Microenvironmental Influence on Pre-Clinical Activity of Polo-Like Kinase Inhibition in Multiple Myeloma: Implications for Clinical Translation

Polo-like kinases (PLKs) play an important role in cell cycle progression, checkpoint control and mitosis. The high mitotic index and chromosomal instability of advanced cancers suggest that PLK inhibitors may be an attractive therapeutic option for presently incurable advanced neoplasias with systemic involvement, such as multiple myeloma (MM). We studied the PLK 1, 2, 3 inhibitor BI 2536 and observed potent (IC50<40 nM) and rapid (commitment to cell death <24 hrs) in vitro activity against MM cells in isolation, as well as in vivo activity against a traditional subcutaneous xenograft mouse model. Tumor cells in MM patients, however, don't exist in isolation, but reside in and interact with the bone microenvironment. Therefore conventional in vitro and in vivo preclinical assays don't take into account how interactions between MM cells and the bone microenvironment can potentially confer drug resistance. To probe this question, we performed tumor cell compartment-specific bioluminescence imaging assays to compare the preclinical anti-MM activity of BI 2536 in vitro in the presence vs. absence of stromal cells or osteoclasts. We observed that the presence of these bone marrow non-malignant cells led to decreased anti-MM activity of BI 2536. We further validated these results in an orthotopic in vivo mouse model of diffuse MM bone lesions where tumor cells interact with non-malignant cells of the bone microenvironment. We again observed that BI 2536 had decreased activity in this in vivo model of tumor-bone microenvironment interactions highlighting that, despite BI 2536's promising activity in conventional assays, its lack of activity in microenvironmental models raises concerns for its clinical development for MM. More broadly, preclinical drug testing in the absence of relevant tumor microenvironment interactions may overestimate potential clinical activity, thus explaining at least in part the gap between preclinical vs. clinical efficacy in MM and other cancers.


Introduction
One of the problems in oncology drug development today is the discordance of highly promising in vitro and in vivo preclinical results with the lack of efficacy observed in patients. Less than 8% of agents that enter phase I clinical trials in cancer eventually become FDA approved for the treatment of any tumor type [1]. Data from our group and others indicate that interaction of malignant cells with their local microenvironment can confer drug resistance, which may account for this gap between the preclinical drug activity and clinical efficacy [2,3].
Polo-like kinases (PLKs) are particularly interesting targets in cancer because of their role in cell cycle progression, checkpoint control and mitosis [4,5]. Tumors with PLK overexpression are associated more frequently with chromosomal instability, DNA aneuploidy and centrosome amplification [6]. In addition, cancer cells are more sensitive to PLK inhibition than normal cells [7]; and PLK expression has been shown to be higher in cancer cells with a high mitotic index [8]. In advanced multiple myeloma (MM), malignant cells have a high mitotic index [9] and chromosomal instability [10], suggesting that PLK inhibitors may be an attractive therapeutic option for this presently incurable disease.
Here we evaluate the activity of the PLK 1, 2, 3 inhibitor BI 2536 in preclinical models of MM and investigate the role of the microenvironment in modulating its anti-MM activity. We observed potent anti-MM activity in traditional drug development experiments, but decreased activity of BI 2536 in bone microenvironmental models. Our results highlight that BI 2536 represents a compound with promising characteristics, but its lack of activity in microenvironmental models of MM raises concerns for its clinical development for this disease. These concerns are compatible with the limited clinical activity that this agent has shown so far in clinical trials in solid tumors, even though clinically achievable concentrations exceed the levels needed for in vitro antitumor activity based on conventional models. These models can overestimate the drug activity because they do not incorporate tumor-microenvironment interactions. More broadly, our study provides a concrete example of the importance of preclinical testing of investigational therapeutics in models that simulate how the non-malignant accessory cells of the tumor microenvironment may confer drug resistance.

