Key points

  • PDGFRβ-STAT5 activity, acting in parallel to the NPM-ALK-STAT3 signaling axis, correlates with an inferior Event Free Survival (EFS) and Cumulative Incidence of Relapse (CI-R) in ALCL.

  • PDGFRβ induced STAT5 activity, increases proliferation by stimulating IL-10 secretion and blocking apoptosis by upregulating Bcl-xL, characterizing PDGFRβ as a novel biomarker.

  • Combined loss of STAT5A and STAT5B is lethal in ALK+ ALCL, suggesting STAT5 as a valid therapeutic target that can be successfully inhibited with the STAT inhibitor AC-4-130.

Introduction

Anaplastic Large Cell Lymphoma (ALCL), characterized by expression of CD30, refers to a family of Non-Hodgkin T cell lymphomas divided into four entities: systemic Anaplastic Lymphoma Kinase (ALK) positive (ALK+ ALCL), systemic ALK negative (ALK− ALCL), primary cutaneous ALCL (pc-ALCL) and breast-implant associated (BIA-ALCL) [1]. In 70% of systemic ALK+ ALCL cases, patients present with the t(2;5)(p23;35) translocation, resulting in abundant cytoplasmic and nuclear expression of the Nucleophosmin-Anaplastic Lymphoma Kinase (NPM-ALK) chimeric protein [2]. The expression of this ligand-independent oncogene is mediated via the NPM1 promoter but drives disease pathogenesis via the ALK kinase domain [3]. NPM-ALK stimulates a plethora of oncogenic signal transduction pathways including JAK/STAT, RAS/RAF/ERK, PI3K/AKT/mTOR and JNK/p38/AP-1, but the key vulnerable nodes for targeting remain illusive [4]. In the context of ALCL, STAT3 has been identified as a crucial signaling modulator downstream of NPM-ALK responsible for disease maintenance [5,6,7]. Additional members of the STAT family, such as STAT1 [8] and STAT5A/B [9], have also been reported in ALCL however their function is not extensively studied and remains enigmatic.

Event Free Survival (EFS) and Overall Survival (OS) rates for ALK+ ALCL patients treated with standard CHOP (cyclophosphamide, doxorubicin, vincristine and prednisone) chemotherapy are relatively succesful [10,11,12,13]. However, current relapse rates range from 20 to 40%, highlighting the urgency to develop alternative therapy approaches for relapsed patients [14]. The restricted expression of ALK in neuronal cells during development make it an ideal drug target with potential to overcome relapse. First-generation ALK tyrosine kinase inhibitor (TKI) crizotinib (Xalkori), initially approved by the Federal Drug Administration (FDA) for Non-Small Cell Lung Cancer (NSCLC) patients harboring ALK fusions, has recently been authorized for pediatric patients with relapsed or refractory systemic ALK+ ALCL (NCT00939770). Second [15,16,17,18,19], third [20], and fourth-generations [21] of ALK TKIs have since been developed and are in clinical use for ALK+ NSCLC [22, 23]. Mounting evidence suggests that patients with ALK-expressing malignancies eventually develop point mutations in ALK [14], rendering them insensitive to ALK inhibition and provoking reoccurrence of a more aggressive disease [6, 24]. Thus, there is a clear need for more effective therapies targeting alternative core signaling pathways.

We identified the AP-1 TFs, cJUN and JUNB, as downstream effectors of NPM-ALK [25], transcriptionally regulating the expression of Platelet Derived Growth Factor Receptor Beta (Pdgfrb) [26]. Strikingly, inhibition of PDGFRβ by the TKI imatinib (Gleevec) induced a complete and sustained remission in a late-stage relapsed ALK+ ALCL patient [26]. Furthermore, our findings resulted in the initiation of a clinical study evaluating imatinib efficacy according to PDGFR status [27]. Imatinib treatment can thus be considered as a promising therapeutic strategy for relapsed ALK+ ALCL. However, the mechanism of action and the extent of PDGFRβ involvement in ALCL pathogenesis still remains elusive.

In this study, we determine the influence of PDGFRβ on ALCL by developing a genetic ALK+ ALCL mouse model lacking PDGFRβ expression in neoplastic CD4+ T cells. Strikingly, genetic loss of Pdgfrb results in a significant increase in survival in line with the observed benefits of PDGFRβ kinase activity blockade via imatinib in patients. Mechanistically, we identify STAT5 as a novel downstream target of PDGFRβ in ALCL. Blockade of STAT5 results in a significant dampening of viability by mediating the pro-survival factor Bcl-xL and the growth-enhancing cytokine IL-10. Additionally, we report the small molecule inhibitor AC-4-130 inhibits both STAT5 and STAT3 activation in our experimental setting, significantly inhibiting tumor development in vivo and highlighting the added benefit of a dual STAT3/5 inhibition. Overall, our findings suggest expression of PDGFRβ in ALCL stimulates a malignant boost, providing an attractive alternative/additive pathway for pharmacologic inhibition.

Materials and methods

Additional materials and resources can be found in Suppl. Materials and Methods.

Transgenic mouse strains

All animal experiments were performed in agreement with the ethical guidelines of the Medical University of Vienna and the Austrian Federal Ministry of Science and Research (Project number: BMWFW-66.009/0183-WF/V/3b/2017; BMWFW-66.009/ 0057-V/3b/2018; BMWFW-66.009/0401-V/3b/2018; 2020–0.103.412). Mice were kept in a pathogen-free environment under standard conditions. Cd4-NPM-ALK transgenic mice [28] were crossed with Cd4-Cre mice [29] and a second strain carrying loxP-flanked Pdgfrb (Exons 4–7) [30] in a C57BL/6 x BALB/c mixed background resulting in the desired genotypes: Cd4-NPM-ALKTg/+ Cd4-Cre+/+ Pdgfrbfl/fl (fl/fl Pdgfrb) and Cd4-NPM-ALKTg/+ Cd4-CreTg/+ PdgfrbΔ/Δ (Δ/Δ Pdgfrb) mice. Mice were genotyped using the primers listed in Suppl. Table 1.

Human tissue microarrays

All human samples were obtained with informed written consent and in accordance with the Declaration of Helsinki.

