Krüppel-like factor 4 blocks tumor cell proliferation and promotes drug resistance in multiple myeloma

Abstract

Krüppel-like factor 4 is a transcription factor with anti-proliferative effects in differentiated cells, but with the ability to reprogram adult cells into cell-cycling pluripotent cells. In cancer, Krüppel-like factor 4 acts as either an anti-oncogene or an oncogene. We analyzed Krüppel-like factor 4 gene expression in multiple myeloma using Affymetrix microarrays. We generated conditionally expressing Krüppel-like factor 4 myeloma cell lines to investigate the function of this gene in myeloma biology. Krüppel-like factor 4 gene expression is high in normal plasma cells, but reduced in primary multiple myeloma cells from two-thirds of patients. It is not expressed by any human myeloma cell line due to promoter methylation. Conditional expression of Krüppel-like factor 4 led to complete cell cycle blockade, mainly in G1 phase, with no major apoptosis. This blockade was associated with induction of p21^Cip1 and p27^Kip1 in cell lines with an intact p53 pathway, and of p27^Kip1 only in those with an impaired p53 pathway. Krüppel-like factor 4 is highly expressed in the poor prognostic MS group with t(4;14) translocation and in the good prognostic CD-1 group with t(11;14) or t(6;14). The apparent contradiction of cell cycle inhibitor Krüppel-like factor 4 expression in patients with poor prognosis could be reconciled since its expression increased the resistance of myeloma cell lines to melphalan. In conclusion, we describe for the first time that Krüppel-like factor 4 could play a critical role in controlling the cell cycle and resistance to alkylating agents in multiple myeloma cells.

Introduction

Krüppel-like factor 4 (KLF4) is a bi-functional transcription factor belonging to the family of Krüppel-like factors. It can both activate or repress genes, depending on its target.¹ KLF4 is expressed in various differentiated cells including intestinal and skin epithelial cells,² monocytes/macrophages and B lymphocytes.³ However, KLF4 is also a so-called stem cell protein. It is one of Yamanaka’s four proteins (OCT4, SOX2, KLF4 and MYC) able to reprogram adult cells into induced pluripotent stem cells.^4,5 KLF4 interacts with OCT4 and SOX2 to bind to the NANOG promoter and confers stem cell pluripotency.⁶ KLF4 can function both as an oncogene and as a tumor suppressor.⁷ The tumor suppressor role of KLF4 is explained in part by co-transcriptional activation of the CDKN1A gene coding for the p21^Cip1 cell cycle inhibitor in collaboration with p53.⁸ KLF4 binds the CDKN1A promoter in a different region than p53 and potentiates the transcriptional activity of p53.⁹ KLF4 also induces transcription of the CDKN1B gene coding for p27^Kip1 cell cycle inhibitor.¹⁰ The KLF4 tumor-promoting role is context-dependent and is observed for instance in case of a ras mutation or cyclin D overexpression that may bypass KLF4-induced growth arrest.⁸ In addition, KLF4 represses the p53-induced expression of the gene coding for pro-apoptotic BAX protein.¹¹

KLF4 is expressed in naïve and memory B lymphocytes.¹² It could be involved in B-cell quiescence since its expression is lost upon B-cell activation and forced KLF4 expression blocks B-cell proliferation.¹³ However, its role in controlling B-cell activation is likely more complex, since KLF4 knockout hampers B-cell proliferation in KLF4^−/− mice.¹⁴ This could be explained by the binding of KLF4 to the CCND2 promoter and induction of CCND2 gene expression.¹⁴ The KLF4 gene is epigenetically silenced in human follicular lymphoma, diffuse large B-cell lymphoma and Hodgkin’s lymphomas.¹² Forced induction of KLF4 results in growth delay in Burkitt’s lymphoma cell lines, but a dramatic apoptosis in Hodgkin’s lymphoma cell lines, through up-regulation of apoptotic proteins.¹²

