Inhibition of PI3K by copanlisib exerts potent antitumor effects on Merkel cell carcinoma cell lines and mouse xenografts

Merkel cell carcinoma (MCC) is a highly aggressive neuroendocrine skin cancer with steadily increasing incidence and poor prognosis. Despite recent success with immunotherapy, 50% of patients still succumb to their diseases. To date, there is no Food and Drug Administration-approved targeted therapy for advanced MCC. Aberrant activation of phosphatidylinositide-3-kinase (PI3K)/AKT/mTOR pathway is frequently detected in MCC, making it an attractive therapeutic target. We previously found PI3K pathway activation in human MCC cell lines and tumors and demonstrated complete clinical response in a Stage IV MCC patient treated with PI3K inhibitor idelalisib. Here, we found that both PI3K-α and -δ isoforms are abundantly expressed in our MCC cell lines and clinical samples; we therefore examined antitumor efficacy across a panel of five PI3K inhibitors with distinctive isoform-specificities, including idelalisib (PI3K-δ), copanlisib (PI3K-α/δ), duvelisib (PI3K-γ/δ), alpelisib (PI3K-α), and AZD8186 (PI3K-β/δ). Of these, copanlisib exerts the most potent antitumor effects, markedly inhibiting cell proliferation, survival, and tumor growth by suppressing PI3K/mTOR/Akt activities in mouse models generated from MCC cell xenografts and patient-derived tumor xenografts. These results provide compelling preclinical evidence for application of copanlisib in advanced MCC with aberrant PI3K activation for which immunotherapy is insufficient, or patients who are unsuitable for immunotherapy.

(score 0) to high expression (score 3), demonstrated that 20% (10 of 50 MCC tumors) had high expression (score 2 and score 3) of PI3K-α isoform, whereas 30% (15 of 50) had no detectable expression (score 0). High PI3K-δ expression was observed in 52% (26 out of 50) of MCC tumors, and no PI3K-δ was detected in 8% of samples ( Fig. 1B,C). Representative immunohistochemistry staining of PI3K-α and -δ in human MCC samples are shown in Fig. 1D. These results demonstrate that class I PI3K isoforms are differentially expressed in MCC cell lines and tissue samples, and indicate that MCCs may respond distinctively to isoform-specific PI3K inhibitors.
Inhibition of PI3K-α/δ by copanlisib elicits the most potent antitumor effects on MCC cell lines compared to other PI3K isoform-selective inhibitors. Next, we tested the responses of the above four MCC cell lines to different PI3K inhibitors, which have distinctive isoform-selectivity, including idelalisib, alpelisib, copanlisib, AZD8186, and duvelisib. The antitumor efficacy of these inhibitors at a series of concentrations from 0 to 10 µM on MCC cell lines was measured by CCK-8 assay, which has been used for assessment of cell viability and proliferation. The half-maximal growth inhibitory concentration (GI 50 ) of these inhibitors on different MCC cell lines was calculated as described previously 49 and shown in Fig. 2A. MCC-3 and MCC-9 cell viability and proliferation was suppressed by all five PI3K inhibitors. Among them, dual-isoform specific inhibitors (copanlisib, AZD8186, and duvelisib) generally showed more potency than single-isoform inhibitors (alpelisib and idelalisib), though MCC-9 was more sensitive to idelalisib (PI3K-δ) than AZD8186 (PI3K-β/δ). Interestingly, inhibition of PI3K-δ (idelalisib) exerted more potent anti-tumor growth effect than PI3K-α inhibition (alpelisib) in MCC-3 and MCC-9 cells, which display predominant PI3K-δ mRNA expression (Fig. 1A). However, this was not the case in MCC-21 cells; although PI3K-δ is highly expressed in MCC-21, this cell line responded poorly to idelalisib (PI3K-δ), AZD8186 (PI3K-β/δ), and duvelisib (PI3K-γ/δ). Instead, MCC-21 proliferation was well repressed by inhibition of PI3K-α (alpelisib) and PI3K-α/δ (copanlisib), suggesting that predominant isoform expression does not fully correlate to responsiveness. Although the underlying mechanisms are potentially intriguing and require further investigation, copanlisib (PI3K-α/δ) demonstrated the most potent anti-tumor efficacy on MCC-3, MCC-9, and MCC-21 cell lines, in which PI3K-δ and -α are the two most abundantly expressed PI3K isoforms. In contrast, MKL-1 cells, which we found had negligible expression of PI3K isoforms, were resistant to all PI3K inhibitors tested ( Fig. 2A). Finally, since inhibition of PI3K-α alone by alpelisib showed less anti-tumor potency on MCC-3 and MCC-9 than other inhibitors, alpelisib was excluded from further experiments.
