Introduction

Pancreatic ductal adenocarcinoma (PDAC) is a lethal disease with a 5-year survival rate of merely 10%–11%.1 Patients typically present with advanced or metastatic disease, making surgical removal impossible and highlighting the need for novel treatment options affecting established tumours.2 With disease incidence expected to rise in the coming years,3 quickly translatable results from preclinical studies are even more critical in order to improve patient outcomes in the near future. With respect to small-molecule inhibitors, promising strategies to serve this need are combination therapies or the repurposing of already-approved drugs or novel compounds already in clinical testing.

Single-cell transcriptomic profiling has shed light on the complex makeup of the disease-distinguishing tumour microenvironment (TME) and its adaptive behaviour to therapy.4 Specifically, PDAC is characterised by a potent immunosuppressive TME, typically excluding cytotoxic T cells from the tumour bed or driving them into functional exhaustion.5 6 Innate immunity, on the other hand, is abundant in the form of tumour-associated neutrophils (TANs) and tumour-associated macrophages (TAMs), thought to provide a potent immunosuppressive environment. Recent work has elucidated that the protumorigenic functionality of TAMs is also associated with their cell lineage, as PDAC harbours two subpopulations of TAMs: the embryonic yolk sac-derived macrophages and bone marrow-recruited monocytes or macrophages.7 In the mouse, the former Ly6clow cell subpopulation promotes fibrosis and PDAC progression, whereas the latter Ly6chigh subtype is enriched for major histocompatibility complex class II (MHCII) genes and antigen sampling, indicative of a benign or even tumour-suppressive phenotype.7 8 The presence of the latter TAM population might also imply that clinical approaches could potentially benefit from a therapeutically induced recruitment of monocytes from the systemic circulation. Phenotypically, TAMs have traditionally been classified into a continuum of states ranging from the MHCII, CD80/86, interleukin (IL)12high M1 state to a tumour-promoting MHCIIlow, IL10high M2 state.9 Reprogramming procancerous cells into tumoricidal TAMs has therefore been of intense interest, as evidenced by the development of, for example, CD40 agonistic antibodies activating antigen-presenting cells.10 Moreover, current translational ambitions not only aim at reprogramming TAMs towards an M1-like state but also facilitate cancer cell recognition by TAMs through antagonising endogenous innate immune checkpoints (‘don’t eat me’ signals) such as CD47, CD24 or B2M, thus promoting cancer cell engulfment.11 12

The dual specificity and tyrosine phosphorylation-regulated kinases (DYRKs) belong to an understudied group of kinases with potential therapeutic impact (the so-called ‘dark kinome’). This is particularly true for DYRK1B (a.k.a. MIRK),13–15 which has been implicated in the aetiology of several malignancies where it regulates a multitude of oncogenic traits.16–18 Despite DYRK1B behaving as an oncogene in many settings, it has also been described as a ‘quiescence kinase’ driving cell cycle exit in ovarian, colon and pancreatic cancer cells.19–21 It is assumed that this function is crucial for resistance against chemotherapy and for disease relapse. Elevated DYRK1B messenger RNA (mRNA) expression can be detected in up to 40% of patients with pancreatic cancer and also in PDAC mouse models.22 In approximately 9% of patients with PDAC, high DYRK1B protein levels are caused by genomic amplification of the DYRK1B gene.17

Here we show that DYRK1B promotes innate immune evasion and represents a promising drug target in PDAC. Although DYRK1B-depleted pancreatic cancer cells display enhanced cell proliferation in vitro, they are unable to support substantial tumour growth in vivo. Mechanistically, we could demonstrate that DYRK1B impacts the cancer cell secretome and profoundly regulates the immune TME. Here, DYRK1B acts on two levels of innate immune evasion. First, this kinase upregulates the surface expression of the self-recognition antigen CD24, protecting cancer cells from attack by phagocytic myeloid cells. Second, DYRK1B excludes macrophages from the tumour, and inhibition of DYRK1B not only results in massive recruitment of tumoricidal myeloid cells to the tumour bed but also increases the levels of phagocytosis of cancer cells and the enrichment of macrophage-activating Th1 cells. Based on these findings and on earlier work on a DYRK1B-mammalian target of rapamycin (mTOR) crosstalk,16 18 23 we trialled a well-tolerated drug combination therapy and found it to significantly extend the survival of animals harbouring highly aggressive autochthonous PDAC by more than twofold. As DYRK inhibitors are currently entering phase I evaluation, our data open new possibilities for timely clinical targeting of critical tumour–stroma interactions.

Materials and methodsCell lines

Mouse pancreatic adenocarcinoma cell lines (named mKpc) were derived from a spontaneous KrasG12D/+, Trp53R172H/+ and Pdx1-Cre (KPC) tumour. The presence of the expected mutations in Kras and Trp53 was verified by Sanger sequencing of PCR-amplified complementary DNAs (cDNA). Immortalised mouse pancreatic stellate cell (mPSC) lines were kindly provided by Albrecht Neesse.24 Human PSC1 (hPSC) cells were a kind gift of Matthias Löhr25; hPSC2 were kindly provided by Rosa Hwang.26 The Panc1 and AsPC-1 cell lines were purchased from Cell Line Service (CLS, Germany); MiaPaca-2 and PA-TU-8988T cells were obtained from DSMZ (Germany). BxPC-3 and SU86.86 cells were purchased from ATCC. All cell lines were cultured in Dulbecco’s Modified Eagle Medium (DMEM (high glucose plus glutamine and pyruvate), Invitrogen), supplemented with 10% foetal bovine serum (FBS; Anprotech, Germany) and 1% penicillin/streptomycin at 37°C with 5% carbon dioxide (CO2). All cells were regularly checked for mycoplasma contamination.