No data are available about an involvement of KLF4 in human multiple myeloma, a neoplasia characterized by the accumulation of a clone of malignant plasma cells primarily in the bone marrow. Numerous genetic abnormalities are present in this cancer and high throughput DNA microarrays have made it possible to classify newly diagnosed patients into eight molecular groups based on gene expression of their primary multiple myeloma cells (MMC).¹⁵ One of these eight groups (MS group) encompasses the 15% of newly-diagnosed patients with the poor prognostic t(4;14)(p16.3;q32) translocation resulting in over-expression of the MMSET gene in all cases and the FGFR3 gene in 70% of cases.¹⁵ Two other molecular groups (CD1 and CD2) comprise patients with aberrant CCND1 expression due to the t(11;14)(q13;q32) translocation or chromosome 11 amplification or CCND3 overexpression due to t(6;14)(p21;q32) translocation. Of note, the KLF4 gene is one of the few genes differentially expressed between the CD1 and CD2 groups.¹⁵

We show here that the KLF4 gene is expressed in normal bone marrow plasma cells, is highly expressed in primary MMC of 26% of the patients - mainly in patients with t(4;14)(p16.3;q32) translocation and in the CD1 group - and is not expressed in 20 human myeloma cell lines (HMCL). KLF4 expression was found to be regulated by its promoter methylation state. A forced induction of KLF4 triggered a stop in cell line growth, with a blockage of cells in the G1 and G2/M phases of the cell cycle, due to p27^Kip1 induction in all HMCL and p21^Cip1 induction in TP53^+/+ HMCL. In addition, forced KLF4 induction protected MMC from drug-induced apoptosis.

Methods

Ethics committee approval

Samples were collected after patients’ written informed consent in accordance with the Declaration of Helsinki and institutional research board approval from the hospitals of Montpellier and Heidelberg University.

Patients’ samples and cell lines

The XG1, XG2, XG3, XG4, XG5, XG6, XG7, XG10, XG11, XG12, XG13, XG16, XG19, and XG20 HMCL were obtained as described elsewhere.^16–18 The SKMM, OPM2, LP1 and RPMI8226 HMCL were purchased from ATTC (LGC Promochem, France). These HMCL were recently fully characterized from molecular and phenotypic points of view.¹⁹ The MMC of 206 patients with previously untreated multiple myeloma were included in this study. These 206 patients were treated with high-dose therapy and autologous stem cell transplantation and this cohort is termed in the following the Heidelberg-Montpellier cohort. The patients’ characteristics are presented in Online Supplementary Table S1. For 12 patients, MMC were harvested both at diagnosis and at relapse. Gene expression profiling of purified MMC was performed using Affymetrix U133 2.0 plus microarrays as described elsewhere²⁰ and data were normalized using the MAS5 Affymetrix algorithm with a scaling factor of 100. The .CEL and MAS5 files are deposited in the ArrayExpress public database (http://www.ebi.ac.uk/arrayexpress/), under accession number E-MTAB-362. Interphase-fluorescence in situ hybridization (iFISH) analysis was performed according to our previously reported standard protocol.²¹ We also used publicly available Affymetrix data (http://www.ncbi.nlm.nih.gov/geo, GSE2658) on purified MMC from a cohort of 345 previously untreated patients from the University of Arkansas for Medical Sciences (UAMS, Little Rock, AR, USA). These patients were treated with total therapy 2²² and this cohort is termed in the following the UAMS-TT2 cohort. Normal bone marrow plasma cells (BMPC) were obtained from healthy donors after they had given informed consent.^23,24 Peripheral blood CD14⁺ monocytes were purified by a FACS cell sorter from healthy donors’ peripheral blood buffy coats purchased from the French Blood Center (France) as indicated elsewhere.²³ Normal plasmablasts or plasma cells were obtained using our three-step in vitro model making it possible to obtain plasmablasts and then plasma cells starting from memory B cells within 10 days.²⁵

Quantitative analysis of KLF4 promoter methylation by pyrosequencing

Genomic DNA was isolated using a DNeasy Blood and Tissue Kit (Qiagen) and converted by bisulfite natrium using an EpiTect Bisulfite Kit (Qiagen). Converted DNA was then amplified by polymerase chain reaction (PCR) and pyrosequenced using biotinylated PCR primers and sequencing primer included in the Hs_KLF4_01_PM PyroMark CpG Assay (Qiagen) on a PSQ 96MA system. The analyzed sequence (CCCGACATACTGACGT-GCTGGCGGGCCACGCGCG) contains six CpG and is located on chromosome 9 region 110,250,599-110,250,633.