Results and Discussion
Anti-MM activity of PLK inhibitor in vitro and in vivo in the absence of bone microenvironmental interactions Because of the activity of BI 2536 in other cancer models and the role of PLKs in cell cycle regulation [4,5], we evaluated a panel of MM cell lines for sensitivity to BI 2536 (Fig. 1a). We observed potent activity with IC 50 values ,40 nM for the majority of cell lines, including lines resistant to established anti-MM agents (e.g. dexamethasone-resistant MM.1R). The BI 2536 concentrations required for in vitro anti-MM activity are within its clinically achievable levels [11]. These in vitro responses were rapid, requiring ,24 hrs of drug exposure to commit cells to death (Fig. 1b). Nonmalignant cells, such as HS-5 stromal cells, THLE-3 hepatocytes and osteoclasts (OCs) had IC 50 values .40 nM (Fig. 1c). The potency and rapid kinetics of BI 2536 activity are also highlighted by the cell cycle analysis (G2/M arrest, followed by increase sub-G0 events; Fig. 1d), rapid cleavage of caspase 3 and PARP within 8 hrs of treatment (Fig. 1e) and formation of monopolar asters (Fig. 1f). In a subcutaneous animal model, BI 2536 significantly suppressed tumor burden (Fig. 1g) and prolonged survival (Fig. 1h), without changes in body weight (Fig. 1i). These preclinical in vitro and in vivo data are similar to those that provided in the past the framework for further clinical development of BI 2536 in other neoplasias. These experiments in other models, however, did not take into account the ability of the bone microenvironment to modulate drug activity.