Chemical compounds

AZ191 was synthesised in-house according to a published protocol.27 KU0063794 was purchased from MedChemExpress MCE/MedChem Tronica and Everolimus from Biomol. Dimethyl sulfoxide (DMSO) was obtained from VWR or Applichem.

Generation of DYRK1B-KO/KD cell lines

The knockouts (KOs) of Dyrk1b in mKpc4 cells were generated according to a procedure described previously.24 For targeting murine Dyrk1b, we used a pU6-Cas9-based plasmid backbone (Addgene #64324) harbouring the following 20 bp guide sequences: CRISPR2-CGGGGCAGGAGCCGCACATC (targets exon 3 of isoform c) and CRISPR5-CTATGCGAAGAAGAAGCGGC (targets exon 4 of isoform c). Individual clones were harvested and screened for loss of DYRK1B in western blotting (WB), yielding clones 2.7 (CRISPR2) and 5.3 (CRISPR5). For human DYRK1B-knockdown (KD) generation, Panc1 cells were transfected with pLKO.1-puro plasmids containing one of the following targeting sequences (5′->3′): shCon-CAACAAGATGAAGAGCACCAA (non-targeting control), shDYRK1B_1-GACCTACAAGCACATCAATGA; shDYRK1B_3- CACGGAGATGAAGTACTATAT. Two days post-transfection, cells were treated with puromycin (2 µg/mL) until single clones appeared, which were subsequently picked, expanded and screened by WB.

Colony assays

Wild-type (WT) and KO mKpc4 cells were plated at a density of 500 cells per well in a six-well plate in DMEM supplemented with 1% penicillin/streptomycin and 10% FBS. After 2 days, cells were washed with Dulbecco’s phosphate-buffered saline (DPBS) and cultured for 6–8 days in 0.5% FBS-containing medium. Subsequently, cells were washed with DPBS and trypsinized, counted or stained with a crystal violet solution (0.2% crystal violet, 10% ethanol (EtOH) and 5% methanol (MeOH)) followed by several washing steps with water to remove excess staining solution. Afterwards, culture plates were air-dried and scanned, and the number of visible colonies was counted. For colony formation in the presence of drugs, the growth medium was supplemented with 0.2 µM KU0063794 or 0.2 µM Everolimus, and control cells received the solvent DMSO.

Bromodeoxyuridine (BrDU) staining

mKpc4 cells plated on six-well plates were serum-starved (0.5% FBS in DMEM) for 2 days, and 10 µM BrDU (Sigma) was added to the growing cells for 4 hours. Cells were trypsinised and harvested by centrifugation at 300g for 5 min. The cell pellet was resuspended in 200 µL of ice-cold DPBS, and 500 µL of ice-cold 70% EtOH was slowly added while the cells were vortexed. After fixation for 30 min on ice, cells were pelleted by centrifugation at 300g for 5 min. The cell pellet was resuspended in 500 µL of 2 M hydrochloric acid containing 0.5% Triton X-100 while vortexing. After incubation at room temperature (RT) for 30 min, the cells were centrifuged at 8000 rpm for 2 min. The cell pellet was resuspended in 500 µL of 0.1 M sodium borate (pH 8.5) to neutralise the acid and centrifuged at 8000 rpm for 2 min. Cells were stained for 30 min at RT in 150 µL of DPBS containing 1% bovine serum albumin (BSA)/0.5% Tween-20 and supplemented with 8 µL of anti-BrDU fluorescein isothiocyanate antibodies (BD Bioscience). Cells were washed once with DPBS containing 1% BSA/0.05% Tween 20, centrifuged and resuspended in 500 µL of DPBS with 1 µg/mL of propidium iodide and analysed by Cytoflex LX Series (Beckman Coulter, USA).

RNA/cDNA/quantitative reverse transcription (qRT)-PCR analysis

Total RNA was extracted using the NucleoSpin RNA II kit from Macherey-Nagel according to the manufacturer’s protocol. 0.5–1 µg of total RNA was used for cDNA synthesis performed with the iScript cDNA Synthesis Kit (BioRad). For quantitative PCR (qPCR) reactions, the Absolute QPCR SYBR Green Mix (Thermo Scientific) was used. qPCR reactions were performed on 96-well plates using the Mx3005P qPCR systems (Agilent). Relative expression was calculated according to the 2ΔΔCt method. For specific information on qPCR primer sequences, please refer to online supplemental file.

RNAseq

Cellular RNA was extracted using the Macherey-Nagel NucleoSpin RNA Isolation Kit, according to the manufacturer’s protocol. The integrity of the total RNA was assessed by capillary electrophoresis. RNAseq libraries were generated using Lexogen Quantseq according to the manufacturer’s protocol and sequenced on an Illumina NextSeq 550 using 75 bp single-end reads. Raw reads were aligned to the human genome (GRCh38) using STAR V.2.6.1d and quantified as counts per million against the protein-coding and lincRNA-coding genes as defined by Ensembl,28 revision 100. Statistical comparisons were performed via unpaired DESeq2.29 Functional annotation was performed with hypergometrical tests and Benjamini-Hochberg correction versus the Molecular Signatures Database.

