Potent anti-leukemic activity of a specific cyclin-dependent kinase 9 inhibitor in mouse models of chronic lymphocytic leukemia

Onset of progression even during therapy with novel drugs remains an issue in chronic lymphocytic leukemia (CLL). Thus, there is ongoing demand for novel agents. Approaches targeting cyclin-dependent kinases (CDK) have reached the clinical trial stage. CDK9 mediating RNA transcriptional elongation is the evolving pivotal CLL CDK inhibitor target. However, more CDK9 selective compounds are desirable. Here, we describe the CDK9 inhibitor LDC526 displaying a low nanomolar biochemical activity against CDK9 and an at least 50-fold selectivity against other CDKs. After demonstrating in vitro MEC-1 cell line and primary human CLL cell cytotoxicity we evaluated the LDC526 in vivo effect on human CLL cells transplanted into NOD/scid/γcnull (NSG) mice. LDC526 administration (75 mg/kg) for 5 days resulted in a 77% reduction of human CLL cells in NSG spleens compared to carrier control treatment. Next, we longitudinally studied the LDC526 impact on circulating CLL cells in the TCL1 transgenic mouse model. LDC526 (50 mg/kg) administration for two days led to a 16-fold reduction of blood CLL cell numbers. Remarkably, residual CLL cells exhibited significantly increased intracellular BCL-2 levels. However, the LDC526 cytotoxic effect was not restricted to CLL cells as also declining numbers of normal B and T lymphocytes were observed in LDC526 treated TCL1 mice. Taken together, our in vivo data provide a strong rational for continued LDC526 development in CLL therapy and argue for the combination with BCL-2 inhibitors.


INTRODUCTION
CLL is the most common leukemia in Western countries [1]. The clinical course of CLL is highly variable. Specific biologic and clinical factors are associated with rapid progression towards more advanced disease stages in some patients whereas others survive for a long period without the need for treatment [2]. Compared to other hematologic malignancies CLL is a nonproliferative leukemia characterized by the over-expression of antiapoptotic BCL-2 family proteins [3]. Survival signals are provided by the highly specialized microenvironment of CLL [4][5][6]. It is possible to achieve minimal residual disease-negative CLL remissions with combination chemo-immunotherapy in a subset of patients [7], however the majority of patients will inevitably relapse. This will likely also hold true in the current era of novel anti-CLL drug approvals targeting Bruton's tyrosine kinase (BTK), phosphatidylinositol-3-kinase delta (PI3Kδ) and B-cell lymphoma/leukemia 2 (BCL-2) [8]. Patients treated www.oncotarget.com Oncotarget, 2018, Vol. 9, (No. 41), pp: 26353-26369

