Differential Targeting of Gr-MDSCs, T Cells and Prostate Cancer Cells by Dactolisib and Dasatinib

Granulocytic myeloid-derived suppressor cells (Gr-MDSCs) promote immune evasion and resistance to immunotherapeutics in a variety of malignancies. Our previous study showed that dual PI3K/mTOR inhibitor Dactolisib impaired the viability and immunosuppressive function of Gr-MDSCs, and significantly synergized with immune checkpoint blockade (ICB) antibodies targeting PD1 and CTLA4 to eradicate metastatic castration-resistant prostate cancer (CRPC) in a preclinical transgenic mouse model. On the contrary, tyrosine kinase inhibitor Dasatinib diminished tumor-infiltrating T lymphocytes and showed no synergic activity with ICB. The understanding of the distinct effects of Dactolisib and Dasatinib on Gr-MDSCs, T cells and prostate neoplastic cells is inadequate, limiting the clinical translation of the combination immunotherapy. To address this question, we applied Reverse Phase Protein Array (RPPA) to profile 297 proteins and protein phosphorylation sites of Gr-MDSCs, T cells and prostate cancer cells isolated from the CRPC model. We found cell type-specific protein expression patterns and highly selective targets by the two drugs, including preferential inhibition of phospho-4E-BP1 in Gr-MDSCs by Dactolisib and preferential suppression of phospho-Src and phospho-p38 MAPK in T cells. Furthermore, transcriptomic profiling of Gr-MDSCs treated with the two inhibitors revealed downregulation of mitochondrial respiration pathways by Dactolisib but not Dasatinib. Overall, these results provide important mechanistic insight into the efficacious combination of Dactolisib and ICB as well as the detrimental effect of Dasatinib on anti-tumor immunity.


Distinct Protein Expression Pattern by PCa Cells, T Cells and Gr-MDSCs in a Mouse CRPC Model
In the same procedure as we reported [16], we induced CRPC formation in CPPSML model by surgically castrating CPPSML males when prostate tumors reached 150 mm 3 measured by magnetic resonance imaging, followed by feeding the mice with an enzalutamide-admixed diet for 4 weeks. At this stage, the mice were euthanized and the prostate tumors were dissected and digested for isolation of primary PCa cells using fluorescence-activated cell sorting (FACS) of GFP + CD45 − cells, or isolation of tumor-infiltrating Gr-MDSCs using magnetic-activated cell sorting (MACS) of CD11b + Ly6G + Ly6C low cells. From the same mice, total T cells were isolated from the spleen using MACS. PCa cells were cultured for 2-3 passages as adherent primary cells before inhibitor treatment, whereas Gr-MDSCs and T cells were treated immediately after isolation to maximize survival. Cells were treated with DMSO (control), Dactolisib or Dasatinib at various concentrations for 2 h before harvest for the RPPA workflow ( Figure 1A, Supplementary Table S1). Unsupervised clustering of the log2 transformed RPPA signals of untreated or DMSO-treated cell samples ( Enrichment of phospho-Akt and phospho-S6 in PCa cells is consistent with the prostate-specific deletion of Pten in the CPPSML model.

Differential Effect of Dactolisib and Dasatinib on Protein Phosphorylation in PCa Cells, T cells and Gr-MDSCs
To identify the treatment effect of Dactolisib and Dasatinib on specific proteins or PTMs in the three cell types, we generated supervised clustering trees of the normalized RPPA data of the control (untreated and DMSO-treated) and inhibitor-treated (Dactolisib or Dasatinib) samples. Dysregulated targets emerged in cell type-specific and inhibitor-specific manners. For Gr-MDSCs, among the 297 unique antibodies, only phospho-4E-BP1 (Thr37/46) was consistently decreased by Dactolisib, but it was unaffected by Dasatinib (Figure 2A), consistent with the biological effect of Dactolisib on inhibiting PI3K/mTOR signaling. In a similar manner, Ser235/236 and Ser240/244 phosphorylation of S6 in PCa cells were downregulated by Dactolisib but remained unchanged under Dasatinib treatment ( Figure 2B). By contrast, phospho-Src (Tyr527) was only downregulated by Dasatinib in PCa cells ( Figure 2B), corresponding to the activity of Dasatinib as a Src inhibitor [19]. For T cells, Dasatinib, but not Dactolisib, effectively downregulated phospho-Src (Tyr416 and Tyr527) and phospho-p38 MAPK (Thr180/182) ( Figure 2C). Together, these results demonstrate that with a short time treatment (2 h), Dactolisib and Dasatinib attenuated specific signaling targets of the PI3K/mTOR or Src/MAPK pathway, respectively. It is interesting to note that the majority of the surveyed proteins and PTMs remained largely unchanged by the inhibitors in all three cell types.