Western blotting

Cells were lysed in 1× sample buffer; separation of lysates by sodium dodecyl sulphate polyacrylamide gel electrophoresis (Bio-Rad) and subsequent blotting on Immobilon-polyvinylidene difluoride membranes (Millipore) was done as described in Schneider et al,30 followed by incubation with the respective primary antibody. After incubation with a corresponding horseradish peroxidase (HRP)-coupled secondary antibody (Cell Signalling Technology), the HRP signal was detected using Pierce ECL Western Blotting Substrate (Thermo Scientific) according to the manufacturer’s protocol. For specific antibodies, see online supplemental file.

Isolation and differentiation of mouse bone-marrow-derived macrophages (BMDM)

Bone marrow was collected from the femur bones of C57Bl/6 mice by flushing with Roswell Park Memorial Institute Medium (RPMI)-1640 medium containing 1% penicillin/streptomycin and passing through a 70-micron cell filter. The cell suspension was centrifuged at 300g for 5 min, and the cell pellet was incubated in erythrocyte lysis buffer (155 mM ammonium chloride, 12 mM sodium bicarbonate and 0.1 mM EDTA) for 3 min at RT, followed by centrifugation at 300g for 5 min. Bone marrow cells were plated on a six-well plate and cultured in RPMI-1640 medium (with glutamine and pyruvate) supplemented with 10% FBS, 1% penicillin/streptomycin and 20 ng/mL of recombinant macrophage colony-stimulating factor (Immunotools, Germany). Cells were incubated for 4–5 days at 37°C in a humid atmosphere containing 5% CO2 to allow differentiation into macrophages.

Isolation of human peripheral blood mononuclear cells (PBMC) and differentiation into macrophages

Chambers of the leucoreduction system with blood from healthy adult volunteers were provided by the Center for Transfusion Medicine and Hemotherapy at the University Hospital Giessen and Marburg. Blood was carefully layered on 15 mL Ficoll (Capricorn Scientific, Germany), and the tube was centrifuged at 1080g for 20 min at RT (without break). After centrifugation, mononuclear cells were collected, resuspended in 40 mL of DPBS and centrifuged at 300g for 5 min at RT. The cell pellet was resuspended in 40 mL of DPBS and centrifuged again at 10g for 10 min at RT. Pelleted cells were resuspended in 10 mL of DPBS, counted and plated as follows: 3×106 cells per well in a six-well plate or 0.7×106 per well in a 24-well plate. Plated cells were cultivated in DPBS for 30 min in an incubator at 37°C in a humid atmosphere with 5% CO2 to enrich for monocytes. Attached cells were washed twice with DPBS to remove poorly attached cells and grown in RPMI-1640 medium supplemented with 5% human serum (Sigma, Germany), 1% penicillin/streptomycin, 2 mM GlutaMAX and 1 mM sodium pyruvate (Sigma) for 8 days at 37°C in a humid atmosphere containing 5% CO2 to allow differentiation into macrophages. Subsequently, differentiated macrophages were washed twice with DPBS and subjected to treatments with supernatant (SN) from human pancreatic cancer cells or used to evaluate phagocytic capability.

Collections of SN from tumour cells and macrophages

For SN production, mKpc4 cells were plated at a density of 2×105 cells per 10 cm culture dish and Panc1 cells at a density of 4×105 cells per 10 cm dish in DMEM supplemented with 1% penicillin/streptomycin and 10% FBS. Two days later, cells were washed with DPBS and the medium was replaced with 0.5% FBS-containing DMEM. Two days later, conditioned media (SN) was collected, centrifuged at 300g for 5 min to deplete cell debris and frozen at −80°C.

To collect conditioned media from macrophages, BMDM were differentiated for 5 days, washed twice with DPBS, treated with SN from mKpc4 cells in a ratio of 1:2 (1/3 volume of SN) and incubated for 24 hours. Control macrophages were incubated with the same proportion of medium containing 0.5% FBS. The next day, the cells were washed with DPBS and either lysed for further purification of total RNA or incubated with DMEM medium containing 2% FBS and 1% penicillin/streptomycin for an additional 72 hours. At this point, macrophages were counted and the conditioned medium was harvested, centrifuged at 300g for 5 min and frozen at −80°C for further analysis.

Human PBMC-derived macrophages were differentiated for 8 days, washed twice with DPBS, treated with SN collected from Panc1 cells in a ratio of 1:2 (1/3 volume of SN) and incubated for 24 hours. Control macrophages were incubated with the same proportion of medium containing 0.5% FBS. The next day, the cells were washed with DPBS and lysed for the purification of total RNA.

Allograft experiment

Mice were kept under specified pathogen-free (SPF) conditions in individually ventilated cages with a 12 hour/12 hour light–dark cycle and a standard altromin housing diet. For the experiment, female C57BL/6 mice were randomly divided into two groups (3–5 animals per cage). 1×105 WT or Dyrk1b-KO (clone 5.3) mKpc4 cells were suspended in a total volume of 150 µL of DMEM medium. The cell suspensions were injected subcutaneously at the posterior flank of the mice. The size of the tumours was measured two times per week using a calliper. Tumour volumes were calculated by the formula (length×width2)/2. The length represents the longer axis and the width represents the shorter axis of the tumour. At the experimental endpoint, mice were euthanised and tumours were removed. The study was approved by the regional agency on animal experimentation (Regierungspräsidium Giessen).