We then further examined the effects of PI3K inhibitors on apoptosis of MCC-3, MCC-9, and MCC-21 cell lines. Cultured MCC cells were treated with three doses (5 nM, 50 nM and 100 nM) of idelalisib, copanlisib, AZD8186 and duvelisib for 24 hours, respectively. DMSO treatment served as a respective negative control for each cell line. Apoptotic rate was measured by Annexin-V and PI (propidium iodide) staining followed by flow-cytometry analysis (Fig. 2B). All four inhibitors induced apoptosis in all three MCC cell lines in a dose-dependent manner with the more prominent effect observed in MCC-3 and MCC-9. Consistently, inhibition of PI3Kα/δ by copanlisib at three doses resulted in the most robust anti-MCC survival effect on three MCC cell lines ( Fig. 2A). Though MKL-1 failed to respond to all PI3K inhibitors tested, we wanted to examine if a higher dose of copanlisib induced apoptosis. MKL-1 cells were treated with copanlisib 1 µM for 24 h, followed by Annexin-V and PI staining and flow cytometry analysis; copanlisib exerted negligible apoptotic effect on MKL-1 cells (Supplementary Fig. 4). In summary, these data indicate that inhibition of PI3K α/δ isoforms by copanlisib had the most potent antitumor growth and survival effects on MCC compared to other PI3K inhibitors.

Copanlisib suppresses MCC colony formation by inhibiting MCC cell proliferation and survival in vitro.
To assess the effect of copanlisib on MCC tumorigenesis in vitro, we performed a clonogenic assay on three MCC cell lines (MCC-3, MCC-9, and MCC-21) responsive to copanlisib treatment in the previous www.nature.com/scientificreports www.nature.com/scientificreports/ experiment (Fig. 2). As shown in Fig. 3A,B, copanlisib treatment significantly decreased the number of colonies formed in methylcellulose medium compared to that in vehicle-treated MCC cells. To further identify the mechanisms by which copanlisib suppresses MCC colony formation, we analyzed cell-cycle progression by flow cytometry in vehicle and copanlisib-treated MCC cell lines. For this purpose, MCC-3, MCC-9, and MCC-21 cells were treated with vehicle or idelalisib or copanlisib at 5 nM and 50 nM concentrations for 24 hours, respectively. Cells were then collected and subjected to BrdU (Bromodeoxyuridine) and PI (propidium iodide) fluorescent staining followed by flow cytometry analysis. Both idelalisib and copanlisib significantly decreased cell populations at S phase, an index of cell proliferation, compared to vehicle-treated controls. Meanwhile, the percentage of apoptotic cells, represented by sub-G1 cell population with DNA fragmentation, significantly increased in idelalisib-and copanlisib-treated MCC cells relative to controls (Fig. 3C). Consistent with the results shown in Fig. 2A, copanlisib exhibited stronger anti-tumor effects than idelalisib. These data indicate that copanlisib attenuates MCC growth in vitro by inhibiting MCC cell proliferation and inducing apoptosis.