Drug resistance

HMCL (2×10⁵/mL) were cultured for 4 days in 24-well flat-bottom microtiter plates in 1 mL complete culture medium with interleukin-6 (2 ng/mL) and increasing concentrations of either melphalan or bortezomib. At the end of the culture, the number of viable cells in culture was determined with a Cell Titer Glo Luminescent Assay (Promega) based on ATP quantitation, which is directly correlated to the number of metabolically active cells. Some drug concentrations were also monitored using phycoerythrin-conjugated annexin V staining and FACS analysis (Boehringer).

Results

KLF4 expression in normal and malignant plasma cells

KLF4 was expressed in purified CD14⁺ monocytes, memory B cells as well as in BMPC of healthy individuals and was not expressed in in-vitro generated plasmablasts using Affymetrix microarrays (probe set 221841_s_at). These data were confirmed using real-time PCR (Online Supplementary Figure S1A), with a high correlation between KLF4 real-time PCR expression and Affymetrix microarray expression (probe set 221841_s_at) in three normal BMPC samples, three monocyte samples, and five primary MMC samples (r=0.83, P=0.001, results not shown). KLF4 gene expression was highly variable in primary myeloma cells from 206 newly diagnosed patients, with log₂ Affymetrix signals ranging from 2 to 13 (Figure 1A). For 12 patients, gene expression profiles of MMC were determined at both diagnosis and relapse, and no significant change in KLF4 expression was found (P=0.47, Online Supplementary Figure S1C). Twenty-six percent of the patients had KLF4^high MMC (i.e. with a KLF4 signal ≥mean value + 2 SD in normal BMPC) and 30% of the patients had a very weak KLF4 signal <100. The KLF4 gene was not expressed by any of 18 HMCL (median value 5, range 1–20). Using western blot analysis, KLF4 protein could be detected in the five samples of purified primary MMC in which the KLF4 gene was expressed (Online Supplementary Figure S1B). Recurrent genetic abnormalities in MMC were documented using iFISH. Patients with t(4;14)(p16.3;q32) translocation had higher expression of KLF4, with 52% of the patients having KLF4^high MMC versus 28% of the patients without t(4;14)(p16.3;q32) (P=0.001, Online Supplementary Tables S2 and S3). This was also the case for patients with del13, whereas no difference was found for other genetic alterations, del17p, 1q21 or t(11;14)(q13;q32).

Figure 1.

KLF4 gene and protein expression in normal or malignant plasma cells. (A) KLF4 gene expression was assayed, using Affymetrix microarray, in purified CD14⁺ monocytes, purified memory B cells, in vitro-generated plasmablasts or bone marrow plasma cells from healthy individuals. Histogram data are the KLF4 gene expression in each sample (probe set 221841_s_at). Gene expression was also assayed in primary MMC from 206 newly diagnosed patients and 20 human myeloma cell lines. The horizontal line is the mean KLF4 expression + 2SD in normal bone marrow plasma cells (BMPC). (B) KLF4 expression was assayed using Affymetrix microarray (probe set 221841_s_at) in purified multiple myeloma (MM) cells from 345 newly-diagnosed patients of the UAMS TT2 cohort. Data are the mean values ± SD of the KLF4 signal in each of the eight molecular groups of patients according to the UAMS classification: PR: proliferation, LB: low bone disease, MS: spiked MMSET expression, HY: hyperdiploid, CD-1 and CD-2: CCND1 or CCND3 expression, MAF: spiked MAF or MAFB expression and MY: myeloid.