KPC mouse model and drug treatment

Mice were kept under SPF conditions in individually ventilated cages with a 12 hour/12 hour light–dark cycle and a standard altromin housing diet. To obtain KPC mice, LSL-KrasG12D/+ and LSL-Trp53R172H/+ mice were crossed with Pdx1-Cre mice, yielding triple-mutant mice: LSL-KrasG12D/+, LSL-Trp53R172H/+ and Pdx1-Cre (KPC). KPC animals of both sexes underwent weekly abdominal palpation starting at the age of 10 weeks. Once a tumour was identified by palpation, a small animal ultrasound was performed to measure its size. Pharmacological treatment was initiated when tumour size reached about 100–150 mm3, calculated by the formula (length×width2)/2. The length represents the longer axis and the width represents the shorter axis of the tumour. Gemcitabine (70 mg/kg; once weekly), AZ191 (5 mg/kg; two times per week) and KU0063794 (5 mg/kg; two times per week) were dissolved in 50% 2-hydroxypropyl-β-cyclodextran (Sigma)/DPBS and injected intraperitoneally. The control group received equal volumes of solvent. Tumour sizes were regularly determined by ultrasound until the animal had to be sacrificed. All studies were approved by the regional agency on animal experimentation (Regierungspräsidium Giessen).

Culture of murine PSCs in a three-dimensional (3D) matrix

PSCs (mPSC4, kindly obtained from Dr Albrecht Neesse) were seeded at a density of 6×105 in a 70 µL matrigel drop containing a 1:1 ratio of DMEM (10% FBS) and growth factor-reduced basement membrane matrix (Corning, Cat #356230) on a 3.5 cm suspension dish (Sarstedt). The matrigel-embedded mPSC4 was covered with 0.5% FBS-containing DMEM and incubated for 48 hours to obtain quiescent mPSC4 (qPSC). Thereafter, qPSC were incubated with 1/3 of the volume of SN from activated BMDM for 48 hours. After differentiation, qPCR-based phenotyping was performed. For this, medium was removed and drops were collected in ice-cold PBS and centrifuged at 1000 rpm for 5 min at 4°C. After removing PBS, Matrigel drops were intensively resuspended in ice-cold PBS and incubated for 30 min at 4°C. Cells were centrifuged (1000 rpm, 4°C, 5 min), and pellets were frozen at −80°C for later RNA preparation.

Immunohistochemistry

For immunohistochemistry of formalin-fixed paraffin-embedded (FFPE) tissue, heat-induced epitope retrieval was performed with EDTA. Staining was performed on a DAKO Autostainer-Plus. After blocking endogenous peroxidases, sections were incubated for 45 min with the respective antibody. Sections were washed and incubated with Dako REAL EnVision HRP Rabbit/Mouse polymer, which reacts with DAB-Chromogen, according to the manufacturer’s protocol. The use of patient material (in the form of tissue microarrays) was approved by the local ethics committee (Ethics Board University Hospital Marburg). The tissue used in this study was surgically resected primary tumour material obtained from patients with PDAC eligible for surgery.

Immunofluorescence (IF) of FFPE tissue and image analysis

For IF using fluorophore-conjugated antibodies, antigen retrieval of tissue sections was achieved by steam-heating in citrate solution (pH 6.0, Mophisto) for 30 min. Sections were then treated with 100 mM glycine for 10 min, washed with Tris-buffered saline with 0.1% Tween 20 detergent (TBS-T) and blocked with 10% goat serum in PBS containing 0.3% Triton for 60 min. Subsequently, sections were incubated overnight at 4°C with primary antibodies diluted in PBS supplemented with 5% goat serum and 0.3% Triton X100. After washing with TBS-T, samples were incubated with corresponding secondary antibodies for 60 min at RT, followed by washing and mounting.

The analysis of the images was done with Imaris 9.9.0 (Oxford Instruments). Macrophages were detected by the spot algorithm. The same was done for CD206 and tumour necrosis factor (TNF). For the analysis of CD206+ macrophages and TNF+ macrophages, the number of double-positive cells in all macrophages was determined. For Ki67, the surface algorithm was chosen, and the colocalized part of macrophages was counted. Pan-cytokeratin was also determined by the surface algorithm. In addition, the classification of the surfaces was selected to distinguish between surfaces with the shortest distance to Ki67 positive surfaces below 1 µm (Ki67+Cytokeratin+ cells) and Ki67−cytokeratin+ cells.