Copanlisib is more potent than idelalisib in suppressing PI3K/AKT/mTOR pathway in MCC cells. We next set out to examine the efficacy of idelalisib and copanlisib in decreasing activities of PI3K and its downstream AKT and mTOR pathways (Fig. 4A). Cells from three MCC cell lines (MCC-3, −9, −21) were treated with vehicle or 5 nM/50 nM of idelalisib and copanlisib for 3 and 24 hours, respectively. Vehicle-treated cells served as negative controls. Whole cell protein lysates were prepared and PI3K pathway activation, as revealed by phosphorylation of AKT, mTOR and their downstream targets, was detected by western blots using specific antibodies as described in Materials and Methods (Fig. 4B). Consistent with the GI 50 data as shown in Fig. 2A, copanlisib inhibited phosphorylation and activation of AKT/mTOR pathway more robustly than idelalisib in all three MCC cell lines (Fig. 4B). Reflecting the inability of idelalisib to suppress MCC-21 cell proliferation ( Fig. 2A), we found that idelalisib had little effect on AKT and mTOR activation in this cell line at both 5 nM and 50 nM concentrations after treatment for 3 and 24 hours (lower panel in Fig. 4B). In contrast, both idelalisib and copanlisib induced quick reduction in phosphorylation of PI3K downstream signaling molecules in MCC-3 and MCC-9 after 3-hour incubation. We observed a rebound of AKT and mTOR phosphorylation after a 24-hour incubation with these PI3K inhibitors, which was more apparent in idelalisib-treated MCC-3 www.nature.com/scientificreports www.nature.com/scientificreports/ prepared with 2 × 10 7 cells of MCC-3, MCC-9, and MCC-21, respectively, and inoculated subcutaneously into the rear flanks of immunodeficient NOD scid gamma (NSG) mice. As described in Materials and Methods, we successfully established, for the first time, two MCC PDX models (PDX-60 and PDX-68). MCC cell lines and PDX tumors exhibited MCC histological features and classical MCC markers (Supplementary Figs. 1 and 2). When xenograft tumor growth approached ~100 mm 3 in volume, mice began receiving 14 mg/kg of copanlisib or vehicle, administered by intraperitoneal injection every other day for up to 6 weeks. Copanlisib treatment had no obvious signs of toxicity as monitored by body weight, food and water intake, and activity (data not shown). As shown in Fig. 5A,B, copanlisib significantly attenuated in vivo growth of all three MCC CDX tumors and two PDX tumors. Of note, although the drug displayed more potent anti-tumor effects in vitro on MCC-3 and MCC-9 than MCC-21 (Fig. 2), copanlisib repressed MCC-21 tumor growth more markedly in vivo. The explanations and mechanisms for the differential effects of copanlisib on MCC in vivo and in vitro are unclear and warrants further study. Next, we performed immunohistochemistry staining of AKT phosphorylation, cleaved caspase-3, and Ki67, as indexes of PI3K activation, apoptosis, and cell proliferation, respectively, in paraffin-embedded xenograft tumor sections (Fig. 5C,D). Copanlisib treatment in mice led to significant inhibition of PI3K activity, induction of tumor cell apoptosis, and decrease in MCC cell proliferation in vivo. These data provide compelling evidence that dual inhibition of PI3K-α and -δ isoforms by copanlisib abrogates MCC tumor growth by inducing tumor cell apoptosis and inhibiting MCC cell proliferation.

Discussion
Despite great advances in our understanding of MCC biology and therapy in recent years, the cellular and molecular mechanisms governing MCC tumorigenesis and metastasis remain largely unknown. Currently, no FDA-approved molecularly targeted therapy exists. Though immunotherapies targeting the PD1/PD-L1 immune checkpoint pathway have been FDA-approved for treatment of advanced MCC 11,12 , a significant proportion of MCC patients are either resistant to immune checkpoint blockade or unsuitable for immunotherapy due to autoimmune or immunosuppressed conditions 1 . There is an imperative need to identify and test novel targeted therapies, which can boost anticancer immunity in addition to their direct cell-autonomous effects on tumor cells, and can be used as alternative treatments for MCC patients who are not suitable for immunotherapy.
The molecules along the signaling network of PI3K/ATK/mTOR pathway regulate most cellular processes involved in cancer development, including cell cycle progression, survival, metabolism, motility and immunity 30 . Hyperactivation of this pathway is commonly detected in many types of cancers, including MCC 4,27,28,30,41 , and oncogenic mutations in PIK3CA gene have been detected in 4-10% of MCC 26,27 . Importantly, our group and others have shown promising anti-tumor effects on MCC by inhibition of PI3K and mTOR 26,27,47,49,58,59 .