Adult TMAs

Formalin-fixed paraffin-embedded (FFPE) tissue specimens from both male and female adult patients, diagnosed with systemic ALK+ (n = 25) or ALK− (n = 25) ALCL, were obtained and reviewed by the local ethic boards of the Medical University of Vienna, Austria (no.1437/2016 and 1221/2019) and the University Hospital Brno, Czech Republic (no. 4–306/13/1). Diagnoses were assigned according to the WHO classification of lymphoid neoplasms.

Pediatric TMA

FFPE tissues specimens from 98 pediatric patients diagnosed with ALK+ ALCL included in the studies NHL-BFM90, NHL-BFM95 or enrolled in the European intergroup trial ALCL99 (NCT00006455) between 1992 and 2006, were obtained. All patients were treated with comparable Berlin-Frankfurt-Münster (BFM)-type chemotherapy, as previously described [12]. Eligibility was determined by detection of oncogenic NPM-ALK: either via NPM-ALK PCR, two color fluorescence in situ hybridization for the translocation t(2;5)(p23;35) or nuclear/cytoplasmic IHC staining for ALK. Staining procedures included bone marrow aspiration cytology and a spinal tap. Bone marrow involvement was defined by cytologically detectable ALCL cells, irrespective of cellular quantity. The studies were approved by the institutional ethics committee of the primary investigator of the NHL-BFM study group.

Immunohistochemistry and whole tissue scans

IHC staining was performed with tissue fixed for 24 hours in formalin before embedding in paraffin blocks, as described [26]. Antibodies used for IHC can be found in Suppl. Table 2. A Panoramic MIDI slide scanner (3DHistech Ltd., Budapest, Hungary) with 40x optics was used to digitalize the tissue sections. Quantitation was performed using Definiens® TM Tissue Studio histomorphometry software (Definiens AG, Munich, Germany). Images were taken with a Zeiss AxioImager Z1 microscope.

Multiplex immunobead cytokine assay

Murine fl/fl Pdgfrb and Δ/Δ Pdgfrb cells were cultured at a density 1 × 105 cells in a 12-well plate and treated with AC-4-130 or DMSO as vehicle control. Post 72 hours of treatment, supernatants were collected and analyzed using ProcartaPlex antibody-based, magnetic bead reagent assay panels for multiplex protein quantitation using the Luminex xMAP technology and instrument platform. Concentrations of cytokines in the supernatants were determined simultaneously with the customized 27-plex immunoassay kit (ProcartaPlex Mouse 27-plex, ThermoFisher Scientific) using magnetic beads. Undiluted frozen samples (50 μL) were processed in 96-well plates according to the manufacturer’s instructions. Standard curves for each analyte were generated by measuring individual standards in duplicate and using the reference concentration supplied by the manufacturer. Measurement was performed on a calibrated Bio-Plex 200 system (Bio-Rad) in combination with Bio-Plex Manager software, version 6.1 (Bio-Rad). The concentrations were calculated from the standard curve using five-parameter logistic (5PL) regression curve fitting.

ChIP-seq and initial processing

ChIP-seq was performed as previously described [31]. The antibodies used for each experiment are listed in Suppl. Table 3. For each ChIP, 5 μg of antibody coupled to 2 μg of magnetic Dynabeads (Life Technologies) was added to 3 ml of sonicated nuclear extract from formaldehyde-fixed cells. Chromatin was immunoprecipitated overnight, cross-links were reversed, and DNA was purified by precipitation with phenol:chloroform:isoamyl alcohol. DNA pellets were resuspended in 25 μl of TE buffer. Illumina sequencing, library construction, and ChIP-seq analysis methods were previously described [31]. Reads were aligned to the human reference genome (hg19) using bowtie v1.2.2 with parameters –k 2 –m 2 –best and –l set to the read length. For visualization, WIG files were created from aligned read positions using MACS v1.4 with parameters –w –S –space = 50 –nomodel –shiftsize = 200 to artificially extend reads to 200 bp and to calculate their density in 50-bp bins. Read counts in 50-bp bins were normalized to the millions of mapped reads, giving RPM values. WIG files were visualized in the IGV browser version 2.7.2.

CUT&RUN sequencing and initial processing

CUT&RUN coupled with high-throughput DNA sequencing was performed using antibodies listed in Suppl. Table 3 and Cutana pA/G-MNase (EpiCypher) according to the manufacturer’s protocol. Briefly, cells were washed and incubated with activated concanavalin A beads for 10 min at room temperature. Cells were then resuspended in antibody buffer containing 0.01% digitonin, 1 ml of each antibody (Suppl. Table 3) was added to individual cell aliquots, and tubes were rotated at 4 °C overnight. The following day, targeted chromatin digestion and release were performed with 2.5 ml of Cutana pA/G-MNase and 100 mM CaCl2. Retrieved genomic DNA was purified with a MinElute PCR purification kit and eluted in 10 ml of buffer EB. Sequencing libraries were prepared with an automated Swift 2S system, followed by 100-bp paired-end sequencing with NovaSeq 6000. Reads were aligned to the human reference genome (hg19) using bowtie v1.2.2 in single-end mode with parameters –k 2 –m 2 –best and –l set to the read length. For visualization, WIG files were created from aligned read positions using MACS v1.4 with parameters –w –S –space = 50 –nomodel –shiftsize = 200 to artificially extend reads to 200 bp and to calculate their density in 50-bp bins. Read counts in 50-bp bins were then normalized to the millions of mapped reads, giving reads per million (RPM) values. WIG files were visualized in the Integrative Genomics Viewer (IGV) browser version 2.7.2. The antibodies used for CUT&RUN are listed in Suppl. Table 3.

Data and code availability

Raw and processed data files were deposited to the NCBI GEO server. Code written in R/python to perform analysis of ChIP-seq and CUT&RUN are available upon request.

Statistical analysis

Log-rank (Mantel-Cox) test, Student’s t test (Wilcoxon-Mann-Whitney correction), and half maximal inhibitory concentration (IC50) statistical analyzes were performed using GraphPad Prism® Software version 8. P-values were defined as indicated in the figure legends: ns = p > 0.05; * = p < 0.05; ** = p < 0.01; *** = p < 0.001; **** = p < 0.0001).