Phagocytosis assay

Human or mouse macrophages were plated on a 24-well plate and differentiated according to the protocol described above. After differentiation, macrophages were either treated for 24 hours with 1/3 volume of conditioned media from corresponding tumour cells or proceeded directly to the phagocytosis assay. One well of macrophages was used to count cells. Plated macrophages were stained for 20 min with CellTracker Green CMFDA (Invitrogen) diluted in plain RPMI medium according to the manufacturer’s instructions. Tumour cells were harvested, counted and an equal number (compared with macrophages) of tumour cells (mKpc4 or Panc1) were stained with CellTracker Deep Red (Invitrogen) diluted in plain RPMI according to the manufacturer’s instructions. Macrophages and tumour cells were washed with DPBS, and an equal number of tumour cells were plated into each well with macrophages. Single stained and unstained cells were also preserved as samples for gating. 14–18 hours later (overnight), cells from wells were trypsinised, harvested, resuspended in 200 µL of DPBS containing 1% BSA and 2 mM EDTA and fixed by the addition of 200 µL 4% paraformaldehyde. Unstained, single-only stained and mixed cells were analysed by flow cytometry using the Cytoflex LX Series (Beckman Coulter).

Macrophage migration

Each bottom of a well of a 24-well plate was loaded with 500 μL of SN from corresponding tumour cells and 500 μL of 0.5% FBS-containing DMEM. Control wells received 1 mL of 0.5% FBS-containing DMEM. Differentiated mouse or human macrophages were trypsinised and counted. Counted macrophages were centrifuged at 300g for 5 min, and the pellet was resuspended in DMEM containing 0.5% FBS. A 500-μL cell suspension with 1×105 macrophages was loaded onto a tissue culture insert with a pore diameter of 8 µM (Sarstedt). 24 hours later, media were aspirated, and cells were washed with DPBS. Cells on the upper surface of the insert were removed by a cotton bud, whereas cells from the bottom surface were stained with a crystal violet solution (0.2% crystal violet, 10% EtOH and 5% MetOH) for 2 hours. Subsequently, inserts were intensively washed with distilled water and air-dried. Inserts with stained cells were placed in a new 24-well plate containing 20% acetic acid in water, incubated on a shaker for 30 min, and the absorbance of the coloured solution was measured at 590 nm using an MWG-Biotech SpectraMax340 spectrophotometer.

Viability assay

Mouse BMDM were differentiated for 5 days and treated for 24 hours with 1/3 volume of SN from corresponding tumour cells. Control wells were treated with 1/3 volume of 0.5% FBS-containing DMEM medium. On the next day, macrophages were washed with PBS and cultivated for an additional 3 days in DMEM containing 0.5% FBS and counted using Neubauer counting chambers.

Surface staining of CD24

mKpc4 and Panc1 cells were trypsinised, harvested and counted. Then, 5×105 cells were resuspended in 100 μL of DPBS containing 1% BSA and antibodies blocking Fc-receptors (Miltenyi, Germany). After incubation for 10 min at RT, 1 µL of phycoerythrin (PE)-conjugated anti-CD24 antibodies (Biolegend) or PE-conjugated isotype control IgG (Biolegend) was added to the tube, and cells were stained for 15 min at 4°C. Cells were washed twice with DPBS containing 1% BSA and 2 mM EDTA and were then analysed using the Cytoflex LX Series (Beckman Coulter). The data were analysed using FlowJo software.

Flow cytometry of mouse tumours

After surgical removal, tumours were digested into single-cell solutions by 200 u/mL Collagenase IV (Worthington), 10 µg/mL DNAse I (Roche) in Hanks’ balanced salt solution, 37°C, 300 rpm agitation and filtered through 100-μm cell strainers (Sysmex). Subsequently, staining with fluorophore-conjugated antibodies was performed. The flow profiles were acquired on an Attune NxT Cytometer (Thermofisher Scientific) or a Cytoflex LX Series (Beckman Coulter).

Cytokine array

Conditioned media (SN) was collected as described above from parental mKpc4 cells and Dyrk1b-KO clone 5.3. Conditioned media was diluted with the provided dilution buffer 1:2 (1/3 volume of SN) and subjected to secretome analysis using the mouse XL cytokine array kit (R&D Systems, USA) according to the manufacturer’s instructions.

Soft agar colony formation assay

Prewarmed sterile 5% Bacto-Agar (Roth, Germany) was mixed 1:10 with warm DMEM containing 10% FBS and loaded on wells of a six-well plate to solidify. 1×104 WT/KO mKpc4 cells were resuspended in 250 µL of warm DMEM containing 10% FBS and mixed with 500 µL of warm agar prediluted as above.

Cells resuspended in agar were loaded on top of solidified agar and grown for 10 days. Formed colonies were counted using a light microscope.

Secretome proteomics

4×105 WT/KO mKpc4 cells were plated on 6 cm dishes in independent triplicates in DMEM containing 10% FBS. Two days later, cells were washed with DPBS, and the medium was changed to DMEM without serum. Conditioned medium was collected after 2 days, centrifuged at 300g for 5 min to deplete debris and snap-frozen in liquid nitrogen for further proteomics studies. Modifying a protocol by Chevallet et al,31 proteins were precipitated after the addition of cOmplete ULTRA Protease Inhibitor Cocktail without EDTA (Roche) as follows: in polypropylene tubes, samples were adjusted to final concentrations of 0.1% and 7.5% of N-lauroylsarcosine sodium salt and trichloroacetic acid, respectively, followed by incubation on ice for 2 hours and centrifugation at 10 000g for 10 min at 4°C. Protein pellets were washed twice using cold tetrahydrofuran (10% initial sample volume) and stored at −20°C. Further sample processing included solution digest, reductive dimethyl labelling and high pH reverse phase chromatographic separation as described, followed by liquid chromatography-tandem mass spectrometric analysis on a Q Exactive HF mass spectrometer as described.32 33 The MaxQuant suite of algorithms (V.2.0.1.0)34 was used to analyse mass spectrometric raw data against the murine Uniprot database (canonical and isoforms; downloaded 2021/02/08_8661807 entries). False discovery rates of 1% were used both at the peptide and the protein group levels. Instrument parametrisation was extracted and summarised using MARMoSET35 and along with the MaxQuant settings and mass spectrometric raw data, it has been deposited with the proteomeXchange consortium via the MASSive repository. The in-house R pipeline autonomics https://doi.org/doi:10.18129/B9.bioc.autonomics was used for downstream bioinformatic analysis.