Recently we first reported PI3K-δ expression in human MCC cells and the first successful clinical application of PI3K-δ inhibitor in a Stage IV MCC patient with PIK3CA mutation 47 . Consistent with the findings reported by Chteinberg et al. 48 , we have found that PI3K-δ isoform can be detected in 92% of archival MCC samples. However, we detected PI3K-δ expression in only 3 out of 4 MCC cell lines, including two MCPyV-negative cell lines (MCC-3 and MCC-9) and one MCPyV-positive cell line (MCC-21). Interestingly, we found only minimal PI3K-δ expression in MKL-1 cell line, which is inconsistent with the report by Chteinberg et al.
We have previously demonstrated that idelalisib was able to resolve liver metastases in a patient with stage IV MCC 47 . However, we chose copanlisib for further in vitro and in vivo studies because, in addition to its relatively low GI 50 value among the inhibitors tested, copanlisib has been recently approved by the FDA for treatment of breast cancer (solid tumor) with an acceptable side-effect profile, and we believe that it has greater translational potential 38 .
In contrast to our findings, Chteinberg et al. report that alpelisib (PI3K-α) more potently suppressed in vitro cell proliferation than idelalisib (PI3K-δ) 48 in the panel of MCC cell lines tested in their laboratory. However, their reported half maximal inhibitory doses (IC 50 ) of idelalisib, ranging from 29.6 µM to 81.9 µM, were well beyond the highest drug concentrations used in our dose response studies (10 µM). Moreover, Chteinberg et al. included two MCPyV-negative MCC cell lines ("MCC13" and "MCC26"), which have been characterized as atypical MCC cell lines 61 . Though the exact explanations for discrepancies between our two studies are unknown, differences in experiment design, cell culture conditions, and biochemical assays may contribute to the variance in our observations.
Although PI3K-α and -β expression was detected, MKL-1 cells were resistant to all five PI3K inhibitors tested. This may be due to the lack of PI3K/AKT/mTOR pathway activation in MKL-1 cell line, as we previously reported 49 , and suggests that this pathway plays a minimal role in MKL-1 tumorigenesis. Mechanistically, results from a series of cellular and biochemical experiments demonstrate that copanlisib inhibits PI3K/AKT/mTOR pathway activities and represses MCC cell proliferation and survival more potently than idelalisib in MCC-3, −9, and −21. Additionally, we found that copanlisib markedly suppresses growth and tumorigenesis of these three MCC cell lines in vitro as assessed by tumor cell colony formation assay, and in vivo as examined in three Scientific RepoRtS | (2020) 10:8867 | https://doi.org/10.1038/s41598-020-65637-2 www.nature.com/scientificreports www.nature.com/scientificreports/ MCC CDX. These in vivo drug efficacy studies were further confirmed in two PDX mouse models of MCC; to our knowledge, this is the first reported preclinical drug study using MCC PDX models.
In the past few years, cancer immunotherapies targeting T-cell immune checkpoint receptors PD-1/PD-L1 have demonstrated great clinical benefits to MCC patients [11][12][13]17,18 . Nevertheless, 50% of MCC patients still Vehicle-treated xenograft tumors served as negative controls. n = 5-10 * p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 compared with vehicle-treated xenograft tumors by paired Student t-test. succumb to their diseases despite immunotherapy, underscoring the need for new therapeutic strategies for those patients as well as those who are not suitable for immunotherapy due to immunosuppressed conditions and/or autoimmune diseases. These alternative therapies may also augment efficacy of immunotherapies and significantly improve clinical benefits when utilized in combination with different types of immune-targeting drugs [19][20][21] . A large-scale survey of cancer genomic and therapeutic databases has identified five candidate genes, namely PIK3CA, BRAF, NF1, NRAS, and PTEN, the targeting of which could be suitable for combination therapy with immunotherapy 62 . Moreover, inhibition of PI3K/AKT/mTOR pathway in other cancers has been shown not only to directly target cancer cells but also modulate tumor microenvironment and tumor-infiltrated immune cells [63][64][65][66][67][68] . Similar to other cancers, PI3K/AKT/mTOR pathway is hyperactive in MCC and inhibition of this pathway has demonstrated significant anti-MCC effects in vitro and in vivo as reported in this study and by others 26,27,47,49,58,59 .