Research Paper
with these novel agents still progress after treatment or may not tolerate them. Treatment failure with the novel agents may occur due to acquired resistance, which in the case of the BTK inhibitor ibrutinib was shown to be mediated by mutations in the ibrutinib-binding site of BTK [9]. Furthermore, patients also progress under treatment with the BCL-2 inhibitor venetoclax. It is assumed that venetoclax resistance is conferred by the overexpression of pro-survival proteins other than BCL-2 such as Myeloid cell leukemia 1 (MCL-1) [10][11][12]. Likely, a priori CLL dependence on MCL-1 rather than BCL-2 [13] conveys decreased venetoclax sensitivity in a subgroup of patients. Additionally, CLL MCL-1 expression is associated with the presence of poor prognostic markers and disease progression [14].
MCL-1 is a protein with a short half-life and its cellular levels are thus susceptible to transient inhibition of RNA transcription [15][16][17]. RNA transcription and in particular elongation are dependent on cyclin-dependent kinase 9 (CDK9) mediated serine phosphorylation of the RNA Polymerase II (RNAPII) carboxyterminal domain (CTD). CDK9 together with its cyclin partners (T or K) forms a functional complex termed positive transcription elongation factor b (pTEFb). The first generation CDK9 inhibitors such as SNS-032 or Alvocidip (flavopiridol) also targeting other cyclin-dependent kinases are capable of inducing apoptosis of CLL cells [18,19]. However, the clinical development of these compounds was negatively impacted by their side effect profile in particular by the occurrence of cytopenias, gastrointestinal symptoms and tumor lysis syndrome [20][21][22]. Likely, the combinatorial inhibition of multiple CDKs contributed to this side effect spectrum. The next-generation CDK inhibitor Dinaciclib specific for CDK1, CDK2, CDK5 and CDK9 was more efficient in inducing CLL apoptosis than flavopiridol [23,24] and exhibited an improved safety profile [25,26]. Nonetheless, the occurrence of cytopenias was still reported in Dinaciclib clinical trials [25,26].
To further increase CDK9 inhibitor specificity and to enable oral administration we developed the novel CDK9 inhibitor LDC526. A recent further pharmacologically optimization of LDC526 resulted in BAY1143572 [27], which has been studied in phase I trials in patients with acute leukemia and solid tumors / lymphomas (ClinicalTrials.gov, Identifier NCT02345382 and NCT01938638, respectively). Here, we report anti-CLL activity of LDC526 in the CLL-derived cell line MEC-1 and in primary CLL cells. Moreover, we demonstrated effective anti-CLL activity of LDC526 in CLL xenografted NSG and TCL1 transgenic CLL mice. In these models LDC526 treatment also decreased non-malignant T cells, which represent an important component of the CLL microenvironment. High BCL-2 expression likely enabled a small fraction of CLL cells to escape LDC526-induced apoptosis.
As CDK9 inhibition should rapidly decrease levels of the prosurvival MCL-1 protein we studied apoptosis of the CLL-derived cell line MEC-1 after short-term LDC526 exposure (4 hours). Indeed, significantly increased MEC-1 apoptosis was observed at LDC526 concentrations of 500 nM and more pronounced at 1 μM ( Figure 1D, 1E). In parallel, we studied intracellular MCL-1 and BCL-2 expression on the single cell level using an intracellular flow cytometric assay ( Figure 1F). In contrast to BCL-2, MCL-1 protein levels were reduced in the presence of increasing LDC526 concentrations ( Figure 1F, 1G, 1H). Longer-term LDC526 exposure also induced apoptosis ( Figure 1I) and decreased the number of viable MEC-1 cells at different concentrations ( Figure 1J). To demonstrate that LDC526 primarily induces apoptosis rather than inhibiting proliferation we studied cell cycle progression by BrdU incorporation at different LDC526 concentrations ( Figure 1K). In contrast to the impact on cell survival, LDC526 did not alter the cell cycle at concentrations up to 1 μM. Solely at very a high concentration such as 10 μM S-phase entry was significantly decreased ( Figure 1L). Next, we studied whether the MEC-1 cell line data were reproducible with primary CLL samples. CLL PBMCs were exposed to LDC526 for four hours and apoptosis was determined with Annexin V staining. Strikingly, significantly increased percentages of apoptosis were observed at concentrations of 500 and 1000 nM (Figure 2A, 2B). Furthermore, increasing LDC526 concentrations led to significantly decreased MCL-1 protein levels while BCL-2 levels were not significantly altered ( Figure 2C, 2D, 2E).
Subsequently, we investigated the in vitro cytotoxic impact of LDC526 on primary CLL cells (patient characteristics, Table 1) in comparison to healthy donor B cells after longer-term LDC526 exposure ( Figure 2F). The www.oncotarget.com   Table 1. Means±SEM are shown, no statistically significant differences between prognostic CLL subgroups (p>0.05) using Mann-Whitney-U test. www.oncotarget.com number of viable (Annexin V-/DAPI-) CLL cells decreased with increasing LDC526 concentrations. Remarkably, normal donor B cells also displayed sensitivity towards LDC526. However, the IC 50 of primary CLL cells was lower than the IC 50 of non-malignant normal donor B cells. This difference in sensitivity was more pronounced at 48 hours than at 24 hours of incubation ( Figure 2F). Finally, we analyzed whether prognostic CLL subgroups (Table 1) had a significant impact on LDC526-induced apoptosis. CLL patients were split into subgroups with respect to fluorescent in situ hybridization (FISH) genetic aberrations, CD38 surface expression, Binet stage and treatment history. In accordance with the assumption that even high-risk CLLs with e. g. p53 aberrations should be susceptible to the transcriptional elongation inhibition by LDC526 no statistically significant differences were observed between subgroups ( Figure 2G). Taken together, LDC526 displayed efficient cytotoxic activity against CLL cells in vitro. As expected, LDC526 decreased CLL MCL-1 protein levels and triggered apoptosis rather than inhibiting proliferation.