Cell Type-Specific Downregulation of Phosphorylation Levels of S6, 4E-BP1, Src and p38 MAPK by Dactolisib and Dasatinib
In order to compare the dose-dependent effect of Dactolisib and Dasatinib on different cell types, we calculated the ratio of phosphorylated over total protein RPPA signal intensity for S6, 4E-BP1, Src and p38 MAPK (identified above as differentially regulated by the inhibitors) and plotted for each cell type at the tested concentrations ( Figure 3). For each cell type, the ratio at different inhibitor concentrations was normalized to DMSO control (0 µM) for each cell type and expressed as percentages to help show the relative changes ( Figure 3). MDSC/T = 16.6, Gr-MDSC/PCa = 8.89). T cells highly express Histone-H3 and Lck compared with Gr-MDSCs (T/Gr-MDSC for Histone-H3 = 3.04, T/Gr-MDSC for Lck = 2.68) and PCa cells (T/PCa for Histone-H3 = 7.76, T/PCa for Lck = 7.01). PCa cells have a higher abundance of a number of proteins and PTMs compared with Gr-MDSCs and T cells, such as E-cadherin, Connexin-43, ARID1A, eIF4G, HES1, phospho-Akt (Ser473), phospho-S6 (Ser235/236, Ser240/244), Phospho-NDRG1 (Thr346), phospho-YAP (Ser127), and TAZ (fold change > 5). Enrichment of phospho-Akt and phospho-S6 in PCa cells is consistent with the prostate-specific deletion of Pten in the CPPSML model. For the ratio of phospho-S6 / total S6, Dactolisib dramatically decreased phosphorylation of Ser235/236 (100% at 0 µM, 10.5% at 0.1 µM) and Ser240/244 (100% at 0 µM, 20% at 0.1 µM) only in PCa cells, but not in T cells or Gr-MDSCs ( Figure 3A). On the other hand, Dasatinib caused little effect on phospho-S6 in either cell type. The higher basal level of phospho-S6 and higher sensitivity to Dactolisib in PCa cells compared with T cells or Gr-MDSCs is consistent with the Pten-deficient status of PCa cells.
To corroborate the results from RPPA, we detected the dysregulated proteins and PTMs by western blot. Dactolisib strongly inhibited S6 phosphorylation at Ser235/236 and Ser240/244 in Gr-MDSCs and PCa cells ( Figure 4A,B). S6 and phospho-S6 were barely detectable in T cells ( Figure 4C). Dactolisib also suppressed 4E-BP1 phosphorylation at Thr37/46 and Ser65 in Gr-MDSCs and PCa cells ( Figure 4A,B), yet 4E-BP1 phosphorylation in T cells was largely resistant to Dactolisib treatment. In contrast to Dactolisib, the effect of Dasatinib on phosphorylation of S6 or 4E-BP1 in all three cell types was much weaker or essentially absent ( Figure 4A-C). However, Dasatinib (but not Dactolisib) inhibited p38 MAPK phosphorylation at Thr180/182, Src phosphorylation at both Tyr416 and Try527 and Lck phosphorylation at both Tyr394 and Tyr505, preferentially in T cells ( Figure 4C). As an internal control, CD3ε remained unchanged in T cells by the drug treatments ( Figure 4C). It is worth noting that the moderate inhibitory effect of Dasatinib on Src phosphorylation in Gr-MDSCs and PCa cells was only observed for Tyr527 but not Tyr416 ( Figure 4A,B). Overall, these results validated the RPPA results and supported (1) the selective downregulation of phospho-S6 and phospho-4E-BP1 by Dactolisib in Gr-MDSCs and PCa cells, and (2) the preferential downregulation of phospho-Src and phospho-p38 MAPK by Dasatinib in T cells.
In order to compare the dose-dependent effect of Dactolisib and Dasatinib on different cell types, we calculated the ratio of phosphorylated over total protein RPPA signal intensity for S6, 4E-BP1, Src and p38 MAPK (identified above as differentially regulated by the inhibitors) and plotted for each cell type at the tested concentrations ( Figure 3). For each cell type, the ratio at different inhibitor concentrations was normalized to DMSO control (0 μM) for each cell type and expressed as percentages to help show the relative changes ( Figure 3).