Patient involvement

Tissue samples of patients with therapy-naive PDAC who underwent surgical resection were provided by the Biobank of the University of Marburg (CBBMR) in accordance with the regulations of the ethics committee of the University of Marburg (AZ 76–17).

Statistics and data accessibility

Statistical comparisons were made of n≥3 experiments using an unpaired two-tailed one-grouped Student’s t-test (using MS Excel or GraphPad Prism) unless otherwise indicated. Significances were indicated as ns (not significant; p>0.05), *p<0.05, **p<0.01 and ***p<0.001. Human Kaplan-Meier curves from public datasets were generated using the R2: Genomics Analysis and Visualisation Platform (http://r2.amc.nl). RNAseq data has been deposited at BioStudies/Array Express with the following accession codes: E-MTAB-13666 (Dyrk1b WT/KO in mKpc4 cells); E-MTAB-13667 (mouse allograft with Dyrk1b WT/KO mKpc4 cells) and E-MTAB-13668 (mouse BMDMs treated with SN from Dyrk1b WT/KO mKpc4 cells). Protein mass spectrometric raw data has been deposited with the proteomeXchange consortium via the MASSive repository.

ResultsThe DYRK1B kinase is enriched in PDAC tumour cells

The DYRK1B kinase can be detected in 12%–39% of all PDAC cases, and in 6%–9%, this is caused by a genomic amplicon encompassing the DYRK1B gene locus (online supplemental figure S1A).17 36 In order to validate these findings, we commenced our studies on DYRK1B in pancreatic cancer by first determining DYRK1B mRNA expression levels in bulk PDAC resection material (Marburg cohort; n=59 patients). As evident from figure 1A, DYRK1B mRNA levels were low but readily detectable and were significantly increased in the basal versus the classical subtype of PDAC (figure 1A). In patients with pancreatic cancer, high DYRK1B expression was associated with poorer overall survival compared with low-expressing patients (figure 1B; Yeh cohort; n=10237), a finding that was validated in an independent second public data set (online supplemental figure S1B; The Cancer Genome Atlas (TCGA) cohort; n=146 patients). Analysis of a recent scRNAseq study38 revealed DYRK1B transcript expression to be primarily confined to epithelial duct cells of the pancreas as well as, to a lesser extent, endothelial and endocrine cells, but not acinar, stellate or immune cells (online supplemental figure S1C,D). In agreement with this observation, DYRK1B protein expression was observed in several human PDAC cell lines, but levels were low in PSCs (figure 1C). In addition, immunohistochemistry of human PDAC tissue supported these results, with DYRK1B being preferentially enriched in the epithelial tumour compartment (figure 1D). In the mouse, Dyrk1b mRNA expression can readily be measured in bulk tissue derived from the well-established KPC model of PDAC39 as well as in epithelial cell-enriched tumour organoids (figure 1E). As shown for the human cell lines above, immunoblotting detected prominent DYRK1B protein levels in murine PDAC cell lines but only in marginal levels in mouse PSCs (figure 1F). Finally, mouse KPC tissue staining with a DYRK1B-specific antibody revealed clear enrichment of this kinase in the epithelial tumour compartment with only minute signals in the surrounding stroma (figure 1G and online supplemental figure S1E for antibody validation). Taken together, we could provide evidence that DYRK1B expression is preferentially restricted to tumour cells (and not the stroma) and negatively associated with overall survival in patients with PDAC.

Figure 1

DYRK1B is expressed in pancreatic cancer cells. (A) DYRK1B mRNA expression (in transcript per million) in bulk tissue derived from patients with PDAC (n=59). Subtype classification as described in Rashid et al.56 Each dot represents one patient (mean±SD). (B) Kaplan-Meier plot depicting overall survival of patients with PDAC37 in relation to DYRK1B expression (Yeh cohort, scan split and log-rank test). (C) DYRK1B immunoblot of human cell lines. Actin was used as a loading control. Shown is one representative blot of n=2. (D) DYRK1B immunohistochemistry of tissue microarrays derived from resected PDAC material. (E) Dyrk1b mRNA expression (in counts per million) from mouse KPC bulk tissue and from KPC-derived primary organoids. Each dot represents one animal or one organoid line. (F) DYRK1B Western blot of murine cell lines. Shown is one representative blot of n=2. (G) DYRK1B immunohistochemistry on mouse KPC tumour tissue. The scale bar is 100 µm. DYRK1B, dual specificity and tyrosine phosphorylation-regulated kinase 1B; mRNA, messenger RNA; PDAC, pancreatic ductal adenocarcinoma; S, tumour stroma area; T, epithelial tumour area; TCGA, The Cancer Genome Atlas.