LDC526 decreased human CLL numbers in CLL xenografted NSG mice
In addition to the cytotoxic effect on CLL cells, selective CDK9 inhibition is likely to have pleiotropic effects in vivo. These include effects on the microenvironment of CLL cells comprising different cell types such as myeloid cells and T cells. Furthermore, proliferating activated CLL cells only exist in specific organ compartments such as lymph nodes, spleen and bone marrow. Therefore, we investigated the impact of specific pharmacologic CDK9 inhibition by LDC526 on CLL in vivo.
We and others previously established human CLL xenograft models using NOD/SCID and NSG mice, respectively [28][29][30]. We chose to use the NSG model for LDC526 experiments since NSG CLL and concomitant T cell engraftment levels are expected to be higher than in the NOD/SCID model [28][29][30]. We intravenously injected PBMCs from n=6 CLL patients into n=57 NSG mice and commenced LDC treatment with a 14 day latency according to the experimental set-up outlined in Figure  3A. To study the effect of doses and treatment duration we chose to administer 50 and 75 mg/kg for three days (NSG transplanted with cells originating from n=3 patients) and for five days (NSG transplanted with cells originating from n=3 additional patients). On experimental days 17 and 21 NSG spleens were harvested and the content of human CLL and T cells was determined by cell counting and flow cytometry ( Figure 3B). The LDC526 impact on splenic NSG human CLL cell numbers was separately evaluated per originating CLL patient as engraftment was shown to vary between patients [28,30]. Compared to carrier control treatment, numbers of human CLL cells decreased after 3 and 5 days of LDC526 treatment, respectively. However, statistical significance regarding the reduction of human CLL cell numbers was only achieved at the LDC526 75 mg/kg dose level ( Figure 3C). In contrast, the numbers of splenic human T cells were significantly reduced at both LDC526 doses administered ( Figure 3C). In order to determine whether there was differential in vivo LDC526 sensitivity of T cell subsets we performed additional staining for human CD4+ and CD8+ T cells present in the spleens of xenografted NSG mice ( Figure 4A). The human CLL T cell compartment originating from n=2 CLL patients analyzed after 3 days LDC526 treatment exhibited a lower in one case and a constant CD4/CD8 ratio in the other case. In contrast, in mice transplanted with cells of n=2 other CLL patients and receiving LDC526 treatment for 5 days we observed a rising CD4/CD8 ratio with increasing LDC526 doses ( Figure 4B). This could mean that CD8+ cytotoxic T cells are more susceptible to longer LDC526 exposure (5 versus 3 days) than CD4+ T cells. However, this differential skew within the T cell compartment could also reflect the heterogeneity of T cells derived from different CLL patients. On days 14-16 of the experimental plan (days 1-3 of LDC526 administration) mice did not show any signs of distress. A mean weight loss of 1.8% was detectable in mice of the LDC526 75 mg/kg group (day 16). However, on the last day of the experiment (day 21) mice of the LDC526 75 mg/kg dose group showed beginning signs of declined general condition with decreased general activity and the mean weight loss was 3.6% (Supplementary Figure 1A). In conclusion, LDC526 displayed significant in vivo anti-human CLL activity. Remarkably, human T cell numbers were already decreased at a lower dose level than human CLL cells.