Discussion
Our study interrogates the differential effect of Dactolisib and Dasatinib on PCa cells, T cells and Gr-MDSCs isolated from the same mouse primary CRPC. Employing the power of RPPA, this study reveals the unique protein expression patterns of the three cell types. Some of the features are expected, such as higher Lck in T cells and higher E-cadherin, phospho-Akt and phospho-S6 in PCa cells. The dramatically higher P-Cadherin level in Gr-MDSCs compared with T cells and PCa cells is surprising and intriguing. As a calcium-dependent cell-cell adhesion glycoprotein, P-Cadherin shares about 67% of homology with E-Cadherin, yet remains far less characterized [20]. P-Cadherin is expressed by myoepithelial/basal cells in a heterogeneous manner in organs like the basal layer of the epidermis, breast, prostate, and part of the digestive tract [20]. Normal hematopoietic cells are not known to express P-Cadherin. Given that Gr-MDSCs massively dysregulate gene expression relative to normal neutrophils [21,22], it will be important to investigate whether human Gr-MDSCs also upregulate P-Cadherin and the role of P-Cadherin in the biological function of Gr-MSDCs. Once validated, P-Cadherin may be a new surface marker and therapeutic target for Gr-MDSCs. P-Cadherin + cells can be targeted for killing using P-Cadherin-targeted radioimmunotherapeutic FF-21101( 90 Y) [23] or antibody-drug conjugates [24].
Upon the 2 h drug treatment, PCa cells, T cells and Gr-MDSCs all exhibit rather limited changes of protein levels and protein phosphorylation levels. This result can be accounted for by two reasons: first, among the proteins and PTMs identifiable by the 297 unique antibodies, Dactolisib and Dasatinib were highly selective and only affected a few targets; second, the short drug incubation duration restricted changes only to the direct targets and more broad secondary effects were not detected. The changes caused by the two inhibitors on PCa cells, T cells and Gr-MDSCs highlight the specificity of Dactolisib and Dasatinib: the dual PI3K/mTOR inhibitor Dactolisib dampened phospho-S6 and phospho-4E-BP1, whereas the TKi Dasatinib attenuated phospho-Src and phospho-p38 MAPK.
The main significance of this study is that through the distinct regulation of the two drugs on PCa cells, T cells and Gr-MDSCs, we can gain insights into the mechanism of action of Dactolisib and Dasatinib and help explain why Dactolisib can, but Dasatinib cannot, cooperate with ICB therapy in treating preclinical CRPC [16]. First, Dactolisib preferentially inhibited phospho-S6 (Ser235/236, Ser240/244) in PCa cells and phospho-4E-BP1 (Thr37/46) in Gr-MDSCs ( Figure 3A,B). Dactolisib inhibits PI3K and mTOR kinase activity by binding to the ATP-binding cleft of these enzymes [25]. mTOR is the key component of mTORC1 complex that promotes protein synthesis largely through the phosphorylation of two effectors, p70 S6 kinase (S6K1, which further phosphorylates S6) and 4E-BP1 [26]. Previous studies demonstrate that S6K1 and 4E-BP1 can be dephosphorylated in different patterns by the same mTORC1 inhibitor (e.g., rapamycin) in a cell type-dependent manner [27,28]. In the CPPSML PCa cells, Pten deficiency leads to hyperactive PI3K/mTOR signaling [16,29], whereas for Gr-MDSCs, several studies have established the essential role of PI3Kδ/γ signaling in the immunosuppressive function [16,30,31]. Due to the differences in upstream activation mechanisms, it is conceivable that mTORC1 in the two cell types elicits differential activation of S6K1 and 4E-BP1, which may explain the differential response of phospho-S6 and phospho-4E-BP1 in the two cell types to the same pan-PI3K inhibitor Dactolisib. The much more significant reduction of Thr37/46 phosphorylation of 4E-BP1 compared with Ser65 phosphorylation in Dactolisib-treated Gr-MDSCs is likely caused by the hierarchical phosphorylation process of 4E-BP1, whereby phosphorylation of Thr37/46 is a priming event for the subsequent phosphorylation of Ser65 [32,33]. It will be interesting for future experiments to define the role of 4E-BP1 phosphorylation in the immunosuppressive function of Gr-MDSCs.