DYRK1B suppresses PDAC cell proliferation in an mTOR-dependent manner

In order to shed more light on the precise role of DYRK1B in pancreatic cancer, we generated CRISPR/Cas9-mediated gene KO in murine PDAC cells (mKpc4) derived from the KPC mouse model39 harbouring mutant KrasG12D and p53R172H. Using two different CRISPR constructs targeting separate exons of the Dyrk1b gene, we identified two clones displaying complete (KO5.3) or almost complete (KO2.7) elimination of the DYRK1B protein (figure 2A), with no apparent changes in levels of the closely related paralogue DYRK1A (online supplemental figure S2A). In line with DYRK1B negatively regulating cell cycle components20 40 and the DREAM (DP, RB-like, E2F And MuvB) complex,41 Dyrk1b-KO cells grew significantly faster in culture than WT cells (figure 2B,C). Similar results were obtained in 3D culture using soft agar conditions (online supplemental figure S2B). KO cells displayed a higher rate of BrDU incorporation, showing that the increase in cell numbers was primarily caused by a faster cell cycle and not by enhanced survival (figure 2D and online supplemental figure S2C). Similar elevated incorporation was also observed upon treatment of Dyrk1b-WT cells with the small molecule DYRK1B inhibitor AZ19142 (figure 2E and S2D). In addition, Dyrk1b-depleted cells were particularly sensitive to inhibition of the pro-proliferative mTOR kinase by either Everolimus (mTORC1 inhibitor) or KU0063794 (dual mTORC1/2 inhibitor), completely abrogating the surplus in proliferation seen upon Dyrk1b loss (figure 2F,G). In agreement with the increased sensitivity against mTOR inhibition, KO cells displayed stronger mTOR and S6 activation compared with WT cells (online supplemental figure S2E).

Figure 2

DYRK1B suppresses PDAC cell proliferation. (A) Western blot depicting DYRK1B protein levels in parental mKpc4 cells (WT) and DYRK1BKO clones 5.3 (KO5.3) and 2.7 (KO2.7). Actin was used as a loading control. Shown is one representative blot of n=3. (B) Colony staining (crystal violet, blue) of Dyrk1b-WT and Dyrk1b-KO mKpc4 clones. Shown is one representative of n=6. (C) Relative cell number (%) of mKpc4 cells grown in clonal density. Mean of n=6±SD (one-tailed t-test). (D) The relative percentage of BrDU-positive Dyrk1b-WT/KO mKpc4 cells. Mean of n=3±SD. (E) The relative percentage of BrDU-positive mKpc4 cells treated with DMSO as a solvent or 1 µM AZ191 for 4 days. Mean of n=3±SD. (F) Relative cell number (%) of WT/KO mKpc4 cells treated with solvent (DMSO) or 0.2 µM KU0063794 or 0.2 µM Everolimus for 6 days. Mean of n=4–5±SD. (G) Colony stain (crystal violet, blue) of WT/KO mKpc4 cells treated with solvent DMSO or 0.2 µM KU0063794. Shown is one representative of n=5. (H) Transcriptome signatures (RNAseq) of KO mKpc4 cells in comparison to WT cells. (I) Relative mRNA expression of TGFβ-pathway genes in WT and KO mKpc4 cells as determined by quantitative reverse transcription-PCR. Shown is one representative of n=3, measured in triplicate (±SD). Asterisks indicate significance versus WT. (J) Representative phase-contrast images of WT/KO mKpc4 cells. The scale bar is 50 µm. BrDU, bromodeoxyuridine; DMSO, dimethyl sulfoxide; DYRK1B, dual specificity and tyrosine phosphorylation-regulated kinase 1B; KO, knockout; mRNA, messenger RNA; PDAC, pancreatic ductal adenocarcinoma; TGFβ, transforming growth factor beta; WT, wild-type.

Intriguingly, transcriptome analysis revealed an upregulation of transforming growth factor β (TGFβ)-related gene signatures in Dyrk1b-KO cells (figure 2H), which were validated using qRT-PCR (figure 2I). In addition, the morphology of both KO clones appeared more mesenchymal as compared with their Dyrk1b WT counterparts (figure 2J). In summary, we could show that loss of Dyrk1b results in the induction of TGFβ pathway genes and mTOR-dependent stimulation of proliferation in pancreatic cancer cells.

Ablation of tumour cell-Dyrk1b stimulates macrophage recruitment

In light of the cell cycle-suppressive impact of DYRK1B in cultured cells, we were interested to investigate the corresponding outcome of Dyrk1b loss in an in vivo setting. To this end, we subcutaneously transplanted Dyrk1b WT or KO mouse PDAC cells into syngeneic C57BL/6 animals (figure 3A). In complete contrast to the cell culture experiments described above, KO cells were significantly growth retarded in this setting and only formed very small tumours, whereas WT cells gave rise to large tumours (figure 3B). Similar findings were also made with the second KO clone (online supplemental figure S3A). In order to obtain further insight into the underlying mechanisms, we performed mRNA sequencing (RNAseq) of bulk tumour tissues. Intriguingly, functional annotation of genes significantly upregulated in KO versus WT tumours yielded several innate immune signatures, such as interferon signalling, lipopolysaccharide, TNF, polyI:C-RNA or IRF3 (figure 3C), suggesting KO tumours to harbour an altered immune microenvironment compared with WT. Downregulated signatures included many general cancer terms (eg, nasopharyngeal carcinoma) as well as signatures associated with cellular differentiation (eg, breast cancer_luminal vs mesenchymal) (online supplemental figure S3B).