LDC526 shows significant CLL cell cytotoxicity in TCL1 transgenic mice
Another widely used preclinical CLL model are TCL1 transgenic mice [31]. The TCL1 transgenics represent a constitutive CLL model not involving the adoptive transfer of cells. This model mirrors the natural course of human IgVH-unmutated CLL including its disease progression [32,33]. In addition, this model has also been widely used by others to test novel therapeutic molecules [34][35][36]. Hence, we used TCL1 transgenic mice to investigate the in vivo anti-CLL activity of LDC526. Consecutive blood sampling allowed us to carry out a longitudinal study in individual TCL1 and control mice. For this, we chose a cohort of TCL1 and littermate control mice, which were 9-12 months old ( Figure 5A). We chose TCL1 transgenics with high peripheral CLL cell (CD19+CD5+) counts (14.6±7.4/nl, mean±SEM). Given the presumably high leukemic burden in these aged TCL1 mice and the reduced general condition of some NSG mice receiving 75 mg/kg for 5 days we decided to dose TCL1 mice with 50 mg/kg daily for two consecutive days ( Figure 5A). As the experiment was longitudinally designed and there was no effect expected of administering carrier to wild-type control mice we omitted this control group and established three experimental groups ( Figure 5A). In order to minimize stress by collecting blood specimens we sampled blood three times, 8 days apart (days -5, +3 and +11; Figure 5A). Compared to blood smears prepared before commencing LDC526 treatment a marked decrease of peripheral blood leukocytes and CLL cells was observed two days post the last LDC526 administration to TCL1 transgenic mice ( Figure 5B). In contrast to the carrier control treated TCL1 mice, white blood cell counts (WBC) dropped in LDC526-treated wild-type and TCL1 mice ( Figure 5C). In comparison to wild-type control mice, where WBC dropped to 34% of baseline values, the WBC decrease in TCL1 mice was more dramatic dropping to 8% of baseline values ( Figure 5D). While hemoglobin counts did not change significantly ( Figure 5E) platelet counts of LDC526treated mice were decreasing. In contrast to platelet counts of LDC526-treated wild-type mice, TCL1 platelet counts were even rising to significantly higher levels (day +11; p<0.01) than measured at baseline ( Figure 5F).
Peripheral blood CLL and non-malignant T and B lymphocyte subsets were defined by flow cytometry using antibodies against CD3, CD5 and CD19 ( Figure 6A). Two days after the last LDC526 application TCL1 CLL cell concentrations decreased while CLL concentrations of TCL1 control mice did not change ( Figure 6B). Peripheral blood CLL counts of individual LDC526-treated TCL1 mice decreased at least 8-fold (Supplementary Figure 2). Remarkably, mean TCL1 CLL numbers decreased 16-fold compared to baseline levels demonstrating a striking antileukemic effect of LDC526 ( Figure 6C). In concordance with TCL1 CLL cells, TCL1 blood T cell concentrations also dropped ( Figure 6D). Compared to the 16-fold decline of TCL1 CLL cells the decrease of T cells was not as dramatic (6-fold). However, in contrast to TCL1 T cells the concentration of wild-type T cells only halved by LDC526 treatment (Figure 6E).
We also determined the absolute non-malignant B cell numbers within the peripheral blood. While nonmalignant B cell numbers of carrier control-treated www.oncotarget.com TCL1 mice remained constant, LDC526 treatment led to a substantial decline of this lymphocyte population ( Figure 6F). Non-malignant B cell numbers of LDC526treated TCL1 mice almost dropped by two log levels to 1.4% of baseline values ( Figure 6G). B cells of LDC526treated wild-type mice also dropped dramatically to 3.7% of baseline levels. As expected, weight loss of LDC526-treated TCL1 mice was more pronounced than the LDC526-induced weight-loss of wild-type mice. All but one LDC526-treated TCL1 mouse, which had to be euthanized due to reduced clinical condition on day +7, regained their weight by day +8 (Supplementary Figure  1B). As the LDC526-associated weight loss was more severe in TCL1 transgenic mice than in wild-type control mice weight loss was at least in part associated with treatment response. In summary, LDC526 displayed an efficient cytotoxic effect against peripheral blood TCL1 CLL cells. Additionally, non-malignant T and B cell numbers also declined after LDC526 treatment.
On day +11 ( Figure 5A) the experiment was terminated as planned and spleens of the experimental mice were harvested. To be able to assess the impact of LDC526 on wild-type spleens we included n=3 of non-treated age-matched wild-type littermate mice as an additional experimental control group in the final spleen analysis. A caveat of this final spleen analysis posed the fact that in contrast to the longitudinal blood analysis, baseline spleen data of individual aged TCL1 mice were not available. As longitudinal analysis was not possible, we now compared spleen parameters between cohorts. Importantly, it is unknown to which extent the amount of circulating peripheral blood TCL1 CLL cells determined at baseline correlates with splenomegaly. Nonetheless, TCL1 spleens of mice receiving LDC526 were macroscopically smaller than spleens of carriertreated TCL1 mice (Supplementary Figure 3A). Consistent with this, splenic weights of LDC-treated TCL mice were significantly lower than the splenic weights of TCL1 mice receiving carrier control treatment suggesting a therapeutic effect (Supplementary Figure 3B). There was also a trend (p=0.09) towards a lower splenic absolute CLL cell content (Supplementary Figure 3C). Here, it has to be taken into account that LDC526 treatment was completed 10 days before the spleen analysis. Thus, as observed in the peripheral blood where CLL numbers increased again by 29% on day +11 compared to day +3 ( Figure  6C) also splenic CLL cell numbers might have recovered during this period. In concordance with the peripheral blood data, TCL1 splenic T cells after LDC526 treatment were significantly decreased compared to TCL1 splenic T cells of carrier control treated mice (Supplementary Figure 3D). The LDC526 B cell cytotoxicity observed in the longitudinal blood analysis was also apparent when comparing absolute splenic B cell numbers of wild-type mice with and without LDC526 treatment (Supplementary Figure 3E). Taken together, spleens of LDC526-teated TCL1 mice were smaller than control TCL1 spleens implying a therapeutic benefit. The data also suggest that LDC526 treatment was capable of reducing TCL1 T cell numbers.