Results

PDGFRβ is a biomarker for aggressive ALK+ ALCL

Imatinib treatment results in tumor regression in Cd4-NPM-ALK transgenic mice [26]. However, the mechanisms of action and whether this effect is dependent on NPM-ALK activity remained to be determined. This is particularly important as inhibition of alternative tumor-promoting pathways in combination with ALK inhibition might prevent the development of drug resistance [6]. We evaluated a set of ALCL biopsies, all taken from children treated with an identical BFM-based chemotherapy in three separate clinical trials, for PDGFRβ expression and correlated expression levels with clinical variables. Patients with PDGFRβ expression on tumor cells (n = 11) had a significantly lower five-year EFS compared to patients lacking (n = 87) membrane-bound PDGFRβ (Fig. 1A). We next analyzed an independent lymphoma patient dataset using the Oncomine database. In the six ALCL patient samples [32], we observed a significant upregulation of PDGFRB expression when compared to healthy donor CD4+ and CD8+ T cells. In contrast PDGFRA was not significantly upregulated (Fig. 1B). Having correlated PDGFRβ expression in patients with poorer EFS, we set out to map the molecular signatures activated by this receptor. We conditionally deleted Pdgfrb in CD4+ T cells in transgenic mice expressing the human NPM-ALK fusion oncogene under control of the murine Cd4 promoter and enhancer (NPM-ALKTg) [28], resulting in littermates either expressing wild type Pdgfrb (fl/fl Pdgfrb) or lacking Pdgfrb in CD4+ T cells (Δ/Δ Pdgfrb) (Fig. 1C). Regardless of sex, NPM-ALKTg mice, develop T cell lymphomas with a high penetrance [28]. We assessed and confirmed deletion of Pdgfrb in T cells by genotyping (Fig. S1A) and immunohistochemistry (Fig. 1D). NPM-ALK protein (Fig. S1B) and mRNA (Fig. S1C) were expressed at similar levels in fl/fl Pdgfrb and Δ/Δ Pdgfrb mice, suggesting that PDGFRβ does not influence NPM-ALK expression per se. Similarly, STAT3, an important downstream modulator of NPM-ALK, was strongly activated as demonstrated by tyrosine phosphorylation in 8 week-old thymi, confirming that NPM-ALK activity is independent of PDGFRβ expression (Fig. S1D) [33].

Fig. 1
figure 1

PDGFRβ is a biomarker for aggressive ALCL. A Tissue microarrays from 98 NPM-ALK+ ALCL patients enrolled to NHL-BFM 90 and 95 studies and the ALCL99 clinical trial, were stained for total PDGFRβ by immunohistochemistry (IHC) and expression levels were correlated with event-free survival (EFS). Staining was quantified according to high (n = 11; orange line) versus low (n = 87; grey line) PDGFRβ expression. B PDGFRB and PDGFRA mRNA transcript levels (log2 median-centered intensity) comparison between (n = 6) ALCL patients and (n = 10, T cells; n = 5, CD4+; and n = 5, CD8+) healthy donors of the Piccaluga dataset [32] extrapolated from the Oncomine database. C Schematic representation of the breeding strategy to obtain a genetic knockout of Pdgfrb in CD4+ T cells harboring the human NPM-ALK+ oncogenic fusion under the control of the Cd4 enhancer. NPM-ALKTg littermates either expressing wild type Pdgfrb (fl/fl Pdgfrb; grey mouse) or lacking Pdgfrb in CD4+ T cells (Δ/Δ Pdgfrb; orange mouse) were generated using Cre-mediated recombination driven by the Cd4 promoter. D Representative pictures of PDGFRβ IHC analysis of fl/fl Pdgfrb and Δ/Δ Pdgfrb mouse thymomas at the experimental end point. Black squares in the left picture represent the area chosen for the magnification depicted on the right. Scale bars: 50 μm. Staining of (n = 6) end point thymic tumors per genotype were scanned and whole-slide quantification was performed using Definiens™ software. E Kaplan Meier cumulative survival analysis of (n = 3) wild type (black line), (n = 19) fl/fl Pdgfrb (grey line) and (n = 24) Δ/Δ Pdgfrb (orange line) mice in biological replicates. Values next to the dotted lines on the x-axis indicate median life expectancy for each genotype. F Quantitative RT-qPCR of Pdgfrb mRNA transcripts from (n = 4) fl/fl Pdgfrb (grey) and (n = 4) Δ/Δ Pdgfrb (orange) primary tumor cell lines. Pdgfrb levels were normalized to 18S ribosomal RNA expression and depicted as fold-change over one fl/fl Pdgfrb biological replicate set to 1. G Hematoxilin and Eosin (H&E)-stained sections of tumor dissemination into the liver of two fl/fl Pdgfrb (grey) and Δ/Δ Pdgfrb (orange) mice at the experimental end point. The black dashed line represents malignant infiltration into secondary organs. Scale bars: 50 μm. A and E p values were determined by the log-rank (Mantel-Cox) test. B, D, F and G Data are shown as the means ± SD and p values were determined by the unpaired two-tailed Student’s t-test (ns = p > 0.05; * = p < 0.05; ** = p < 0.01; *** = p < 0.001; **** = p < 0.0001)

Full size image

We next compared tumor development and survival rates of fl/fl Pdgfrb and Δ/Δ Pdgfrb mice. Strikingly, log-rank analysis of Kaplan-Meier survival curves indicated a 1.7-fold increase in survival for the Pdgfrb knockout mice (fl/fl Pdgfrb median survival: 19.3 weeks; Δ/Δ Pdgfrb median survival: 31.9 weeks) (Fig. 1E). Of note, both animal cohorts presented with identical thymic tumor phenotypes at the experimental end point (Fig. S1E), despite the total loss of Pdgfrb in tumor cells (Fig. 1F). These data suggest that PDGFRβ expression in tumor cells shortens the time to tumor development but does not prevent NPM-ALK driven lymphomagenesis. In addition to increased survival, Δ/Δ Pdgfrb mice had a decreased thymic tumor-to-body weight ratio (Fig. S1F), but spleen-to-body weight ratios were similar among the genotypes (Fig. S1G), consistent with an absence of splenic tumor development regardless of PDGFRβ status. Histopathological analysis of various organ parenchymas showed decreased dissemination of tumor cells to the kidney and liver in Δ/Δ Pdgfrb mice, whereas no significant differences were observed when comparing dissemination into the heart and lungs (Fig. 1G, S1H). Altogether, these results demonstrate that PDGFRβ expression facilitates ALK+ tumor formation and dissemination, and that our generated genetic mouse model is a valid tool to study the effects of PDGFRβ in ALK+ ALCL pathogenesis.