Figure 3

Ablation of tumour cell-Dyrk1b stimulates macrophage recruitment and inhibits in vivo tumour growth. (A) Scheme outlining the allograft experiment. (B) Tumour size changes over time (subcutaneous allograft growth in C57BL/6 mice). Animals received either Dyrk1b-WT (black curve) or KO (clone 5.3; red curve) mKpc4 cells (n=12 each) on day 0. Shown is the mean±SEM. (C) Gene signatures (RNAseq) upregulated in the KO allograft (vs WT). (D) Representative histology of resulting tumours from B. The upper panel depicts H&E; other panels depict staining with corresponding antibodies (scale bar 150 µm in all panels). (E–I) Quantification of IHC staining intensity of slides from panel D. Each dot represents one tumour (relative IHC intensity, mean of n=4–5 ± SD (one-tailed t-test). (J) Immunofluorescent staining of allografted tumours. The scale bar is 30 µm. (K) Flow-cytometric quantification of F4/80-MHCII double-positive (M1) TAMs from tumours in panel B. Each dot represents one tumour (mean±SD). (L) Flow-cytometric quantification of F4/80-CD206 double-positive (M2) TAMs from tumours in panel B. Each dot represents one tumour (mean±SD). (M) IHC of tumours from panel B. Arrows depict individual T cells. The scale bar is 100 µm. (N–S) Flow-cytometric quantification of T cell subtypes in WT/KO5.3 tumours. Each dot represents one tumour (mean±SD). DYRK1B, dual specificity and tyrosine phosphorylation-regulated kinase 1B; IHC, immunohistochemistry; KO, knockout; MHCII, major histocompatibility complex class II; PDAC, pancreatic ductal adenocarcinoma; TAM, tumour-associated macrophages; WT, wild-type.

Following up on the observation of a potentially altered TME in the KO setting, we stained WT and KO tumour tissue with various antibodies detecting proliferation (Ki67), myofibroblastic cancer-associated fibroblasts (myCAFs; α-Smooth muscle actin (α-SMA)), inflammatory CAFs (iCAFs; Platelet-derived growth factor receptor β (PDGFRβ)), vasculature (CD31) or tumour-associated macrophages (TAMs; F4/80) (figure 3D). Intriguingly, tumour cell proliferation as assessed by KI67 staining was unaffected in vivo, arguing that the microenvironment (and not tumour cell-intrinsic proliferative alterations) was responsible for the observed growth retardation (figure 3E). Blood vessel density (CD31 staining) was moderate (figure 3F), while iCAF (PDGFRβ staining) abundance was profoundly reduced in KO tumours without significant changes in myCAFs (α-SMA) (figure 3G,H). One of the most striking differences between WT and KO was however observed with F4/80, revealing a significantly increased abundance of macrophages in KO tumours (figure 3D,I), a finding also corroborated with the second KO clone (online supplemental figure S3C). Interestingly, TANs appeared not to be recruited (online supplemental figure S3D). Dual immunofluorescence, as well as flow cytometry studies, revealed that the KO-associated TAMs were TNF-positive and MHCII-positive, suggesting a polarisation towards an M1-like state (figure 3J,K and online supplemental figure S3E). In contrast, the M2-marker CD206 was unaffected by the Dyrk1b genotype (figure 3L and online supplemental figure S3F,G). The increased influx of macrophages (Mphs) into KO tumour tissue was also reflected in the corresponding RNAseq data set, where pan-Mph genes (Cd68 and Msr1), as well as M1-polarised Mph genes (Cd80 and Csf2) presented with significantly higher levels in KO tumours (online supplemental figure S3H,I). In contrast, M2-polarised Mph genes (Cd163 and Serpinb2) trended to an inverse correlation (online supplemental figure S3J). Interestingly, and in contrast to the cancer cells, the majority of the KO-associated TAMs were non-proliferative (online supplemental figure S3K–M), strongly suggesting that they were recruited to the tumour site and not generated by proliferative expansion of a resident pool.

As TAMs can impact T-cell differentiation and activation, we next investigated WT/KO tissue for the presence of CD4+ and CD8+ T cells and found the former to be more frequent in KO versus WT tumours (figure 3M–O). Importantly, these CD4+ cells were highly enriched for the Th1 markers interferon-gamma (IFNγ), TNF and T-bet (figure 3P,Q and online supplemental figure S3N,O). Considering that Th1 cells can boost M1-TAM activation, these results point towards a reciprocal crosstalk between Th1 and M1-TAMs in the Dyrk1b-deficient TME. In contrast, the frequency of neutrophil-attracting RORγt+ Th17 cells was significantly reduced (figure 3R). We also observed an increased presence of IFNγ+/TNF+ CD8+ T-cells, indicating enhanced differentiation of cytotoxic lymphocytes, despite comparable absolute frequencies in WT versus KO tumours (online supplemental figure S3S).

In conclusion, we find the loss of Dyrk1b to exert a strong growth inhibition in an in vivo setting, despite fostering cell proliferation in vitro. This tumour-suppressive effect was associ