In summary, we have confirmed abundant PI3K-δ expression in MCC and also demonstrated that PI3K-α is commonly expressed across MCC cell lines and archival MCC tumors. Furthermore, we have shown that inhibition of PI3K/AKT/mTOR pathway by copanlisib (PI3K-α/δ) suppresses MCC cell proliferation and survival more potently than other PI3K inhibitors with single/dual isoform specificities. Copanlisib attenuates MCC cell-line derived xenograft tumor growth by inhibiting MCC proliferation and stimulating apoptosis, and this therapeutic efficacy was further evaluated and confirmed in MCC patient-derived tumor xenograft models. Thus, this study provides compelling evidence for the application of copanlisib as monotherapy and/or potentially in combinatorial therapies for a subset of advanced MCCs, as well as other solid tumors with PI3K activation for which standard therapies are insufficient.

Generation of MCC cell line-derived and patient-derived xenograft models in mice. MCC
cell line-derived xenograft (CDX) mouse models were generated using six-week-old female immunodeficient NSG mice (Jackson Laboratory, strain #005557). In brief, 2 × 10 7 MCC cells were suspended in Matrigel (BD Biosciences; catalog # 354248) and subcutaneously inoculated on right rear flanks. Palpable tumor growth appeared within 3 to 5 days of inoculation, and treatment per protocol began when tumors reached approximately 100 mm 3 volume. To generate MCC patient-derived xenograft (PDX) mouse models, we obtained excess surgical tissue from consenting MCC patients at the University of Arkansas for Medical Sciences (UAMS) in accordance with the Declaration of Helsinki and relevant institutional guidelines for human studies, under study protocols approved by the Institutional Review Board (IRB). Animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at UAMS, in accordance with laboratory animal care and use guidelines set by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International. Briefly, excess fresh MCC tissues not needed for clinical diagnosis were processed and sectioned into 2-to 4-mm 3 pieces. Non-necrotic pieces were subcutaneously implanted into the rear flanks of immunodeficient NSG mice. Per standard parlance 69,70 , this initial engraftment of human tumor tissue was termed as "F 1 " generation; successful engraftments were subsequently allowed to grow until approaching tumor endpoint (~1500mm 3 volume), harvested, processed, biobanked and/or passaged into further immunodeficient NSG mouse cohorts. Each successive mouse-to-mouse passage was numbered consecutively as F 2 generation and so forth. RNA, gDNA, and whole tissue samples were obtained from tumors in each generational cohort, characterized by RT-PCR and immunohistochemistry, and compared to originating tumors to validate each MCC PDX lineage. In this study we utilized F 5 generation of our PDX-60 lineage and F 6 generation of PDX-68 lineage, which exhibit classical MCC morphology and express classic MCC markers (see Supplementary Fig. 2), to expand tumor-bearing mouse cohorts for copanlisib preclinical drug studies. Tumor-bearing mice were randomly divided into control and treatment groups (n = 5-10 for each condition) receiving copanlisib treatment. Copanlisib was formulated in PEG400/ acidified water solution with pH ~4.5 and administered at 14 mg/kg via intraperitoneal injection every other day. Control mice received vehicle only. Mice were monitored daily and tumors were measured using digital calipers. Tumor volume (TV) was calculated as L x W 2 /2, where length (L) is the longer dimension and width (W) is the shorter dimension. The therapeutic efficacy of copanlisib on tumor growth in each CDX and PDX was defined by tumor growth inhibition, calculated as (TGI) = [1−(TV copanlisib /mean TV vehicle )] × 100.