PDGFRβ promotes autocrine growth of ALK+ ALCL cells via IL-10 secretion

We next set out to address the mechanisms behind the observed prolongation in survival upon loss of PDGFRβ. Analysis of tumors at the experimental end point did not reveal any differences in the percentage of proliferating cells (Fig. S2A). Hence, we investigated proliferation capacity over time, as analysis of established tumors is a static observation and as such not representative of the process of lymphomagenesis. To do so, we performed in vitro proliferation assays seeding primary tumor cells at a low cell density (1 × 105 cells). In this restricted growth environment, we observed decreased proliferation for Δ/Δ Pdgfrb primary tumor cells (Fig. 2A), consistent with delayed tumor development in Δ/Δ Pdgfrb mice observed in vivo (Fig. 1E). Comparable to primary tumor cell lines, CRISPR/Cas9 knock-out of PDGFRβ in ALK+ cells (Fig. S2B) also led to a decrease in proliferation when seeded at a low cell density (Fig. 2B). We next inoculated immunodeficient (NOD.Cg-Prkdcscid, Il2rdtm1Wjl/SzJ) mice with fl/fl Pdgfrb and Δ/Δ Pdgfrb primary tumor cells at low (1 × 105) and ten-fold higher (1 × 106) densities and monitored tumor initiation and development over time. Animals inoculated with a high density of fl/fl Pdgfrb cells reached the maximum tumor volume cut-off (2000 mm3) around 23 days post inoculation whereas animals inoculated with Δ/Δ Pdgfrb cells were sacrificed around 26 days post inoculation (Fig. S2C). We performed longitudinal analyzes of the tumor growth curves over the entire duration of the experiment and observed no difference in tumor volume between NSG mice inoculated with high density fl/fl Pdgfrb and Δ/Δ Pdgfrb cells (Fig. S2D). However, when NSG animals were inoculated with a low density of Δ/Δ Pdgfrb cells, we observed a significant delay in tumor growth (Fig. 2C) resulting in an average of 38 days to reach the experimental end point, compared to 29 days for the PDGFRβ expressing tumors (Fig. 2D).

Fig. 2
figure 2

PDGFRβ promotes autocrine growth in ALCL. A Live cell counts of (n = 4) fl/fl Pdgfrb (grey) and (n = 4) Δ/Δ Pdgfrb (orange) primary mouse tumor cell lines seeded in 6-well plates at low confluency (1 × 105 cells). Cells were stained with Trypan blue and live cell counts were recorded on days two, four and six. B Representative pictures of a fl/fl Pdgfrb cell line (PdgfrbEV; empty vector transduced) subjected to CRISPR/Cas9 deletion of Pdgfrb (PdgfrbΔCRISPR) and seeded in a 96-well plate for a limiting dilution assay. Right graph: arbitrary fluorescence units measured using a resazurin assay were measured and correlated to the matched empty vector control (n = 5). C Longitudinal analysis of tumor volume increase of 6-week old immunodeficient (NOD.Cg-Prkdcscid, Il2rdtm1Wjl/SzJ) female mice inoculated with either (n = 8) fl/fl Pdgfrb (grey) or (n = 8) Δ/Δ Pdgfrb (orange) primary mouse tumor cell lines at a low concentration (1 × 105 cells/flank). D Kaplan Meier cumulative survival analysis of 6-week old NSG female mice inoculated with either (n = 8) fl/fl Pdgfrb (grey) or (n = 8) Δ/Δ Pdgfrb (orange) primary mouse tumor cell lines at a low concentration (1 × 105 cells/flank). Values next to the dotted lines on the x-axis indicate median life expectancy before tumor size reaches the 2000 mm3, the set threshold volume. E Heatmap depicting cytokine concentrations (pg/ml) in the supernatants of (n = 3) fl/fl Pdgfrb and (n = 3) Δ/Δ Pdgfrb primary mouse tumor cell lines cultivated in vitro at a low density (1 × 105 cells). F Quantitative RT-qPCR of Il-10 mRNA transcripts from (n = 8) fl/fl Pdgfrb (grey) and (n = 6) Δ/Δ Pdgfrb (orange) primary mouse tumor cell lines. Data was normalized to 18S ribosomal RNA expression. G Gene comparison between (n = 6) ALCL patients and healthy donors of the Piccaluga dataset [32] extrapolated from the Oncomine database [34]. Heatmap colors represent z-scores normalized to depict relative values within rows, where blue represents least expressed and red represents most expressed. A, B and F Data are shown as means ± SD, and p values were determined by the unpaired two-tailed Student’s t-test. D p value was determined by log-rank (Mantel-Cox) test (ns = p > 0.05; * = p < 0.05; ** = p < 0.01; *** = p < 0.001; **** = p < 0.0001)

Full size image

Considering the role cytokines play in maintaining homeostatic T cell survival and proliferation, we hypothesized a change in the cytokine profile may be responsible for the observed effects. ProcartaPlex cytokine ELISA analysis indeed revealed that IL-10 was significantly reduced in the supernatant of Δ/Δ Pdgfrb tumor cells cultured in vitro (Fig. 2E, S2E), as were transcript levels of IL-10 (Fig. 2F). In contrast, IL-19, a cytokine of the IL-10 super-family, was detected at higher levels in the supernatant of Δ/Δ Pdgfrb tumor cells, perhaps compensating for the lack of IL-10 (Fig. 2E, S2F). Interestingly, the six ALCL patients with upregulated PDGFRB expression [32] (Fig. 1B) also showed a downregulation of IL-19 and an upregulation of IL-10 mRNA levels in tumor cells compared to healthy control cells (Fig. 2G). Finally, using genome-wide DNA methylation data [35], we identified IL-10 hypomethylation in five ALK+ ALCL patients when compared to peripheral blood-derived activated CD3+ T cells from five healthy donors (Fig. S2G). In contrast, human control cell lines showed DNA hypermethylation of the IL-10 locus. In summary, our data suggests PDGFRβ orchestrates autocrine signaling in ALK+ ALCL cells that is sensitive to cellular density and IL-10 secretion.