Residual TCL1 CLL cells after LDC526 treatment exhibit high levels of Bcl-2
Inhibition of CDK9 activity prevents RNA transcriptional elongation and deprives cells from the short-lived antiapoptotic protein MCL-1 leading to apoptosis [37]. Given the redundancy between the antiapoptotic proteins MCL-1 and BCL-2 we intended to study the effect of LDC526 on longitudinal cellular BCL-2 levels. We measured Bcl-2 protein levels in TCL1 lymphocyte subsets by flow cytometric analysis on days -5 and +3 ( Figure 7A) of the experimental set-up outlined in   Figure 5A. As mentioned before, peripheral blood TCL1 CLL cells on day +3 were diminished to 6.4% of baseline levels by LDC526 treatment ( Figure 6C). Remarkably, residual CLL cells exhibited significantly increased intracellular Bcl-2 levels after LDC526 exposure. Moreover, T cells that were decreased in numbers by LDC526 treatment also displayed significantly higher Bcl-2 levels compared to levels measured before LDC526 application ( Figure 7B). We speculate that inherent higher cellular Bcl-2 levels of a subset of TCL1 CLL and T cells might have protected these cells from apoptosis by LDC526-mediated MCL-1 loss. Alternatively, LDC526 treatment might have induced upregulation of Bcl-2 by a so far unknown mechanism in residual cells.

DISCUSSION
An increasing number of studies support the notion that pan-CDK inhibitors such as flavopiridol (Alvocidip) exert their primary anti-leukemic CLL activity by interfering with transcription via CDK9 inhibition rather than inhibiting the cell cycle [18,38,39]. Because pan-CDK inhibitors were associated with a narrow therapeutic window, it is anticipated that specific CDK9 inhibition would selectively inhibit transcription and not affect cell cycle progression and might therefore exhibit a more favorable safety profile. Compared to other CDK inhibitors such as flavopiridol and SNS-032 our novel CDK9 inhibitor LDC526 exhibited improved selectivity

