for 40 min at 37 °C. Diethylaminobenzaldehyde (15 μM), a specific ALDH inhibitor, was used as a negative control for each reaction. Samples were then centrifuged, resuspended in fresh assay buffer, and stored on ice until flow cytometric analysis. Sample acquisition was performed in a BD LSRFortessa Cell Analyzer cytometer (BD Biosciences). A total of 10,000 cells were analyzed for each test and control sample pair, and the percentage of ALDHhigh cells was determined using FlowJo software (version 10.0.7; Tree Star, CA, USA).
Cell death and caspase activity assays
HCT116 cells were plated in 24-well ultra-low attachment plates and stem cell medium at a density of 500 cells/well. Resulting 8-day tumorsphere cultures were treated for 3 days with FOLFOX (1.25 μM oxaliplatin plus 50 μM 5-FU), or FOLFIRI (1 μM irinotecan plus 50 μM 5-FU) chemotherapeutic regimens78, alone or in combination with 4 μM XMD8-92. The in vitro cytotoxic effect of chemotherapy was evaluated using ToxiLight bioassay kit (Lonza) to measure the amount of adenylate kinase (AK) released from plasma membrane-damaged cells into tumorsphere supernatants, following the manufacturer’s instructions. Further, the activity of effector caspases-3 and -7 was measured using Caspase-Glo 3/7 Assay (Promega, WI, USA). For this purpose, tumorspheres were collected at 300 × g for 7 min, dissociated into single cells, and resuspended in fresh growth medium. Cell suspensions were then mixed with an equal volume of Caspase-Glo 3/7 reagent, and incubated for 30 min at room temperature, protected from light. Resulting luminescence was measured using the GloMax-Multi+ Detection System (Promega). Experimental AK release and caspase-3/-7 activity levels were normalized by the number of cells per well.
For gene expression analysis, tumorspheres were grown and treated as described for Aldefluor activity evaluation. Additionally, comparative studies were conducted by culturing colon cancer cells as monolayers in traditional medium supplemented with FBS. Total RNA was extracted using Ribozol (VWR International, PA, USA), treated with RNase-free recombinant DNase I (Roche, Mannheim, Germany), and reverse-transcribed to complementary DNA using the NZY First-Strand cDNA Synthesis Kit (NZYTech, Lisbon, Portugal), all according to the manufacturers’ instructions. Quantitative real-time PCR was performed in 5 μL duplicate reactions on a 384-well QuantStudio 7 Flex Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific), using the SensiFAST SYBR Hi-ROX kit (Bioline, London, UK), following manufacturer’s protocol. Primer sequences are listed in Supplementary Table S1. For each sample, quantification of gene expression was performed using the relative standard curve method and normalized to ACTB levels.
Total protein isolation and immunoblotting
Total protein extraction and immunoblot analysis were performed as previously described36. Briefly, 40 μg of total protein extracts were denatured, separated on 8 or 10% sodium dodecyl sulfate polyacrylamide electrophoresis gels, and transferred onto nitrocellulose membranes. Steady-state protein levels were evaluated using primary rabbit antibodies reactive to ERK5 (#3372), OCT4 (#2750; Cell Signaling Technology, MA, USA), SOX2 (#AB5603, Merck Millipore, MA, USA), p-MEK5 (#sc-135702), PARP (#sc-7150), NF-κB (#sc-372), IκBα (#sc-371), or XIAP (#sc-11426; Santa Cruz Biotechnology, CA, USA); or primary mouse antibodies against MEK5 (#sc-135986), NANOG (#sc-134218, Santa Cruz Biotechnology), or p-IκBα (#9246; Cell Signaling Technology). β-actin (#A5541; Sigma-Aldrich) and GAPDH (#sc-32233) were used as loading controls. Following incubation with appropriate horseradish peroxidase-conjugated secondary antibodies (Bio-Rad Laboratories, CA, USA), the proteins of interest were detected by chemiluminescence using SuperSignal reagents (Pierce, Thermo Fisher Scientific), on a ChemiDoc XRS+ imaging system (Bio-Rad). Densitometric analysis was performed using the Image Lab software (version 5.1; Bio-Rad).
Gene expression profiling
Differential gene expression between DMSO- and XMD8-92-treated tumorspheres was evaluated using the Human Cancer Stem Cells RT2 Profiler PCR Array (PAHS-176Z; Qiagen, MD, USA) according to the manufacturer’s instructions. For each condition, pools were obtained by combining equal amounts of total RNA from five different experiments. Complementary DNA synthesis was performed with 800 ng of DNase I-treated RNA (Roche) and the RT2 First Strand Kit (Qiagen). Real-time PCR was run on a 384-well QuantStudio 7 Flex Real-Time PCR System (Thermo Fisher Scientific), using the RT2 SYBR Green ROX qPCR master mix (Qiagen). Duplicate reactions for all genes, as well as quality controls for genomic DNA contamination, reverse transcription efficiency, and PCR array reproducibility were included. Data analysis was performed using the GeneGlobe online platform (https://www.qiagen.com/geneglobe/). Relative gene expression over control samples was determined as per the comparative cycle threshold (ΔΔCt) method and normalized to the geometric mean of B2M and HPRT reference genes (ΔCt = Ctreference − Cttarget; ΔΔCt = ΔCtXMD8-92 − ΔCtDMSO). A cutoff value of log2-fold change (ΔΔCt) ≥ 1 was defined for the selection of differentially expressed transcripts. Genes with Ct values above 34 or standard deviations between technical replicates superior to 0.5 were excluded from analysis. Results for each detectable gene are shown in Supplementary Table S2.
NF-κB luciferase reporter assay
NF-κB transcriptional activity was measured using the Cignal NF-κB Pathway Reporter Assay Kit (Qiagen), according to the manufacturer’s specifications. Briefly, HCT116 cells were seeded at 2 × 104 cells/well on 96-well plates and transfected with 100 ng of luciferase construct harboring NF-κB response elements using Lipofectamine 3000 (Invitrogen, Thermo Fisher Scientific). Non-inducible and constitutively expressed firefly luciferase constructs were used as negative and positive controls, respectively. A constitutive Renilla luciferase vector was included in all mixes (40:1) to normalize transfection efficiency and monitor cell viability. Sixteen hours post transfection, cells were treated with 4 μM XMD8-92 or DMSO vehicle control. Luciferase activities were assayed 8 h after treatment using the Dual-Luciferase Reporter Assay System (Promega).
All data are expressed as mean ± standard error of the mean from at least three independent experiments. Statistical significances were determined using unpaired two-tailed Student’s t-test. Values of p < 0.05 were considered statistically significant.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Edited by I. Lavrik
The authors thank Dr. Robert C. Doebele (University of Colorado, CO, USA) for the kind gift of the pWPI-GFP expression construct encoding constitutively active MEK5. We also wish to thank Hospital São Francisco Xavier (Lisbon, Portugal) for providing clinical grade 5-fluorouracil, oxaliplatin, and irinotecan. This work was supported by Fundação para a Ciência e a Tecnologia (FCT) through fellowships SFRH/BD/88619/2012 (S.E.G.) and SFRH/BD/96517/2013 (D.M.P.). This project received funding from European Structural & Investment Funds through the COMPETE Programme and from National Funds through FCT under the Programme grant SAICTPAC/0019/2015.