We focused on studying KLF4 as the primary example because it is frequently de-regulated in B-cell tumors [28,29,30,31,32,33,34,35,36,37,38], but functional in vivo data in advanced B-ALL disease is missing. We performed reverse genetics to decipher the role of KLF4 in certain B-cell tumors. To examine both tumor formation and tumor maintenance, we began with the in vitro surrogate model of EBV-induced B-cell transformation.
KLF4 blocks EBV-induced transformation of B-cells in vitro
We chose this in vitro model of B-cell oncogenesis, which reprograms B-lymphocytes into B-blasts (Fig. 1a) [44,45,46] because we had previously detected, strong and persistent downregulation of KLF4 mRNA levels during EBV-induced B-cell transformation (Supplementary Figure S2A and ref. [47]).
Two different variants of KLF4, wildtype KLF4 (wtKLF4) and a variant form of KLF4 (mutKLF4) were cloned into recombinant viruses (Supplementary Figure S2B and C). Upon infection, they coexpressed wtKLF4 or mutKLF4 together with viral genes in primary B lymphocytes. mutKLF4 lacks the two C-terminal zinc fingers representing the DNA binding domain, which disables the basic functions of KLF4 as a transcription factor, but the protein-protein interactions of KLF4 are not affected (Supplementary Figure S2B).
Naïve human B-lymphocytes were infected with EBV to induce their transformation into B-blasts by viral oncogenes (Fig. 1a, mock). To study the influence of KLF4 for the process of transformation, EBV was genetically modified and used as a shuttle vector to express wtKLF4 or mutKLF4 together with viral oncogenes in resting B-lymphocytes that undergo EBV-induced oncogenesis (Fig. 1b, Supplementary Figures S2C, 3A).
While mutKLF4 expression showed no major effects in EBV-infected B-cells compared with mock-EBV infection, the maintenance of wtKLF4 levels completely abrogated EBV-induced B-cell transformation, reduced EBV-induced B-cell proliferation and induced apoptosis in the infected cells (Fig. 1c, Supplementary Figure S3B). Cell cycle analysis indicated that KLF4 expression prevented S-phase entry in EBV-infected B-lymphocytes, which is required for the generation and expansion of B-blasts. Thus, KLF4 acts as a cell cycle inhibitor in the model of in vitro EBV-induced oncogenesis and conversion of normal B-cells into B-blasts.
In conclusion, persistent expression of KLF4 blocks human B-cell transformation in a model of EBV-induced oncogenesis.
This finding is in line with published data that suggest a role of KLF4 as tumor suppressor in BCR-ABL-mediated transformation of murine pre-B-cells into B-ALL [39]. Based on these findings and in line with our previous observation and published results that i) KLF4 is severely downregulated in primary pediatric B-ALL cells [35,36,37,38], especially after therapy [43] (Supplementary Figure S4A), and ii) among adult ALL samples, the aggressive Ph+-like B-ALL subgroup showed low expression levels of KLF4 similar to those of T-ALL [22] (Supplementary Figure S4B); we went on to investigate a putative role of KLF4 in leukemia models of advanced disease.
Inducible transgene expression in PDX cells in vivo
We asked whether KLF4 is functionally relevant for established B-cell tumors in vivo and whether it might represent a putative therapeutic target. As a clinically relevant model for functional genomic studies of established tumors, we used orthotopic PDX obtained from two children with relapsed B-cell precursor ALL (clinical data for both patients are listed in Supplementary Table S1). Similar to pediatric [35,36,37,38] and adult (Supplementary Figure S4B) primary B-ALL samples, both PDX B-ALL models revealed low KLF4 mRNA and protein levels (Fig. 2a and b).
For the conditional expression of KLF4 in PDX B-ALL samples, we engineered a tetracycline-response system to study the characteristics of PDX models independently of tumor cell transplantation and at different time points, e.g., before, during and after treatment and at the minimal residual disease stage.