Figure 7: Residual CLL cells and T cells post LDC526 treatment express high levels of BCL-2. (A) Representative histograms of intracellular anti-Bcl-2 and corresponding isotype control staining of TCL1 CLL cells (CD5+CD19+) and T cells (CD5+CD19-). (B)
Graphs comparing the CLL and T cell intracellular Bcl-2 expression of individual mice on days -5 and +3. The Bcl-2 expression was quantified as median fluorescence intensity ratio (MFI ratio: MFI anti-Bcl-2/MFI isotype control antibody). * p<0.05. www.oncotarget.com for CDK9. Meanwhile, LDC526 was optimized further in terms of pharmacological parameters resulting in BAY1143572 [27], which has been studied in phase I trials in patients with acute leukemia and solid tumors / lymphomas (ClinicalTrials.gov, Identifier NCT02345382 and NCT01938638, respectively).
Here, we first studied the anti-CLL activity of LDC526 in vitro to provide the basis for subsequent in vivo studies carried out in the CLL NSG xenograft and the TCL1 transgenic CLL mouse model. Recently, other groups described other CDK9 inhibitors with improved CDK9 selectivity profiles such as CDKI-73 and LY2857585 [40][41][42]. The biochemical LY2857585 selectivity for CDK9 versus for example CDK7 was lower (22-fold; [42]) than the LDC526 CDK9 selectivity versus CDK7 (52-fold) described here. The degree of ex vivo CLL cell killing by LDC526 in the sub-micromolar range was comparable to the CLL cell killing achieved by LY2857585 and CDKI-73 [40][41][42].
Next, we moved on to in vivo testing of LDC526 in two independent mouse models of CLL. To our knowledge, there are only two other studies using the complementary CLL NSG xenograft and the TCL1 transgenic CLL mouse in preclinical drug testing [43,44]. This is the first study investigating the impact of a specific CDK9 inhibitor in CLL in vivo models. The use of in vivo models has the advantage of providing evidence whether the interaction of CLL and stroma influences the response to CDK9 inhibition. This is of particular importance since the interaction of CLL with stroma in vivo leads to increased CDK9 activity and elevated MCL-1 expression in CLL cells [45,46]. Here, we demonstrated that LDC526 has dose-dependent in vivo anti-CLL activity in the NSG xenograft model. The insufficient LDC526 anti-CLL activity of the 50 mg/kg dose might well be explained by prevailing high stroma-induced expression levels of antiapoptotic proteins by CLL cells in the NSG spleen, which were not present in the in vitro setting.
In contrast to human CLL cells, the lower LDC526 50 mg/kg dose was already capable of inducing significant anti-human T cell activity within the NSG spleen. Altogether, the LDC526 anti-human T cell activity appeared more pronounced than the anti-CLL activity in the NSG spleen environment. This might be due to fact that activated T cells involved in an alloreactive (in the case of NSG, xenoreactive) response were shown to be especially sensitive to CDK9 inhibition [47,48]. This is consistent with the upregulation of CDK9 upon T cell activation [49]. Since the proliferation of CLL cells was shown to be dependent on the presence of T cells in NSG spleens [29], the LDC526-mediated T cell reduction might have concomitantly contributed to the decreased splenic CLL cell numbers. However, it should be noted that residual T cells after 5 days of LDC526 therapy were primarily CD4+ T cells, which were shown to be indispensable for CLL proliferation in NSG spleens [29]. TCL1 mice mirror the course of human CLL as aged mice develop the clonal accumulation of CD5+CD19+ B cells in the blood and hematopoietic organs. Importantly, TCL1 mice display a response to CLL treatment analogous to human patients [32]. As disease onset and progression can be variable in TCL1 mice, we carried out a longitudinal study focusing on the course of peripheral blood CLL cell counts before and after LDC526 treatment. In parallel, we determined standard blood counts. In contrast to mean WBC dropping 12-fold in TCL1 mice after LDC526 treatment, mean platelet concentrations did not decrease below 50% of baseline levels. The decrease in platelet numbers was expected as CDK9/pTEFb is involved in megakaryocyte maturation [50]. Possibly the platelet drop turned out relatively subtle because other CDKs involved in megakaryocyte differentiation and functions were not targeted by LDC526. Treating TCL1 mice with LDC526 resulted in a reduction of blood