PDGFRβ activates STAT5 in ALK+ ALCL

It has recently been shown that resistance to ALK inhibition via crizotinib is mediated by aberrant upregulation of IL-10RA rewiring the STAT3 signaling pathway in ALCL. STAT3 consecutively binds to the promoters of IL-10, IL-10RA and IL-10RB, maintaining oncogenic signaling regardless of NPM-ALK phosphorylation [6]. As we detected elevated IL-10 levels in primary tumor cells expressing PDGFRβ, we hypothesized that overexpressed PDGFRβ might additionally fuel STAT3 activity in ALCL. Supporting this hypothesis, end stage tumors developing in Δ/Δ Pdgfrb animals indeed exhibited a decrease in both active STAT3 and STAT5 levels when compared to those from fl/fl Pdgfrb mice (Fig. 3A). Because both PDGFRβ and NPM-ALK are potent tyrosine kinases, we produced kinase dead (KD) versions of NPM-ALK (K210R or D309A) or PDGFRβ (K634A or D826A) via Site Directed Mutagenesis to further understand the phosphorylation cascade (Fig. S3A). PDGFRβ and NPM-ALK in either wild type (WT) or KD forms were expressed in HEK293FT cells and as expected, neither of the KD versions were phosphorylated on the indicated tyrosine residues normally associated with activity (Fig. S3B). As anticipated, WT NPM-ALK induced STAT3 phosphorylation [36] (Fig. 3B, S3C). In contrast, PDGFRβ selectively induced STAT5 phosphorylation (Fig. 3B, S3D), suggesting that NPM-ALK and PDGFRβ have preferential STAT3 or STAT5 tyrosine kinase substrates. This was confirmed by imatinib inhibition of PDGFRβ recombinant protein activity, which prevented STAT5 phosphorylation in an in vitro kinase assay (Fig. 3C). In addition, the fl/fl Pdgfrb cell lines subjected to CRISPR/Cas9 deletion of PDGFRβ similarly showed a dampening of phosphorylated and total STAT5 (Fig. 3D). Finally, co-staining for PDGFRβ and STAT5 indicated both factors are expressed in tumor cells and do not originate from different cell populations (Fig. S3E).

Fig. 3
figure 3

PDGFRβ activates STAT5. A Western blot showing protein levels of PDGFRβ, phospho (p) STAT3, total STAT3, phospho (p) STAT5 and total STAT5 in thymomas excised from (n = 4) fl/fl Pdgfrb and (n = 4) Δ/Δ Pdgfrb mice at the experimental end point. GAPDH serves as the loading control. The molecular weight of analyzed proteins in kiloDaltons (KDa) is shown on the left. B Western blot showing protein levels of PDGFRβ, phospho (p) NPM-ALK, total NPM-ALK, phospho (p) STAT5, total STAT5, phospho (p) STAT3 and total STAT3 in HEK293FT transfected cell lysates. GAPDH serves as the loading control. The molecular weight of analyzed proteins in kiloDaltons (KDa) is shown on the left. C Western blot showing phosphotyrosine-100 (pTyr-100) levels following a kinase assay with PDGFRβ and STAT5 recombinant (r) proteins. The kinase assay was performed in the absence (−) or presence (+) of ATP and 5 μM of Imatinib. The highlighted bands indicate phospho (p) PDGFRβ and (p) STAT5. Total STAT5 serves as the loading control. D Western blot showing protein levels of PDGFRβ, phospho (p) STAT5 and total STAT5 in two fl/fl Pdgfrb primary tumor cell lines subjected to CRISPR/Cas9 mediated deletion of Pdgfrb. GAPDH serves as the loading control. The molecular weight of analyzed proteins in kiloDaltons (KDa) is shown on the left. E Representative pictures of PDGFRβ, phospho-STAT5 and total STAT5 IHC analysis and grading scheme to quantify staining of tissue microarrays from human ALCL tumor samples. F PDGFRβ expression levels were divided into either high (n = 7; grey) or low (n = 13; orange) and correlated to total STAT5 as detected by IHC. G Correlation between phospho (p) STAT5 and total PDGFRβ IHC scores of ALCL tissue specimens (n = 16). Pearson correlation, p = 0.03 and DFn = 1.14. F data is shown as the mean ± SD, and the p value was determined by the unpaired two-tailed Student’s t-test (ns = p > 0.05; * = p < 0.05; ** = p < 0.01; *** = p < 0.001; **** = p < 0.0001)

Full size image

We next correlated PDGFRβ and STAT5 expression in ALCL patient biopsies by dividing samples into two cohorts according to PDGFRβ high and low IHC-scores (Fig. 3E). Interestingly, patients with a high PDGFRβ IHC-score also had high STAT5 expression levels (Fig. 3F). We analyzed a second cohort of 16 ALCL patient samples to validate our findings and found a significant positive correlation between PDGFRβ expression and active levels of STAT5 (Fig. 3G), supporting our in vitro findings.

Inhibiting the PDGFRβ-STAT5 axis induces apoptosis of ALK+ ALCL cells

STAT5 plays an essential role in maintaining self-renewal capacities of hematopoietic cells via the tight regulation of downstream targets such as Granzyme B, Perforin, Osm, Hif2a, and Bcl-2 family members [37, 38]. As STAT5 has been reported to have anti-apoptotic functions in hematopoietic cells [39], we next assessed apoptosis in the murine tumors. IHC analysis for Cleaved Caspase 3 (CC3) revealed an increase in clusters of apoptotic cells in Δ/Δ Pdgfrb tumors compared to fl/fl Pdgfrb controls (Fig. 4A). This increase in apoptotic clusters correlated with a decrease in levels of STAT5 (Fig. S4A) and its anti-apoptotic target Bcl-xL (Fig. 4B). The ALCL patient biopsies, previously identified to have high PDGFRβ IHC-scores, also had elevated Bcl-xL levels (Fig. 4C, S4B), indicating that our murine model mirrors the molecular signatures observed in human patients.