Effects of the bone microenvironment on PLK inhibitor activity in vitro and in vivo
Emerging literature indicates that the activity of different anticancer agents can be modulated by the non-malignant accessory cells of the microenvironment in which tumor cells are located [2]. We therefore recently developed an in vitro model, tumor cell compartment-specific bioluminescence imaging (CS-BLI), which allows high-throughput scalable evaluation of investigational agents in co-cultures simulating the tumor-stromal interactions [2]. This allows us to identify compounds with increased, as well as others with deceased, antitumor activity in the presence of stromal cells [2]. Importantly, we used of this in vitro co-culture platform in tandem with our orthotopic animal models and validated in vivo these observations in conditions which reflect the tumor microenvironment and simulate the patient condition [2].
Based on this experience, we assessed the effect of the bone microenvironment on the anti-MM activity of BI 2536. We first tested the compound in vitro against MM cell lines cultured alone or in the presence HS-5 stromal cells, differentiated OCs or primary bone marrow stromal cells from MM patients. Drug activity was selectively assessed using the CS-BLI assay. BI 2536 was less active against MM.1S (Fig. 2a) In addition to our in vitro microenvironment assessment of BI 2536, we tested the compound in our in vivo orthotopic model of diffuse myeloma lesions [12]. In this model, tumor cells home to the bone in multiple sites, consistent with the clinical presentation of myeloma in patients. We observed no significant activity of BI 2536 when administered at 50 mg/kg 2x weekly (Fig. 2g) and no significant prolongation of survival in treated mice vs. vehicle treated controls (Fig. 2h). Although this dose was higher than in the s.c. model, we still did not observe significant toxicity, as evidenced by the lack of weight loss in the BI 2536 treated mice compared to vehicle-treated controls (Fig. 2i). The MM.1S cell line model utilized in our in vivo studies was one of the most drug-responsive cell lines in vitro, however, its in vivo response in the s.c. model, but not the diffuse bone model, illustrates that tumor cells highly sensitive to a given drug in the absence of the bone microenvironment may become more drug-resistant in its presence.
One of the biggest problems in oncology remains the low success rate of translating preclinical results to effective clinical drugs. Among the various reasons for this discrepancy, our data suggest that the tumor microenvironment plays a role in decreasing anti-cancer drug efficacy. PLK inhibitors such as BI 2536, although active against tumors studied outside the context of their bone microenvironment, are less active in models that take this interaction into account. The application of both in vitro and in vivo models that simulate how tumors interact with the microenvironment in patients would improve the prioritization of PLKs inhibitors, and other compounds, for further development. Utilizing orthotopic in vivo models, which include all the elements of the local tumor microenvironment, in conjunction with co-culture in vitro screening offers a better simulation of clinical condition than conventional models. Relying exclusively on in vitro coculture assays, especially if supra-pharmacological doses and nonrelevant durations of exposure are used, may over-estimate the ability of a given drug to overcome microenvironment-dependent drug resistance. This may explain why prior studies using BI5236 at 2.5 mM (a dose higher than the Cmax in patients) did not detect in vitro stroma-induced resistance of tumor cells [13].
Interestingly, multiple clinical trials of BI 2536 in solid tumors [11,14,15,16] have shown limited objective clinical responses in clinical trials, despite achievable levels of BI 2536 far exceeding those showing in vitro activity in solid tumor models. This raises the possibility that, similarly to our observations in MM models, solid tumor cells may also exhibit a microenvironment-dependent resistance to BI 2536. Furthermore, other clinical trials of agents in the same class have shown limited clinical activity as well [17], suggesting that the lack of clinical activity so far for this drug class could be attributable in part to microenvironmental factors, although further studies are warranted.
Although we observed microenvironment resistance to BI 2536 in preclinical models, there may be subpopulations of MM patients who may respond to this drug despite these microenvironment interactions. To evaluate this we examined the gene expression data of myeloma patients from the Dutch HOVON trial (N = 320) for the expression of PLK transcripts. We observed that PLK1 (201429_at) is expressed at higher levels in patients compared to the other PLK isoforms (Fig. 3a). Comparing PLK1 transcript (202240_at) in various stages of disease, we observe a significantly higher expression level in plasma cell leukemia (PCL) patients compared to individuals with the premalignant monoclonal gammopathy of undetermined significance (MGUS) (P,0.05) and a trend compared to MM patients, although not statistically significant (Fig. 3b). Interestingly, we observed significantly higher PLK1 transcript (37228_at) expression in the proliferative (PR) subtype of MM compared to CD1 (cases with CCND1 upregulation because of t(11;14) translocation), CD2 (cases with CCND3 upregulation because of t(6;14) translocation), HY (hyperdiploid MM), LB (''Low Bone'' disease group, characterized by low number of magnetic resonance imaging (MRI)-defined focal bone lesions and low expression of DKK1), and MS (cases with MMSET overexpression due to t(4;1;4) translocation) subtypes (1-way ANOVA; p,0.0001; Dunn's multiple comparison post-hoc tests for each of the comparison between subtypes, p,0.05), but not compared to the MF subtype (cases with overexpression of MAF/ MAFB) of MM patients (p.0.05; Fig. 3c). The PR subtype of MM is the one that most closely resembles the molecular profiles of MM cell lines [18], such as those evaluated in our study. This observation suggests that PLK inhibition may still be an interesting putative therapeutic approach for MM patients harboring a proliferative transcriptional signature in their tumor cells. However, our observations that BI 2536 is subject to microenvironment-dependent drug resistance suggest that caution is warranted for further clinical development in MM, even for patients of the PR subtype. Preclinical models currently applied in most cancers don't take into account tumor microenvironmental interactions, and therefore may overestimate the clinical activity of investigational agents. Incorporating the microenvironmental component into preclinical studies may reconcile the discordance between preclinical and clinical efficacy and thereby improve the bench to bedside translation of effective therapies. In conclusion, BI 2536 is a less attractive agent for development in MM based on our microenvironment data, unless further studies determine clinically applicable methods to overcome this microenvironment-dependant resistance.

Materials and Methods
Compound BI 2536 (CAS 755038-02-9, C 28 H 39 N 7 O 3 , MW 521.66) was synthesized by published methods [19], characterized by 1 H and 13 C nuclear magnetic resonance, solubilized in DMSO (in vitro) or 0.1N HCl/0.9% NaCl (in vivo) and used as indicated in figure legends.