KLF4 expression in molecular subgroups of patients with newly-diagnosed multiple myeloma

Using publicly-available gene expression profiling of purified MMC from the UAMS TT2 cohort, the KLF4 gene was also found to be highly expressed in patients with t(4;14)(p16.3;q32) translocation and spiked MMSET expression (MS group) in agreement with iFISH data (Figure 1B). The frequency of patients with KLF4^high MMC (i.e. KLF4 signal mean^BMPC + 2SD^BMPC) was 69% for the MS group and was significantly higher (P<0.05) than those for CD-2 (5%), hyperdiploid (6%), low bone disease (6%), MAF (0%), proliferative (0%) and myeloid groups (4%). The KLF4 gene was also highly expressed in patients in the CD-1 group (50% of patients with KLF4^high MMC), unlike patients in the CD-2 group. The CD-1 and CD-2 groups have spiked CCND1 or CCND3 expression due to the t(11;14)(q13;q32) or t(6;14)(p21;q32) translocations.¹⁵KLF4 expression had no prognostic value for event-free survival or overall survival in the TT2 cohort, as in our Heidelberg-Montpellier cohort, or in both cohorts grouped together. This is hardly surprising since KLF4 is over-expressed in the MS poor prognosis group and in the CD-1 good prognosis group. This holds true when analyzing prognosis among patients within subgroups. In particular, patients with t(4;14)(p16.3;q32) with either high or low KLF4 expression did not have a different overall survival (results not shown).

Epigenetic silencing of KLF4 in myeloma cell lines

A loss in KLF4 expression in MMC was not due to a loss in KLF4 allele as assayed using Affymetrix 50K Nsp mapping arrays. Indeed, none of 18 HMCL or 60 primary MMC investigated exhibited loss in the KLF4 gene (100-kb sensitivity, results not shown). However, 3/18 HMCL and 21/60 patients exhibited a gain (median one additional copy, range 1 to 3) of at least a 100-kb region containing the KLF4 gene, without a correlation with KLF4 expression (results not shown).

Treatment of HMCL with 5-azacytidine, a demethylating drug, induced KLF4 expression suggesting lack of KLF4 expression could be due to KLF4 promoter methylation (Online Supplementary Figure S2). Thus, KLF4 promoter methylation was assessed using pyrosequencing in four purified monocyte samples, ten HMCL (XG19, OPM2, LP1, XG6, XG7, RPMI, U266, NAN1, XG11 and XG16), and ten primary MMC purified from patients with diverse KLF4 gene expression (Figure 2). An inverse correlation between KLF4 expression and KLF4 promoter methylation was obvious with monocytes and HMCL (P<0.05). Monocytes, which highly expressed the KLF4 gene (Online Supplementary Figure S2B), had poor KLF4 promoter methylation (mean methylation of the six investigated CpG ranged from 4% to 5.5%), whereas the KLF4 promoter region was highly methylated in the ten HMCL that failed to express the KLF4 gene. Between 13% and 82.5% of the six CpG were methylated in the ten HMCL (Figure 2). In patients’ primary MMC, the mean methylation of the six CpG ranged from 7% to 40% and was not significantly correlated to KLF4 gene expression. This is hardly surprising since patients’ primary MMC are more heterogeneous than HMCL, with putative different MMC clones harboring various KLF4 promoter methylation states. In addition, the pyrosequencing method we used enables investigation of the methylation status of only 1% of about 500 CpG islets located in a 10 Kb region including the KLF4 start codon. Thus, in the case of partial methylation, many other CpG islets than the six investigated ones could contribute to the regulation of KLF4 expression.

Figure 2.

KLF4 promoter methylation in human myeloma cell lines. Genomic DNA was isolated from four purified peripheral blood monocyte populations from four healthy donors, ten HMCL samples, and primary MMC from ten patients. DNA was converted by bisulfite natrium, amplified by PCR and pyrosequenced (chromosome 9 region 110,250,599-110,250,633). Circles represent CpG dinucleotides, and the gray scale indicates methylation status ranging from no methylation (white) to complete methylation (black) of the analyzed sample. Mean bisulfite sequencing of the six analyzed CpG islands in the KLF4 promoter are also plotted above each sample.