CLL cells to 6.4% of baseline values, which is a striking response for a twoday oral therapy. Comparability between the LDC526 CLL responses in TCL1 blood versus the NSG spleen is limited. Besides the species disparities between CLL cells studied, the difference observed in the magnitude of CLL response obtained in blood (TCL1) versus spleen (NSG) could be explained by the micro-environmental pro-survival signals CLL cells are exposed to in the NSG spleen which are not provided in the TCL1 circulation. In the NSG spleen pro-survival and proliferation signals are presumably delivered to human CLL cells within patientderived follicular structures that were described to be surrounded by human T cells [29].
Our study demonstrated in vivo cytotoxicity of non-malignant T cells towards specific CDK9 inhibition in the murine as well in the human CLL NSG system. Interestingly, TCL1 T cells exhibited a higher degree of LDC526 sensitivity than T cells of wild-type control mice without CLL. Others and we have shown skewing of the T cell compartment in CLL patients and in TCL1 mice towards an effector memory phenotype [51][52][53]. This might explain the higher LDC526 sensitivity of TCL1 CLL T cells compared to controls since the effector memory T cell subpopulations were shown to upregulate CDK9 and Cyclin T1 [49]. Moreover, LDC526 targeting of stromal non-malignant T cells exerting a CLL-supportive impact might have indirectly contributed to the LDC526 in vivo anti-CLL effect. On the other hand CD8+ T cells even though mainly dysfunctional in the CLL setting [51,52,54] might still exert some anti-tumoral effects that would be depleted by LDC526 therapy. Furthermore, an increased rate of infectious complications due to decreased lymphocyte counts would be expected to emerge under LDC526 treatment.
Recently, direct in vivo cytotoxicity of the selective CDK9 inhibitor BAY1143572 against malignant T cells (adult T cell leukemia/lymphoma, ATL) was shown. BAY1143572 also exerted a cytotoxic effect against normal CD4+ T cells in vitro, which was not as prominent as the effect against ATL cells [27].
Remarkably, we demonstrated that LDC526 has a cytotoxic effect on non-malignant B cells in vivo. This effect involved normal B cells of diseased CLL TCL1 mice as well as circulating and splenic B cells of wild-type mice. LDC526 induced MCL-1 loss is likely responsible for the depletion of B cells since MCL-1 is required for the survival of most normal B cell subsets [55]. However, the LDC526-induced B cell loss was not complete since circulating B cell numbers were rising 10 days post the cessation of LDC526 treatment and splenic B cell numbers of LDC526-treated wild-type mice corresponded to 31% of splenic B cell numbers of untreated mice.
We demonstrated in TCL1 CLL mice that residual CLL cells post LDC526 treatment exhibited higher Bcl-2 levels than CLL cells before treatment. Interestingly, others described a similar upregulation of BCL-2 in CLL cells post CDK inhibitor exposure in vitro (24 hours) and patient treatment (100 mg/m 2 ) with the pan-CDK inhibitor SNS-032 [18,21]. The mechanism of how CDK9 inhibition might lead to higher cellular BCL-2 levels remains elusive. However, as transcription is suppressed by CDK9 inhibition upregulated BCL-2 mRNA expression appears an unlikely cause. Instead, CLL cells with a priori higher BCL-2 levels might exhibit a selective advantage to survive MCL-1 depletion by CDK9 inhibitor treatment. Remarkably, higher BCL-2 levels within lymphoma cells were shown to correlate with an increased sensitivity to venetoclax treatment [56]. Therefore, we think high BCL-2 levels in residual CLL cells after CDK9 inhibitor therapy represent a rationale for combining CDK9 inhibitor treatment with a BCL-2 inhibitor such as venetoclax to achieve deeper CLL remissions or potentially cure. In line with this approach, CDK inhibition with flavopiridol or dinaciclib synergized with venetoclax to induce apoptosis of lymphoma cell lines in vitro [57,58].
In conclusion, we describe the novel specific CDK9 inhibitor LDC526 and demonstrate effective LDC526 anti-leukemic activity in two independent preclinical CLL mouse models. Residual CLL cells after LDC526 treatment were characterized by higher BCL-2 levels. Therefore, specific CDK9 inhibitors are primary candidates for combination therapy with BCL-2-inhibitors to further deepen and prolong CLL treatment responses.