Fig. 4
figure 4

Inhibiting the PDGFRβ-STAT5 axis induces apoptosis. A Representative pictures of cleaved caspase 3 (CC3) IHC analysis of (n = 19) fl/fl Pdgfrb and (n = 20) Δ/Δ Pdgfrb mouse thymomas at the experimental end point. Black squares in the left pictures represent the area chosen for the magnification depicted on the right. Scale bars: 50 μm. Whole-slide scans were quantified using Definiens™ software (right graph). B Representative pictures of Bcl-xL IHC analysis of (n = 18) fl/fl Pdgfrb and (n = 11) Δ/Δ Pdgfrb mouse thymomas at the experimental end point. Black squares in the left pictures represent the area chosen for the magnification depicted on the right. Scale bars: 50 μm. Whole-slide scans were quantified using Definiens™ software (right graph). C PDGFRβ expression levels were divided into either high (n = 7; grey) or low (n = 13; orange) and correlated to Bcl-xL IHC of patient primary tumor samples. D BCL2, MCL1, BCL2L1 (Bcl-xL) gene expression comparison between (n = 6) ALCL patients and healthy donors of the Piccaluga dataset [32] extrapolated from the Oncomine database [34]. Heatmap colors represent z-scores normalized to depict relative values within rows, where blue represents least expressed and red represents most expressed. E Normalized ChIP-seq alignment track for STAT3 and STAT5 (unstimulated or stimulated with either IL-2, or IL-21) in murine CD4+ T cells compared to IgG, shown at the gene loci for Bcl2l1 (Bcl-xL). Read densities (y-axis) were normalized to reads per million reads sequenced in each sample. F Representative FACS plots of fl/fl Pdgfrb and Δ/Δ Pdgfrb murine primary tumor cells co-stained with 7-aminoactinomycin D (7AAD) and Annexin V. The right graph indicates the percentage of necrotic, living, early or late apoptotic cells in both genotypes as indicated. G Western blot showing protein levels of CC3 and PDGFRβ in a fl/fl Pdgfrb primary tumor cell line following AC-4-130 treatment. GAPDH serves as the loading control. The molecular weight of analyzed proteins in kiloDaltons (KDa) is shown on the left. A, B, C, D and E Data are shown as means ± SD, and p values were determined by the unpaired two-tailed Student’s t-test (ns = p > 0.05; * = p < 0.05; ** = p < 0.01; *** = p < 0.001; **** = p < 0.0001)

Full size image

To confirm the importance of Bcl-xL in ALCL, we made use of the Piccaluga Lymphoma dataset [32] and observed significant upregulation of BCL2L1 (Bcl-xL) expression when compared to healthy donor T cells, whereas BCL2 and MCL1 were not significantly upregulated (Fig. 4D). These data suggest that STAT5 drives cell survival by activation of pro-survival pathways preventing apoptosis. Indeed, by re-analyzing several publicly available ChIP-Seq datasets derived from murine and human cell lines or ex-vivo models, we identified binding of both active STAT3 and STAT5 to the promoter of Bcl-xL (Fig. 4E, S4C and D). As a biological read-out, we performed a 7AAD/Annexin V FACs co-staining and identified higher levels of apoptosis in our primary tumor cells (Fig. 4F, S4E), whereas the cell cycle was not significantly affected (Fig. S4F). Western blot analysis revealed that inhibition of STAT activity using the SH2 domain small molecular weight inhibitor AC-4-130, successfully induces apoptosis via CC3 (Fig. 4G, S4G). Our data suggest that genetic deletion or inhibition of the PDGFRβ-STAT5 axis induces a CC3-mediated apoptotic phenotype in ALK+ ALCL.

STAT5A and STAT5B are essential for unrestricted cell proliferation

To determine whether the two STAT5 gene products mediate apoptosis to the same extent on a PDGFRβ+ background, we proceeded to delete either STAT5A, STAT5B or both genes simultaneously (Fig. S5A). CRISPR/Cas9 deletion of either Stat5a or Stat5b resulted in complete gene knock-out within the bulk population. On the contrary, the guide RNA designed to target both Stat5a/b only resulted in a knockdown of either gene product within the bulk population (Fig. 5A). Knockdown of Stat5a/b resulted in a decrease in proliferation similar to that achieved following single Stat5a or Stat5b knockout (Fig. 5B), suggesting at least one of the two gene products is necessary for survival. To confirm this hypothesis, we next attempted to isolate single clones from Stat5a/bΔCRISPR bulk populations. Sanger sequencing coupled with Western blot analysis suggested simultaneous deletion of Stat5a/b is lethal for ALK+ ALCL cells as we never obtained a clone with a complete deletion (Fig. S5B).

Fig. 5
figure 5

STAT5A/B are essential for unrestricted proliferation. A Western blot showing protein levels of phospho (p) STAT5 and total STAT5 in fl/fl Pdgfrb following CRSIPR/Cas9 mediated knockout of either STAT5A, STAT5B or both genes. GAPDH serves as the loading control. The molecular weight of analyzed proteins in kiloDaltons (KDa) is shown on the left. B Heatmap depicting cell viability of fl/fl Pdgfrb primary tumor cells following CRSIPR/Cas9 deletion of either STAT5A, STAT5B or both genes. Cells were seeded in a 96-well plate in limiting dilutions and arbitrary fluorescence units were measured using a resazurin assay. C Representative FACS plots of Stat5aΔCRISPR, Stat5bΔCRISPR, Stat5a/bΔCRISPR, Stat3ΔCRISPR or MybΔCRISPR cells 39 days post-transduction. The left graphs represent ‘Count vs. mCherry’ and the right graphs represent ‘GFP (FITC-A channel) vs. mCherry (ECD-A channel)’. D Heatmap representing the survival of Lenti-EF1As-Cas9-P2A-GFP and U6-IT-mPgk-mCherry vector expressing cells over time. Viability was calculated as the percentage of mCherry+ cells relative to the negative non-targeting control (RosaΔCRISPR) for each condition on day 9. MybΔCRISPR was used as a positive control. E Cell viability of Stat5a/bΔCRISPR double knock out cells normalized to individual Stat5aΔCRISPR cells over time. Data is plotted as the percentage of mCherry+ cells relative to Day 9 post-transduction. F Cell viability of Stat5a/bΔCRISPR double knock out cells normalized to individual Stat5bΔCRISPR cells over time. Data is plotted as the percentage of mCherry+ cells relative to Day 9 post-transduction. B, D, E and F Data are shown as means ± SD

Full size image

To track the effects of the double deletion of Stat5a/b over time we next performed a CRISPR/Cas9-based competition assay (Fig. S5C). We firstly transduced our primary tumor cell line derived from a fl/fl Pdgfrb mouse with a vector expressing SpCas9 and GFP. Following validation, we transduced the stably SpCas9-GFP-expressing clones with a vector expressing our sgRNAs of interest and mCherry. A competition assay was used to monitor competing growth kinetics of sgRNA-expressing mCherry and SpCas9-GFP positive (mCherry+) cells versus non-targeting SpCas9-GFP (GFP+) cells. Deletion of either Stat5a, Stat5b or Stat3 did not result in a significant decrease in viability indicated by the elevated expression of mCherry+ cells 39 days post transduction (Fig. 5C, S5D). On the contrary, the double deletion of Stat5a/b resulted in a decrease in mCherry+ vs GFP+ cells, suggesting the loss of both gene products impacts viability (Fig. 5D). Normalizing the percentage of Stat5a/bΔCRISPR mCherry+ cells to either Stat5aΔCRISPR or Stat5bΔCRISPR indicated no selective preference for either gene product over time (Fig. 5E, F). These data suggests that loss of Stat5a/b is lethal for PDGFRβ+ ALK+ ALCL cells, providing a rational for the use of STAT inhibitors.