Inducible expression of KLF4 did not affect myeloma cell survival but blocks myeloma cell proliferation

To look for the biological effect of KLF4 in myeloma cells, four HMCL - XG7, XG19, LP1, XG2 - were transduced with a combination of two lentiviral vectors allowing doxycycline-inducible KLF4 expression. Two HMCL –XG2 and LP1 – had no functional p53 because of a monoallelic deletion of the TP53 gene and inactivating mutation in the remaining allele and XG7 and XG19 HMCL had wild-type TP53 gene.¹⁹ Adding doxycycline induced KLF4 transgene expression in the four HMCL-TR-KLF4 at levels similar to those found in primary monocytes using real time reverse transcriptase PCR (Figure 3A). Doxycycline treatment also induced KLF4 protein expression in the four HMCL-TR-KLF4, as detected by western blotting (Figure 3B) or immunofluorescence (Online Supplementary Figure S3). Doxycycline-induced KLF4 resulted in a blockade of HMCL-TR-KLF4 growth that was obvious on and 3–4 days after addition of the doxycycline (Figure 3C). The cell growth blockade was not due to induced apoptosis, since no significant increases in annexin V⁺ LP1-TR-KLF4, XG2-TR-KLF4 or XG19-TR-KLF4 cells were found on day 4 of doxycycline stimulation. Only for XG7-TR-KLF4 cells was a slight, significant increase in annexin-V⁺ cells detected (from 15% to 25%, P<0.05) (Online Supplementary Figure S4).

Figure 3.

Inducible KLF4 expression in human myeloma cell lines blocks myeloma cell line growth. (A) Four HMCL-TR-KLF4 were treated for 2 days with doxycycline (dox). KLF4 gene expression was assayed using real time PCR and data are the relative KLF4 expression compared to untreated XG2-TR-KLF4 (value 100). Stars indicate significant increases of KLF4 gene expression using a Wilcoxon test for pairs (P<0.05). (B) KLF4 protein expression was assayed 4 days after Dox treatment of the HMCL-TR-KLF4 using western blotting. (C) The HMCL-TR-KLF4 were treated or not with Dox and cultured for 8 days. Data are the mean counts ± SD of viable cells using trypan blue exclusion in five separate experiments.

Doxycycline-induced cell growth blockade was due mainly to a cytostatic effect with accumulation of myeloma cells in the G1 and G2/M phases of the cell cycle and decrease of cells in the S phase. Detailed data using DAPI and anti-BrdU staining are shown for XG19 HMCL (Figure 4A) and summed up in Figure 4B for all four HMCL. Of note, myeloma cells could not resume cell cycling after removal of doxycycline. This is likely explained by the use of a KLF4 transgene lacking a 3’ region conferring KLF4 RNA instability.¹¹ Thus, given the stability of KLF4 transgene RNA, the myeloma cells stop cycling, but can no longer proliferate and gradually die. We, therefore, used a short-interfering (si) RNA, which can down-regulate doxycycline-induced KLF4 expression. This siRNA was also able to revert cell growth blockade conferred by doxycycline-induced KLF4, unlike a scrambled control siRNA (Online Supplementary Figure S5).

Figure 4.

Forced KLF4 expression blocks cell cycling in myeloma cell lines. XG19-TR-KLF4 cells were cultured with or without doxycycline (dox) for 4 days and cell cycle was quantified using DAPI staining and BrdU incorporation, labeling with anti-BrdU antibody and flow cytometry analysis. (A) XG2-TR-KLF4, XG7-TR-KLF4, XG19-TR-KLF4 and LP1-TR-KLF4 were analyzed as described for XG19-TR-KLF4 cells. The figures represent data from three independent analyses. (B) *Indicates a significant increase in G0/G1 phases; **A significant decrease in S-phase; ***A significant increase in G2/M phases using a Wilcoxon test for pairs (P<0.05).