Cells and cell culture
The MEC-1 cell line was obtained from the DSMZ (Braunschweig, Germany) and was cultured in RPMI (10% fetal bovine serum). A LDC526 stock solution was prepared with DMSO, diluted and added to the cultures to reach the specified concentrations. Samples (percentage of CD19+CD5+ CLL cells >90%) of n=37 CLL patients (

Mice
All mice were bred and housed at the University Hospital Essen animal care facility. TCL1 mice [31] were back-crossed more than 6 generations into the C57BL/6 background and were genotyped as previously described [51]. Age-matched TCL1 wild-type littermates served as controls. Peripheral blood counts were determined with the VetABC analyzer (scil, Viernheim, Germany). Blood smears were Wright-Giemsa stained. For the human CLL xenograft experiments eight to 14-week-old NOD. Cg-Prkdc scid Il2rg tm1Wjl /SzJ (NOD/scid/γc null ; NSG) mice were sublethally irradiated (3.25 Gy) 24 hours before transplantation. Freshly isolated 1.0 x 10 8 PBMC suspended in 0.2 ml of RPMI were intravenously transplanted as previously described [30]. For oral administration, LDC526 was diluted in polyethelene glycol (PEG M n 400, Sigma-Aldrich) and administerd by gavage. Animal experiments were performed in accordance with institutional guidelines approved by the Animal Care Committee of the University Hospital Essen.

In vitro enzymatic kinase assay for CDKs
IC50 values for CDK inhibitors were determined using the fluorescence resonance energy transfer (FRET)based LANCE Ultra KinaSelect Ser/Thr kit (Perkin Elmer). Kinase activity and inhibition were measured according to the manufacturer's instructions and as previously described [59]. Briefly, a specific ULight MBP peptide substrate (50 nM final concentration) was allowed to get phosphorylated by a CDK-cyclin pair in enzymatic buffer (50 mM Hepes pH 7.5, 10 mM MgCl2, 1 mM EGTA, 2 mM dithiothreitol) containing ATP at the concentration of the Km values of the individual kinases for 1 hour at room temperature. Subsequently, phosporylation was detected by addition of specific europium (Eu)-labeled anti-phospho-antibodies (2 nM), which upon binding to the phospho-peptide give rise to a FRET signal. FRET signals were recorded in a timeresolved manner in a Perkin Elmer EnVision reader. Purified cyclin-kinase pairs were obtained from the following suppliers: Carna Biosciences (CDK1-Cyclin B1, CDK6-Cyclin D3, CDK7-Cyclin H-MAT1), ProQinase (CDK2-Cyclin A) and Invitrogen (CDK9-Cyclin T1). Broad selectivity kinase in vitro profiling with LDC526 was carried out using n=219 recombinant kinases (Millipore/Merck, Darmstadt, Germany).

Imaging
Macroscopic spleen images were taken with a Canon IXUS 1100 HS digital camera (Canon, Krefeld, Germany). Images of stained blood smears of LDC526 treated mice were taken through the 10x and 50x objective lenses of an Axioskope microscope (Zeiss, Jena, Germany).

Statistical analyses
Data were analyzed with GraphPad Prism 7 software. Differences were assessed by unpaired or paired Student t tests if not indicated otherwise. Two-sided probabilities of less than 0.05 were considered significant. Nonlinear regression sigmoidal dose-response curve fit was used to determine the IC 50 values and to generate graphs. All data are displayed as mean±SEM.

ACKNOWLEDGMENTS
We thank the staff of the central animal facility (ZTL) of the University Hospital of Essen. We would like to thank Carlo Croce for his permission to use TCL1 transgenic mice.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.

FUNDING
JRG received grant support from the Stem Cell network of North-Rhine-Westphalia. The activities of the LDC have been co-funded by the Max-Planck Foundation,