With these said, our validation studies with western blot still revealed substantial downregulation of both phospho-S6 and phospho-4E-BP1 by Dactolisib treatment in Gr-MDSCs ( Figure 4A), consistent with our previous publication [16]. Moreover, at the transcriptomic level, Dactolisib treatment (but not Dasatinib) on Gr-MDSCs caused an interesting downregulation of genes involved in the mitochondrial electron transport chain ( Figure 5C, Figure S1B,C). It is known that mTOR signaling coordinates energy consumption by regulating mRNA translation and synthesis of nucleus-encoded mitochondria-related components of complexes I and V [34]. Specifically, mTORC1 controls mitochondrial activity and biogenesis by selectively promoting translation of nucleus-encoded mitochondria-related mRNAs via inhibition of 4E-BP [35]. A previous study demonstrated reprogrammed metabolism in tumor-infiltrating MDSCs in various murine tumor models, whereby these cells upregulated fatty acid uptake and β oxidation, mitochondrial biogenesis and oxygen consumption [36]. Inhibition of fatty acid oxidation by etomoxir attenuated the immunosuppressive activity of MDSCs and synergized with immunotherapy [36], similar to our observation of the synergistic anti-tumor efficacy of Dactolisib and immunotherapy [16]. Collectively, these results suggest that a major effect of Dactolisib on Gr-MDSCs is inhibition of mTOR and oxidative phosphorylation, which compromises fatty acid oxidation and leads to loss of immunosuppression.
Dasatinib preferentially inhibited Src and p38 MAPK phosphorylation in T cells, providing a molecular explanation for the effect of Dasatinib on diminishing T cells in the CRPC microenvironment [16]. Src-family kinases Lck and Fyn in proximal T-cell receptor signal transduction play critical roles in T-cell activation, differentiation, and tolerance [37]. Dasatinib inhibits Src, Lck and Fyn [38]. While the RPPA platform did not have antibodies recognizing phospho-Lck, Fyn or phospho-Fyn, the antibodies for Src and phospho-Src (Tyr527, Tyr416) provided surrogates for evaluating the effects of Dasatinib on T cells as well as PCa cells and Gr-MDSCs. Src activity is regulated by tyrosine phosphorylation at two sites, but with opposing effects [39]. Phosphorylation at Tyr527 in the carboxy-terminal tail by Csk renders Src in a closed, inactive conformation. Upon dephosphorylation of Tyr527, Src changes into an open conformation and self-phosphorylates Tyr416 to become activated [39]. Dasatinib reduced Tyr527 phosphorylation in all three cell types, yet only reduced Tyr416 phosphorylation in T cells ( Figure 3C, Figure 4C), suggesting complete loss of Src activity preferentially in T cells. Additional validation experiments with western blot showed dampened Lck phosphorylation by Dasatinib in T cells ( Figure 4C), which further helps explain the depletion of tumor-infiltrating T cells by a month-long Dasatinib treatment in the CPPSML CRPC model [16]. Moreover, Dasatinib preferentially inhibited the phosphorylating activation of p38 MAPK ( Figure 3C, Figure 4C), a protein with important roles in activating T cells especially CD4 + T cells to release TNF-α [40,41]. These results collectively help explain the failure of Dasatinib to synergize with ICB in treating CRPC [16]. Therefore, while Dasatinib may block MDSC expansion in certain scenarios [42,43], its adverse effect on T cells [16,44] makes its application in combination immunotherapy unpromising.
Taken together, our studies highlight the advantage of using a targeted proteomics approach such as RPPA to dissect the impact of inhibitors on primary tumor cells and immune cells. By finding distinct protein expression patterns and differential sensitivity of protein phosphorylation to Dactolisib and Dasatinib, our results offer some in-depth explanations on the molecular basis of the success of Dactolisib/ICB combination and the failure of Dasatinib/ICB combination in the preclinical CRPC model, and strengthen the idea of combining PI3K inhibitors and immunotherapy in treating refractory malignancies. RPPA provides a robust and quick snapshot of some of the most important proteins in cells. Nevertheless, an obvious limitation is that the number of antibodies is still low compared with the entire proteome. Future studies employing mass spectrometry-based proteomics and phosphoproteomics [45] will reveal the effect of targeted therapeutics on Gr-MDSCs in more comprehensive and precise manners. A caveat of the study is that the 2 h treatment of various cells in vitro for RPPA analysis may not represent what these cells experience in vivo when a patient is treated with Dactolisib or Dasatinib. Therefore, conclusions drawn from these in vitro cell models with short drug treatments should be further validated with specimens collected from possible clinical trials that use Dactolisib or Dasatinib to treat patients with PCa.