Regulation of cell cycle proteins by KLF4

Doxycycline-induced KLF4 expression induced significant up-regulation of p27^Kip1 in the four HMCL: XG2, XG7, XG19, and LP1 (P=0.05, Figure 5, Online Supplementary Figure S6). Although p27^Kip1 was weakly expressed in the wild XG2 HMCL, it was reproducibly increased upon KLF4 induction in three separate experiments. KLF4 strongly induced p21^Cip1 in two HMCL – XG7 and XG19 –actually the two cell lines with wild-type TP53 genes. It did not induce p21^Cip1 in the other two HMCL with TP53 inactivating mutations - XG2 and LP1.¹⁹ Of note, p53 was strongly expressed in these two TP53-mutated HMCL and mutated p53 was increased by KLF4 expression. No change in p16 was found (results not shown).

Figure 5.

Forced KLF4 expression induced p27^Kip1 and p21^Cip1 in myeloma cell lines. The HMCL-TR-KLF4 were cultured for 5 days with or without doxycycline (dox) and protein was assayed using western blot analysis. Membranes were blocked and stained with polyclonal rabbit anti-KLF4, anti-p53, anti-p27^Kip1 antibodies, or mouse anti-p21^Cip1 antibody. A mouse monoclonal anti-β-actin antibody was used as the control. The antibody binding was revealed using peroxidase-conjugated secondary antibodies and an enhanced chemiluminescence detection system.

KLF4 expression protects some myeloma cell lines from melphalan-induced cell death, but not bortezomib-induced cell death

The ability of KLF4 to block cell cycling of myeloma cells and its high expression in patients with t(4;14)(p16.3;q32) could explain the poor survival of patients treated with chemotherapy that targets the cell cycle and stem cell transplantation. First, in 50 previously untreated patients with t(4;14)(p16.3;q32), there was a significant (P<0.05) inverse correlation between KLF4 expression in MMC and a gene expression-based proliferation index,³² suggesting that the in vitro observation of a KLF4 cytostatic effect on myeloma cell lines could occur in vivo (Online Supplementary Figure S9). We then investigated whether KLF4 expression could protect HMCL from the apoptosis induced by the two major drugs used in the treatment of MM, melphalan and bortezomib. Four HMCL were used: XG7 harboring the t(4;14)(p16.3;q32) translocation, XG1 the t(11;14)(q13;q32) translocation and XG2 and XG19 without these translocations. Different setups were tested to look for drug resistance. Cells were pulsed with doxycycline for 1, 2 or 3 days, then treated with increasing drug concentrations for 2 to 3 days (Online Supplementary Figure S8). The setup resulting in a maximum KLF4-induced drug resistance was the one pulsing cells for 2 days with doxycycline prior to a 3-day drug treatment. In five different experiments, forced KLF4 expression reproducibly induced a partial but significant protection to 16 or 32 μM melphalan for XG7 (P<0.001) and XG2 (P<0.05) HMCL using an ATP assay. Forced KLF4 expression did not protect XG1 or XG19 cells from melphalan-induced killing and did not protect any of the four HMCL from bortezomib-induced killing. The partial protection from melphalan-induced apoptosis that forced KLF4 expression conferred was also evidenced using annexin V staining. XG7 and XG2 expressing KLF4 both showed a significant (P<0.05) increase in cell viability following treatment with 25 μM melphalan, unlike XG1 and XG19 cells expressing KLF4 (Figure 6).

Figure 6.

Forced KLF4 expression protects myeloma cells from melphalan-induced cell death. Results shown are the percentages of viable cells monitored using phycoerythrin-conjugated annexin V staining for the four cell lines treated with 25 μM melphalan. * indicates a statistically significant difference from that obtained in the control group using a Student’s t test (P≤0.05).

We investigated whether the KLF4 protective effect could be explained by modulation of the p53 pathway induced by melphalan. In two wild-type TP53 XG7 and XG19 HMCL, melphalan induced p53 stabilization and p21 activation and forced KLF4 expression did not affect this effect. Regarding mutated TP53 HMCL, melphalan induced no p21 and an increase in mutated p53. Again, forced KLF4 expression did not change this profile.