Targeting STAT3/5 in vivo is therapeutically relevant in ALK+ ALCL

Having identified the PDGFRβ-STAT5 axis as an important oncogenic driver in ALCL, we next assessed the efficacy of the STAT inhibitor AC-4-130 in vivo. The original publication that identified AC-4-130 as a selective STAT5 inhibitor, postulated AC-4-130 might also influence STAT1 and STAT3 activity [40]. Due to the dependance of the JAK/STAT signaling pathway in ALCL we hypothesized AC-4-140 would be an attractive therapeutic strategy in our PDGFRβ+ ALCL model. We engrafted fl/fl Pdgfrb and Δ/Δ Pdgfrb tumor cells subcutaneously into 6-week old NSG mice. Once tumors reached palpable dimensions, mice were treated daily with either vehicle (10% DMSO, 5% Cremophore in PBS) or single agent AC-4-130 (25 mg/kg) (Fig. S6A). AC-4-130 treatment administered via intraperitoneal injection was well tolerated with no significant decrease in animal weight nor toxicity (Fig. S6B, C and [40]). Longitudinal analysis of tumor volume over time indicated that Δ/Δ Pdgfrb tumors grew at a significantly slower rate than fl/fl Pdgfrb tumors, mirroring the phenotype observed in the transgenic mouse model (Fig. 1E). Strikingly, single agent AC-4-130 treatment led to a delay in tumor growth in both experimental cohorts relative to vehicle controls (Fig. 6A). All experimental animals were sacrificed when the fl/fl Pdgfrb vehicle control tumors reached 2000 mm3 in size. At the experimental end point, AC-4-130 treated mice, in both cohorts, showed a significant decrease in tumor-to-body weight ratio (Fig. 6B). This experiment was repeated with a second set (biological replicates) of fl/fl Pdgfrb and Δ/Δ Pdgfrb tumor cells. Consistently, treatment with AC-4-130 resulted in a significant decrease of tumor volume (Fig. S6D) and in tumor-to-body weight ratio (Fig. S6E). Thus, our data suggest that STAT3/5 inhibition is a rational therapeutical option for ALK+ ALCL irrespective of PDGFRβ expression status.

Fig. 6
figure 6

Targeting STAT5 in vivo is therapeutically relevant for ALK+ ALCL. A The tumor volume (mm3) increase of (n = 3) fl/fl Pdgfrb and (n = 3) Δ/Δ Pdgfrb inoculated 6-week old female NSG mice treated with either vehicle or AC-4-130. B Representative macroscopic pictures of excised left and right flank tumors. The right graph represents tumor weight to body weight ratio (%) at the experimental end point. C Representative pictures of the H&E analysis of (n = 3) fl/fl Pdgfrb and (n = 3) Δ/Δ Pdgfrb inoculated 6-week old female NSG mice treated with either vehicle or AC-4-130. The black squares in the left pictures represent the area chosen for the magnification depicted on the right. Scale bars: 50 μm. D Representative pictures of Ki67 and PDGFRβ IHC analysis of (n = 3) fl/fl Pdgfrb and (n = 3) Δ/Δ Pdgfrb inoculated 6-week old female NSG mice treated with either vehicle or AC-4-130. The black squares in the left pictures represent the area chosen for the magnification depicted on the right. Scale bars: 50 μm. Whole-slide scans were quantified using Definiens software (right graphs). E Western blot showing protein levels of phospho (p) STAT3, total STAT3, phospho (p) STAT5 and total STAT5 of end point tumors excised from (n = 3) fl/fl Pdgfrb and (n = 3) Δ/Δ Pdgfrb inoculated 6-week old female NSG mice treated with either vehicle or AC-4-130. GAPDH serves as the loading control. The molecular weight of analyzed proteins in kiloDaltons (KDa) is shown on the left. Phospho (p) STAT3 levels over total STAT3 and phospho (p) STAT5 levels over total STAT5 are depicted as relative volume in %. B, D, and E Data are shown as means ± SD, and p values were determined by an unpaired two-tailed Student’s t-test (multiple t-test application on GraphPad) (ns = p > 0.05; * = p < 0.05; ** = p < 0.01; *** = p < 0.001; **** = p < 0.0001)

Full size image

Residual masses following AC-4-130 treatment were harvested and morphologically analyzed via IHC. Macroscopically, xenografted tumors recapitulated the histological features of ALK+ tumors derived from Cd4-NPM-ALK transgenic mice. H&E staining revealed small, round monomorphic cells with characteristically necrotic/apoptotic regions in the treated tumors (Fig. 6C). Consistent with treatment efficacy, immunostaining for Ki67 revealed a significant decrease in proliferating lymphoma cells upon AC-4-130 treatment (Fig. 6D). As previously reported, AC-4-130 treatment also resulted in a decrease in PDGFRβ expression (Fig. 6D and [40]). Western blot analysis of excised tumors indicated that AC-4-130 treatment decreases the activity of STAT5 and STAT3 in vivo (Fig. 6E and S6F), potentially due to the interwoven signaling network between NPM-ALK/STAT3 and PDGFRβ/STAT5.