PDX cells were consecutively transduced with three lentiviral constructs (Fig. 2c and d and Supplementary Figure S5A) to express (i) the tet activator (rtTA3) together with luciferase to allow in vivo bioluminescence imaging; (ii) the tet repressor (tetR) together with one of two different fluorochromes as molecular markers for competitive in vivo assays; and (iii) a mock control, wtKLF4 or mutKLF4 gene (Supplementary Figure S2B) in conjunction with the fluorochrome mCherry and under the control of the tet-responsive element (TRE) [42] (Fig. 2d and Supplementary Figure S5A). We selected a low transduction rate of 1% for each vector to permit the integration of only a single copy of each transgene per cellular genome. Although silencing of the CMV promoter is common in leukemic cells, the TRE-regulated mini-CMV promoter was not epigenetically repressed in our PDX models (data not shown). Doxycycline (Dox) treatment (indicated by the green background in all figures) of transduced B-ALL cells induced the expression of KLF4 together with mCherry, which serves as a proxy to monitor the coexpression of the wtKLF4 or mutKLF4 protein (Fig. 2e and Supplementary Figure S5B and C). Upon Dox withdrawal, the expression of mCherry and KLF4 returned to baseline levels, demonstrating the reversibility of the system (Supplementary Figure S5C). Notably, KLF4 protein expression in the induced state was moderate and did not exceed the levels observed in PBMCs, avoiding undesirable overexpression (Fig. 2e). Importantly, feeding Dox to mice induced the expression of mCherry/KLF4 in PDX B-ALL cells in vivo (Supplementary Figure S5B), enabling the analysis of KLF4-mediated effects at any given time point in mice. The combination of different marker fluorochromes provided the opportunity to investigate PDX cells with conditional wtKLF4 and control (mock) alleles in the same animal by pairwise competition experiments, abrogating animal-to-animal variability.
Taken together, the results showed that the newly established inducible system of transgene expression in PDX of acute leukemias appeared to be functional and represent an attractive method to identify and study leukemia vulnerabilities in individuals and to test established and potential drugs for their anti-leukemic potency in vivo.
Re-expressing wtKLF4 reduces the growth and homing of B-ALL PDX cells in vivo
In the first step, we determined whether low KLF4 expression levels were essential and required for the survival and proliferation of B-ALL tumors in vivo. Two PDX B-ALL samples, consisting of ALL-265 (Figs. 3, 4 and 5) and ALL-199 (Supplementary Figure S6) cells, that were transduced with either mock or wtKLF4 alleles were transplanted into groups of mice. When cell homing and early engraftment were completed, the mice were fed with Dox to induce transgene expression during the exponential growth phase of the pre-established PDX B-ALL tumors (Fig. 3a and Supplementary Figure S6A).
Tumor growth of wtKLF4-expressing B-ALL PDX cells was reduced compared with that of control cells, as determined by in vivo imaging (Fig. 3b and Supplementary Figure S6B). Consequently, KLF4 re-expression resulted in a reduction in splenomegaly compared with that in mock gene-bearing mice (Fig. 3c, d and Supplementary Figures S1 and 6CD). KLF4 significantly reduced the fraction of cells in S-phase, accompanied and eventually caused by upregulation of the bona fide KLF4 target gene p21 (Figs. 3e and Supplementary Figure S6E and Table S2), but also induced cleavage of PARP and caspase-3 (Fig. 3f and Supplementary Figure S6F) indicating that KLF4 impaired tumor growth by inducing cell cycle arrest and apoptosis. Comparable results were obtained in a B-ALL cell line (Supplementary Figure S7) supporting our observations in vivo.
The effect of KLF4 on tumor growth was further investigated in competition assays in vivo when identical numbers of mock-transduced control cells and wtKLF4-transduced cells derived from the same PDX model were mixed and injected into the same animal (Fig. 4a, indicated in orange). As an additional control, a second group of mice was injected with mock- and mutKLF4-transduced cells (indicated in purple in Fig. 4a). The leukemic load was monitored by in vivo imaging, and the cell ratios of the subpopulations were quantified using the respective fluorochromes.