Discussion

KLF4 is a complex transcription factor that is expressed in differentiated cells but also in pluripotent stem cells. It acts as either a tumor suppressor or tumor enhancer.¹ We show here that the KLF4 gene is expressed in healthy mature plasma cells, but was not expressed by any of 18 HMCL tested. The lack of KLF4 expression is not due to gene deletion as no loss in the KLF4 locus was found in primary MMC or HMCL. On the contrary, gains in the KLF4 locus were found in MMC from one-third of patients or HMCL, but without correlation with KLF4 gene expression. Many aberrant DNA copy number variations, comprising coding sequences, are found in MMC from previously untreated patients, but do not result in any change in expression of the vast majority of these genes.³³ These copy number variations are likely created by genomic instability occurring in MMC without specifically targeting a pathway and this is probably the case for copy number variations targeting the KLF4 locus. Actually, KLF4 is silenced in these HMCL through promoter methylation, as demonstrated by induction of KLF4 expression using 5-azacytidine treatment and pyrosequencing. Epigenetic silencing of the KLF4 promoter has already been documented in epithelial tumors,³⁴ as well as in non-Hodgkin’s and Hodgkin’s lymphoma cells.¹² Whereas there was a perfect correlation between KLF4 expression and KLF4 promoter methylation in monocytes and HMCL (P<0.05), this was less clear for patients’ primary MMC. This is hardly surprising since patients’ primary MMC are more heterogeneous than HMCL, with putative different MMC clones harboring various KLF4 promoter methylation states. In addition, the pyrosequencing method we used enables investigation of the methylation status of only 1% of about 500 CpG islets located in a 10 kb region including the KLF4 start codon. Thus, in the case of partial methylation, many other CpG islets than the six investigated ones could contribute to the regulation of KLF4 expression. Finally, besides promoter methylation, other mechanisms may regulate KLF4 gene expression, such as availability of transcription factors (p53, retinoic acid receptor, CDX2, SP1 or SP3),¹ histone modifications or nucleosome positioning. KLF4 gene expression is for instance up-regulated by the histone deacetylase inhibitor trichostatin A through acetylation of histones H3 and H4.³⁵ Regarding transcription factors regulating KLF4 transcription, we did not find differences in expression of TP53, retinoic acid receptor, CDX2, SP1 or SP3 genes between KLF4^high and KLF4^low MMC.

This KLF4 silencing in HMCL is understandable since a conditional expression of KLF4 transgene resulted in a blockade of the growth of the HMCL tested. The arrest in HMCL growth was due mainly to a blockade of the cell cycle since KLF4 induced no significant apoptosis in three of the four HMCL and only weak apoptosis in XG7. The KLF4 growth inhibitory effect on HMCL was independent of their p53 status and was observed in two HMCL - XG7 and XG19 - carrying wild type TP53 genes and two - XG2 and LP1 - carrying mutated TP53 genes. This could be explained by up-regulation of p27^Kip1 in all HMCL, in agreement with previous findings showing that KLF4 can bind and activate the CDKN1B promoter. p27^Kip1 inhibits cyclin D/CDK4–6, cyclin E/CDK2 and cyclin A/CDK2 complexes. In the two wild-type TP53 HMCL, KLF4 induced an increase in p21^Cip1, in agreement with its ability to co-activate CDKN1A expression together with p53,⁹ whereas in the two HMCL with inactive p53, no p21^Cip1 could be detected using western blotting. The complete KLF4-induced growth inhibition in wild-type or mutated TP53 HMCL differs from previous findings with lymphoma cell lines.¹² Forced KLF4 expression induced a slight growth retardation of Burkitt’s cell lines, and a dramatic apoptosis of Hodgkin’s lymphoma cell lines.¹² Actually, KLF4 interacts with many partners, and the presence or absence of these partners may change its biological effects.¹