Discussion

Imatinib has previously been identified as an effective treatment for PDGFRβ+ ALK+ ALCL relapse patients, however the mechanisms of action and the factual contribution of PDGFRβ remained unanswered [26]. We show here that T cell-specific deletion of PDGFRβ in a genetically engineered mouse model mimicking human ALCL in vivo, leads to delayed tumor growth and prolonged survival. To decipher the central signaling executor amongst the large repertoire of downstream targets activated by PDGFRβ, we used a tool kit of in vitro and in vivo models and identified that PDGFRβ signals preferentially via STAT5. The activation of the PDGFRβ/STAT5 axis in tumor cells results in a malignant boost by increasing autocrine-dependent proliferation and evading apoptosis. Treatment with the STAT5 SH2-domain inhibitor AC-4-130, results in a reduction in tumor growth in ALK+ ALCL, irrespective of PDGFRβ expression. Interestingly, we observed a decrease in both STAT3 and STAT5 following in vivo AC-4-130 treatment. This effect on STAT3 and STAT5 was already observed by Wingelhofer et al., who discussed that AC-4-130 targets first cellular STAT5 at pharmacologically relevant concentrations while having smaller effects on STAT3 and STAT1 [40]. We therefore suggest these “smaller effects” are accentuated in diseases such as ALCL that centrally rely on STAT3 activity to maintain malignancy. Our findings suggest that STAT5 plays a crucial role in ALCL oncogenesis and that inhibition of the PDGFRβ/STAT5 axis, but also of the NPM-ALK-STAT3 axis, both upstream via imatinib/ALK inhibitors and downstream with AC-4-130, is therapeutically relevant in ALK+ ALCL.

Murine thymic lymphoma cells lacking PDGFRβ proliferate at slower rates both in vitro (Fig. 2A) and when implanted subcutaneously into NSG mice (Fig. 2C). In contrast to ALK+ T cells lacking Tyk2 [8], deletion of PDGFRβ resulted in reduced proliferation suggesting that PDGFRβ acts as a malignant boost. Cytokine screening indicated that IL-10 is significantly downregulated in Δ/Δ Pdgfrb primary tumor cell lines providing an explanation for the observed density-dependent proliferation defect. IL-10 is one of the most abundant cytokines in peripheral blood of children diagnosed with ALCL [41] and it is readily detected in both systemic and cutaneous ALCL [42]. Our data echoes previous findings which demonstrated that IL-10 is central for TYK2-mediated STAT1 and STAT3 activity in ALCL [8]. More recently, the IL-10R was found to directly activate STAT3, bypassing NPM-ALK, to bind to the promoters of IL-10, IL-10RA and IL-10RB [6]. Our findings, which are supported by two independent lymphoma patient datasets (Fig. 1B), propose PDGFRβ is yet another potent kinase in the list of regulators of IL-10 expression in ALCL and that targeting the PDGFRβ/STAT5/IL-10 axis is an attractive therapeutic strategy. In parallel to the observed decrease of IL-10, we detected an increase in secreted IL-19 upon loss of PDGFRβ (Fig. 2E). IL-19 is a member of the IL-10 cytokine superfamily and forms part of a gene cluster alongside IL-10 and IL-20 located on chromosome 1q32 [43, 44]. Interestingly IL-19, which has been reported to play both pro- and anti-inflammatory roles, induces STAT1 and STAT3 activation via binding to IL-20Rα and IL20-Rβ [44, 45]. It is thus attractive to speculate that in the absence of PDGFRβ, alternative pathways capable of mediating STAT phosphorylation, such as IL-19/IL-20Rα/β-STAT3 node, could compensate for decreased STAT activity.

To date STAT3 [46] dominates the spotlight of ALCL research although recent findings have begun to unravel the important involvement of STAT1 [8] and STAT5 [47]. In particular, STAT5A/B exhibit important functions in the pathogenesis of hematological neoplasias due to their involvement in lymphoid and myeloid cell differentiation [38, 48,49,50]. In human ALK+ ALCL cell lines, the two different STAT5 gene products were identified to play divergent functions [9]. Nevertheless, our data indicate both STAT5A and STAT5B are equally expressed in PDGFRβ+ ALK+ ALCL. Deletion of either Stat3, Stat5a or Stat5b did not result in cell death nor a detectable upregulation of other Stat family members, advocating for a defined Stat function during ALK+ ALCL development. However, we did not achieve a complete knockout of both Stat5a/b gene products simultaneously, suggesting that at least one of these remained active and was sufficient to maintain cellular homeostasis (Fig. 5D, S5B). In contrast to peripheral T cell leukemia/lymphoma (PTCL) [48, 51, 52], we did not observe a selective preference for maintenance of either Stat5a or Stat5b in malignant ALK+ cells expressing PDGFRβ. Kinetic tracking of viability, using mCherry expression, indicated knockdown of both Stat5a/b resulted in a significant decrease in cell viability, however this population was not completely lost over time, hinting towards sequential activation of STAT3/5 activity in ALK+ ALCL.

The elevated levels of STAT3/5 in ALK+ patient samples, renders them attractive therapeutic targets as blocking would preferentially affect malignant lymphocytes while leaving homeostatic lymphopoiesis unaffected. In vivo and in vitro AC-4-130 treatments resulted in an induction of apoptosis mediated via CC3 and Bcl-xL in ALK+ lymphocytes. Interestingly, we found that STAT3 binds to super-enhancers that regulate high expression of Bcl-xL in both ALK+ and ALK− ALCL lymphoma cell lines [53] (Fig. S4C). Further in silico investigations revealed binding of STAT3 and STAT5 in murine primary CD4+ T cells and hematopoietic precursor cells (HPC7) [54, 55], suggesting a potential co-dependence at a transcriptional level (Fig. 4E, S4D). We postulate that this mode of action might be maintained in other tumors expressing elevated PDGFRβ levels such as cervical cancer [56] and castration-resistant prostrate cancer [57]. Interestingly, STAT3 and STAT5 hyperactivation has also been reported in these models [58, 59], rendering them potential druggable PDGFRβ targets. Extrapolating our findings from ALCL into a more general context, we hypothesize that multi-layered combinatorial treatments targeting the PDGFRβ-STAT5 signaling axis via imatinib upstream and AC-4-130 downstream might bolster response rates. Targeting single or multiple pathways at different hierarchical levels would in theory reduce the possibility of developing resistance to one specific drug. Nevertheless, future investigations into STAT small molecule inhibitors and degraders are urgently required to better understand the off-target effects.

In conclusion, we identify the PDGFRβ/STAT5 axis acts as a booster of malignancy, operating in parallel to the oncogenic NPM-ALK-STAT3 signaling cascade, resulting in a more aggressive ALCL disease entity. Our results highlight the importance of both STAT5A and STAT5B in ALK+ ALCL and suggest that blockade of STAT3/5 activity represents a new therapeutic strategy for the treatment of ALCL refractory patients. However, the use of STAT inhibitors will need to be rigorously assessed for their clinical applicability in the future. Thus, targeting activators upstream of the STATs, such as PDGFRβ and NPM-ALK, remains a highly relevant approach.