In the competition experiments, mice injected with the wtKLF4 mixture displayed reduced tumor growth in vivo, although 50% of cells in the wtKLF4 mixture were control cells (Fig. 4b, left panels). Cells transduced with the wtKLF4 allele had a severe growth disadvantage and nearly disappeared within 30 days (Fig. 4b, lower right panel) whereas mock-transduced cells expanded readily. In control mice, the ratio of the two subpopulations remained constant over time (Fig. 4b, upper right panel) suggesting that genetic manipulation per se did not affect the oncogenic properties of the leukemic PDX models.
The in vivo competitive assay (Fig. 4a and b) confirmed the growth-inhibitory effect of wtKLF4 that we observed in previous experiments (Fig. 3). Among the major advantages of competitive in vivo assays is that they can correct for high intermouse variation and have a high reliability and sensitivity, and thus they represent a powerful tool to identify, even moderate functional contributions to tumor maintenance in vivo.
Ectopic expression of KLF4 impaired B-ALL growth and survival in vivo, but KLF4 might also affect engraftment into the murine bone marrow. To investigate this possibility, we analyzed the capacity of KLF4-expressing PDX cells to home in on the murine bone marrow in retransplantation experiments. Mock-transduced or wtKLF4 or mutKLF4 expressing PDX cells were isolated from the first recipient mice that received Dox at an advanced disease stage (Supplementary Figure S8A). Equal numbers of cells were combined and retransplanted into second-generation mice and cell homing was assessed 3 days later (Supplementary Figure S8A). The ratio of transplanted mock versus mutKLF4-expressing cells was not altered in the bone marrow after 3 days, but in reisolated cell populations obtained from mice that had received mock- and wtKLF4-transduced cells, the abundance of the latter was strongly reduced (Supplementary Figure S8B), indicating that wtKLF4 expression impairs homing of PDX B-ALL cells in vivo.
Taken together, the results demonstrated that the re-expression of KLF4 reduced the overall fitness of PDX B-ALL cells in vivo. KLF4 functions as a cell cycle inhibitor and pro-apoptotic factor in PDX B-ALL resulting in decreased growth of established tumors and an impaired ability to infiltrate the murine bone marrow.
Treatment-surviving cells are especially sensitive towards the re-expression of KLF4
In B-ALL, persistent minimal residual disease (MRD) is the most highly predictive factor for disease-free survival [11], emphasizing the need to target and eradicate MRD. The number of primary MRD cells isolated from each patient is small by definition, so elaborate in vivo models are necessary to functionally characterize the MRD cell subpopulation, which is a serious technical challenge [43]. The inducible expression system provides the means to characterize the role of individual genes in MRD.
To mimic MRD in mice, we titrated a combination of the routine drugs vincristine and cyclophosphamide at clinically relevant doses to effectively reduce the tumor burden by several orders of magnitude over 5 weeks. When MRD was achieved (defined as less than 1% leukemia cells in the bone marrow) chemotherapy was discontinued and Dox was administered to induce KLF4 expression during the phase of tumor regrowth (Fig. 4c).
The treatment responses of the control cells and wtKLF4 mixture were comparable, as indicated by in vivo imaging (Fig. 4d, left panel). In contrast, upon Dox administration in MRD, tumor regrowth was clearly diminished in mice carrying PDX cells transduced with the wtKLF4 mixture (Fig. 4d, left panel). Subfractional analysis revealed that the expression of wtKLF4 (but not mutKLF4) was reduced, and finally, even cellular regrowth was prevented in vivo, which depleted wtKLF4 cells to undetectable levels within 2 weeks (Fig. 4d, right panel). When comparing therapy-naïve, wtKLF4-expressing cells (Fig. 4b) with similar cells obtained after chemotherapy in MRD (Fig. 4d, lower right panel), KLF4 inhibited the growth of MRD cells more dramatically than that of therapy-naïve PDX B-ALL cells (Supplementary Figure S9). It thus appears that MRD cells are especially sensitive to KLF4 expression, suggesting that any regimen that induces KLF4 expression in MRD might be especially effective for tumor consolidation therapy.
KLF4 sensitizes B-ALL PDX cells to chemotherapy in vivo
Low KLF4 expression levels were a prerequisite for the in vivo growth of PDX B-ALL samples, especially after mice underwent experimental chemotherapy. Our subsequent experiments addressed the question of whether the re-expression of KLF4 sensitizes patients’ B-ALL cells to conventional chemotherapy during routine ALL treatment in vivo in the setting of induction therapy.