Immunohistochemistry. Dissected MCC xenograft tumors and MCC patient tumor samples (collected under protocols approved by the UAMS IRB in accordance with relevant guidelines) were fixed overnight in 10% neutral buffered formalin and paraffin-embedded by routine histology procedure. Five micrometer tissue section slides were prepared, processed for antigen retrieval, and stained as described before 49,58,59 . Samples were incubated with specific primary antibodies for p-Akt-Ser473 (1:50), cleaved caspase-3 (1:100) and Ki67 (1:2000) at 4 °C overnight. Samples were then incubated with goat anti-rabbit-secondary antibody for one hour at room temperature, followed by development with horseradish peroxidase detection system. Slides were viewed under Cell cycle analysis by flow cytometry. Cell cycle distribution in MCC cell populations was detected by BD Pharmingen BrdU Flow Kits (BD Biosciences; San Jose, CA). MCC cells were seeded at a cell density of 2 × 10 5 per well in 6-well plates and treated with idelalisib, alpelisib, copanlisib, AZD8186 and duvelisib for 24 hours as described before 59 . BrdU incorporation was detected using FITC-conjugated anti-BrdU antibody followed by 7-AAD staining per manufacturer's protocol. Cell cycle detection was performed via FACSAria flow cytometer and analyzed by FlowJo software (version 10.4.2) and cell cycle distribution was reported as the percentage of cells in G0/G1, S, and G2/M populations.

Determination of apoptosis by flow cytometry. Apoptotic cells in each control and treatment
group were detected by Annexin V-FITC apoptosis-detection kit (BD Biosciences; San Jose, CA). Briefly, MCC cells were plated in 6 well plates (2 × 10 5 per well) and treated with idelalisib, copanlisib, AZD8186 or duvelisib for 24 hours at indicated concentrations. At the end of incubation, cells were collected and stained with Annexin-V-FITC/Propidium Iodide (PI) followed by FACSAria (BD Biosciences) analysis within an hour of staining. Cell death was scored by the following criteria, set by appropriate gating: (a) early apoptotic cells (PI negative, FITC Annevin-V positive), (b) late apoptotic or dead cells (doubly positive for both FITC Annevin-V and PI), and (c) live cells (doubly negative for Annexin-V and PI). Statistical analysis was performed using FlowJo software (version 10.4.2).

RNA extraction and gene expression analysis. Total RNA was isolated from MCC cells via RNeasy
Kit (Qiagen) per manufacturer's instructions. Complementary DNA was generated from MCC mRNA using High-Capacity cDNA Reverse Transcription Kit (Life Technologies). Quantitative real-time-PCR (qRT-PCR) was performed with a StepOne Plus Real-Time PCR System (Applied Biosystems) as described previously using specific TaqMan Gene Expression Assay primers purchased from Applied Biosystems: PI3Kα, PI3Kβ, PI3Kγ and PI3Kδ and MRPS2 (mitochondrial ribosomal protein S2). Triplicate runs of each sample were normalized to MRPS2 mRNA to determine relative expression.
Western blot. MCC cells were harvested and processed for Western blot analysis as described previously 49,58,59 . Xenograft tumor tissues harvested from mice were homogenized in 2% SDS lysis buffer and processed as described previously 49 . Briefly, whole cell protein lysates (10-30 µg per lane) were resolved by 8% or 12% SDS-PAGE gel electrophoresis and transferred onto PVDF membranes by a semidry blotting system (Bio-Rad, Hercules, CA). Membranes were blocked in 5% fat-free milk/Tris-buffered saline for 1 hour at RT and incubated with specific primary antibodies at 4 °C overnight, followed by one hour RT incubation with secondary antibodies conjugated with horseradish peroxidase. Visualization of immunoreactive proteins was achieved using ECL detection reagent per manufacturer's instruction. Alpha-tubulin was used as a loading control and all immunoblotting data represent contiguous lanes. Statistical analysis. All measurements were made in triplicate, and all values are represented as mean ± SD.
Statistical analyses were performed with Student t test or one-way analysis of variance (ANOVA) using GraphPad prism software (v6.0; San Diego, CA). P value < 0.05 was considered statistically significant.