Whereas KLF4 was not expressed by any HMCL, in agreement with its suppressor function, its expression in primary MMC was highly variable, suggesting a more complex function. The expression of 177 genes was significantly correlated with that of KLF4 in MMC and the expression of 458 genes was different between HMCL with or without forced KLF4 expression, with only five genes common to the two lists. Ingenuity pathway analysis of genes differentially expressed in HMCL-TR-KLF4 with or without doxycycline highlighted a “Cell death and survival” network composed of five genes: PSEN1, IL1B, EP300, NFκB1 and RELA. None of these five genes was found when analyzing genes correlated to KLF4 in primary MMC. As KLF4 has a role in inducing pluripotent stem cells, an enrichment of ‘stemness’ genes was looked for in KLF4^highversus KLF4^low primary myeloma cells. No enrichment was found using ingenuity pathway analysis or using a recently-reported cell-cycle unrelated ‘stemness’ gene signature in myeloma cells³⁶ (results not shown).

KLF4 expression was very low in primary MMC from the majority of patients and high in 26% of patients (KLF4 signal ≥mean value + 2 SD in normal BMPC), namely patients with the t(4;14)(p16.3;q32) translocation and a subgroup of patients with high CCND1 or CCND3 expression (CD-1). As commented above, several mechanisms - transcription factors, epigenetics - could account for this high KLF4 expression in these two subgroups of patients. We did not find differential expression of a gene coding for transcription factors known to control KLF4 gene expression, being aware that the activity of a transcription factor is mainly regulated by protein modifications. It is noteworthy that two genes, coding for SET domain histone methyltransferases involved in gene regulation and DNA repair pathways, could regulate KLF4 expression. These genes are highly expressed in two groups of patients: MMSET in patients with t(4;14)(p16.3;q32) and SET7/9 in patients in the CD-1 group. MMSET is a histone methyltransferase that methylates H3K27, H3K36 and H4K20. It is involved in DNA repair controlling 53BP1 recruitment through histone H4 methylation.³⁷ It is also an EZH2 effector³⁸ and controls expression of genes coding for cell cycle and survival pathways (MYC, NK-κB, Rb, Bax, Sox2). SET7/9 methylates H3K4 and is a tumor suppressor activating p53 activity through p53 methylation and acetylation.^35,36 The fact that KLF4 has a tumor suppressor role in MMC could explain why positive KLF4 regulation in MMC with t(4;14)(p16.3;q32) or belonging to CD-1 group is abrogated in HMCL through promoter methylation in HMCL.

It is noteworthy that forced KLF4 expression can partially protect some HMCL from melphalan-induced MMC killing, but had no influence on the effect of bortezomib. Melphalan binds DNA creating mostly monoadducts and 5 to 10% interstrand crosslinks,³⁹ which are repaired by the Fanconi anemia pathway and homologous recombination in replicating cells or by less known mechanisms in non-replicating ones.⁴⁰ The partial KLF4 protective effect from melphalan is not related to blockade of the p53 apoptotic effect, as reported for a bladder cancer cell line.⁴¹ Indeed, KLF4 conferred protection in two out of four HMCL, one with wild-type TP53 (XG7) and the other with mutated TP53 (XG2) and no protection to the wild type TP53 XG19 cell line or mutated TP53 XG1 and LP1 cell lines. We confirmed that forced KLF4 expression did not affect activation of p53 and p21 by melphalan with or without forced KLF4 expression (Online Supplementary Figure S7). In contrast, forced KLF4 expression alone induced p53 stabilization and p21. A likely explanation is that KLF4 overexpression may protect a small portion of cells from the toxicity of alkylating agents by blocking replication machinery in the presence of interstrand crosslinks.⁴² The lack of KLF4 protection from bortezomib-induced killing is likely explained by the fact that bortezomib targets proliferating cells as well as non-proliferating cells.⁴³ It targets protein degradation, which is central to protein recycling, in particular for HMCL that produce large amounts of immunoglobulins. This lack of bortezomib resistance induced by forced KLF4 expression in vitro is in line with the improvement of response rate in patients with t(4;14)(p16.3;q32) translocation and KLF4^high MMC treated with bortezomib.⁴⁴

In conclusion, we show here that KLF4 participates in the complexity of multiple myeloma, likely being an anti-proliferative agent in MMC but also protecting them from cell cycle-specific drugs.