PDX B-ALL control and wtKLF4 mixtures were injected into mice, and Dox was administered together with chemotherapy at a high tumor burden (Fig. 5a). In this experiment, Dox was administered simultaneously with chemotherapy, which was administered at a dose intended to stop tumor progression but not reduce the tumor size to facilitate the identification of KLF4-induced phenotypes.
Accordingly, chemotherapy prevented further tumor progression in mice injected with control cells as intended, but it also reduced the tumor load in mice injected with PDX cells transduced with control and wtKLF4-expressing vectors (Fig. 5b, left panel). When the subcellular fractions of these animals were analyzed, the number of wtKLF4- but not mutKLF4-expressing cells was significantly decreased by chemotherapy, and therefore the wtKLF4-expressing cells were outcompeted and lost within less than 40 days (Fig. 5b, right panel). KLF4 re-expression also sensitized NALM-6 cells to chemotherapy in vitro (Supplementary Figure S10) confirming the results obtained from PDX B-ALL cells in vivo. The data suggest that the upregulation of KLF4 may synergize with standard therapeutic regimens, i.e. conventional chemotherapy to eliminate B-ALL cells in patients.
Azacitidine-induced cell death partially depends on KLF4
Next, we searched for drugs capable of upregulating KLF4. We first tested the small molecule APTO-253, which was developed as a KLF4-inducing drug [48] but was recently shown to also regulate MYC [49]. APTO-253 indeed upregulated KLF4 in our experiments and sensitized cell lines towards vincristine treatment in vitro (Supplementary Figure S11), indicating that KLF4 can, in principle, be reupregulated by drugs in B-ALL.
We speculated that existing approved drugs might be able to reupregulate KLF4 in B-ALL to facilitate clinical translation. The demethylating agent 5-azacitidine (Aza) was shown to upregulate KLF4 expression in other tumor entities [22, 28, 31, 50]. Although Aza can reverse DNA hypermethylation, it regulates protein expression on multiple levels, and multiple mechanisms have been shown to contribute to the antitumor activity of Aza [51, 52]. Aza is increasingly used in clinical trials on patients with hematopoietic malignancies and MRD, e.g., in acute myeloid leukemia, to prevent or delay relapse [53]. Here we found that, already within 48 h, Aza upregulated KLF4 levels in B-ALL PDX cells as well as B-ALL cell lines (Fig. 6a and Supplementary Figure S12A), and decreased cell viability in vitro at clinically relevant doses (Supplementary Figure S12A).
To assess whether Aza-induced cell death in B-cell tumor cells is mediated by increased levels of endogenous KLF4 protein, we generated KLF4 knockout (KO) cells, which is a challenge for PDX. For B-ALL PDX, cells were transduced with a lentiviral vector to introduce Cas9 (Fig. 6b and Supplementary Figure S12B). In subsequent steps, a pool of 3 vectors, each encoding a single KLF4 targeting guide RNA, and a reporter construct were introduced to enrich gene-edited cells. PDX B-ALL cells with stable KO of KLF4 could readily be established (Fig. 6c).
Stimulation of PDX B-ALL cells with Aza at clinically relevant concentrations reduced cell viability in control cells but not in PDX B-ALL cells with KLF4 KO (Fig. 6d). Similar results were obtained in three B-ALL cell lines (Supplementary Figure S13), using either individual sgRNAs or pools of 3 sgRNAs to generate KLF4 KO cells. These data suggest that KLF4 at least partially mediates the effect of Aza on B-ALL cells.
Taken together, the results indicate that KLF4 might represent a novel therapeutic target for B-ALL. Our data also show that Aza, an established drug, can upregulate KLF4. Introducing Aza into standard polychemotherapy protocols for B-ALL patients with the intention of raising KLF4 levels might reduce tumor burden and increase sensitivity to conventional chemotherapy. Patients with B-ALL might benefit from this treatment option; further investigation for proof-of-concept should be performed in clinical trials.