Inhibitor Treatment
Gr-MDSCs, T cells and PCa cells isolated and plated in respective medium as described above were added with DMSO, Dactolisib and Dasatinib at different concentrations for 2 h in a standard tissue culture incubator with 37 • C and 5% CO 2 . Next, the cells were washed with inhibitor-free and serum-free medium once and harvested as pellet and stored at −80 • C before submitting to the RPPA analysis.

Reverse Phase Protein Array (RPPA)
RPPA was conducted by the RPPA Core Facility at MD Anderson Cancer Center following the standard protocols at the facility, the details of which can be found on the core facility website (www.mdanderson.org/research/research-resources/core-facilities/functional-proteomics-rppacore.html). The dataset returned from the facility included raw data in log2 value and normalized data in linear value.

Western Blot
For western blot analysis, cell lysates prepared by the RPPA core were used following the procedure as previously described [16]. Primary antibodies were mostly purchased from Cell Signaling Technology

Microarray and Analysis
RNA was purified from inhibitor treated primary Gr-MDSCs (6 h) using RNeasy Mini Kit (Qiagen, Germantown, MD, USA). Microarray (Affymetrix Mouse Genome 430 2.0 Array) was conducted by the Sequencing and Microarray Facility at MD Anderson Cancer Center following the facility's standard procedure. The data were analyzed in RStudio, including oligo 1.50.0 package [46] and arrayQualityMetrics 3.42.0 package [47] used for quality control and calibration, ArrayExpress 1.46.0 package [48] and mouse4302.db 3.2.3 package used for probe annotation, and limma 3.42.2 package [49] was used for differential expression analysis. p value of 0.05 or 0.01 by unpaired t-test was used as threshold of significantly upregulated or downregulated genes by Dactolisib or Dasatinib treatment, respectively. For pathway analysis, the gene lists were imported to RStudio with clusterProfiler 3.14.3 package [50] used for GO term and KEGG pathway analysis and ReactomePA 1.30.0 package [51] used for Reactome analysis. The microarray raw and normalized data were deposited to NCBI GEO database with accession number GSE146083.

Statistical Analysis
Heatmaps were constructed with median-centered log2 values in Figure 1 and normalized log2 values in Figure 2 using package pheatmap 1.0.12 in RStudio. Graphs in Figure 3 were generated using normalized linear values. Significance test in Figures 1C and 5B was by unpaired t-test in RStudio.  Table S1. RPPA data normalized in linear values, generated by the RPPA Core Facility; Table S2. Differential RPPA protein levels between MDSC and T cells (unpaired student t-test); Table S3. Differential RPPA protein levels between MDSC and PCa cells (unpaired student t-test); Table S4. Differential RPPA protein levels between PCa cells and T cells (unpaired student t-test);