Intravesical Ty21a Vaccine Promotes Dendritic Cells and T Cell–Mediated Tumor Regression in the MB49 Bladder Cancer Model

Bladder cancer is the fourth and eighth most common malignancy among men and women, respectively (1, 2). Approximately 75% of bladder cancers are diagnosed as non–muscle-invasive (3), and according to specific tumor stage and grade characteristics, intravesical immunotherapy with Bacillus Calmette–Guérin (BCG) is used to prevent recurrence and/or progression (4). However, BCG immunotherapy is associated with significant adverse events, mainly bladder irritation, general malaise, and fever, but also severe though rare complications (5). In addition, treatment failure may occur in 30% to 40% of cases (6), hence the necessity for alternatives. Along this line, we reported that the attenuated Salmonella enterica serovar Typhi Ty21a live vaccine-strain against typhoid fever was effective at inducing regression of established bladder tumors using the immunocompetent orthotopic MB49 bladder cancer model (7), which closely mimic non–muscle invasive bladder cancer (NMIBC) in mice (8). Besides the excellent safety profile of the oral Ty21a vaccine Vivotif, as confirmed worldwide in more than 200 million vaccinees over the last 30 years (9), the absence of bacterial survival in the bladder (7) points to it as a viable candidate to test in NMIBC patients. Toward this goal, it is crucial to examine whether the local adverse events associated with BCG may also be induced by intravesical Ty21a. Here we characterized the inflammatory effects of Ty21a on the bladder, but also investigated the innate and/or adaptive immune mechanisms underlying tumor regression in the MB49 bladder cancer model. Using morphology, IHC, and flow cytometry, we comparatively characterized inflammation and immune cell infiltration in the bladder upon the different intravesical treatments and functionally determined key effector cells. Altogether, we highlighted mechanisms of intravesical Ty21a therapy that may have the potential for safe and efficient implementation in NMIBC patients.

The MB49 cell line, derived from a carcinogen-induced urothelial carcinoma in male C57Bl/6 mice in 1979 (8), was kindly provided in 2009 by Professor A. Loskog (Uppsala University, Sweden), amplified for 1 week (2 passages), and aliquots were frozen. Luciferase-expressing (MB49-luc) and green fluorescent protein (GFP)-expressing (MB49-GFP) cells were generated by transfection with lentiviral vectors encoding for firefly luciferase and GFP, respectively (kindly provided by Prof. D. Trono, EPFL, Lausanne, Switzerland) of one MB49 aliquot after 1 passage. MB49-luc and MB49gfp were amplified during 1 week (2 passages), tested as free of mycoplasma and aliquots frozen. Further amplifications (2 passages) from the initial stock were performed in 2012 and 2018 to generate new aliquots. In all experiments, a new aliquot of cells is used within 10 days after thawing.

Seven- to 10-week-old female C57Bl/6 wild-type mice (Charles River) were used, and all experiments were performed in accordance with Swiss law and with approval of the Cantonal Veterinary Office of Canton de Vaud, Switzerland. Bladder tumors were established in deeply anesthetized mice that were urethrally catheterized using Introcan 24Gx3/4 catheters (Braun). A 15-minute pretreatment with 100 μL 22% ethanol was performed before instillation of 500,000 MB49-luc (or MB49-GFP) cells in a volume of 50 μL. MB49-luc tumor growth was monitored by bioluminescence 15 minutes after intraperitoneal (i.p.) injection of D-luciferin (Promega, L8220, 150 μg/g of body weight) in the Xenogen imaging system (Xenogen/IVIS Caliper Life Science, kindly provided by cellular imaging facility, CIF/UNIL, Lausanne, Switzerland). One hundred percent of the mice will develop bladder tumors. Bioluminescence monitoring of MB49-luc tumors is very efficient for assessing tumor establishment and growth during the first 3 weeks; however, uncontrolled loss of luminescence of the growing tumors can then often appear (10), requiring additional monitoring by palpation, hematuria, and overall health status of the mice, which were euthanized if they reached > 15% weight loss.

Ty21a bacteria were prepared by resuspension of the lyophilized content of a Vivotif capsule (PaxVax) into 750 μL of PBS, resulting in ca. 3 × 10⁹ CFU/mL. BCG bacteria were prepared by resuspension of one vial of oncoTICE (Essex Chemie SA) in 1 mL of PBS, resulting in ca. 3 × 10⁸ CFU/mL. Further dilutions were made in PBS as required to achieve the indicated bacteria numbers used in the experiments.

Bacterial suspensions (50 μL) were instilled by urethral catheterization, as described above. The retention time in the bladder was ca. 1 hour, until the mice awoke from the anesthesia and would spontaneously urinate. Classic treatment (Table 1) starts 1 day after intravesical tumor-cell instillation and is administered 4 times at weekly intervals (days 2, 9, 16, and 23). In a more stringent setting, a single intravesical treatment instillation was administered at day 2 (24 hours after intravesical tumor-cell instillation). Tumors can be detected by bioluminescence at day 5 and grow until days 9 to 12 at a similar rate, irrespective of instillation with PBS versus BCG or Ty21a.

Three-micrometer paraffin-embedded hematoxylin–eosin-stained sections were used for morphologic evaluation of bladder specimens, including the urothelial mucosa, muscular, subserosa, and serosa layers. The inflammation was evaluated using an inflammatory scoring system as previously reported (11). Score 0: absence or minimal inflammation or epithelial changes. Score 1: mild inflammation within the lamina propria accompanied by mild chronic edema, hemorrhage, or urothelial changes, zonal fibrosis in lamina propria. Score 2: moderate inflammatory infiltrate in the lamina propria and focal extension of the inflammation into the muscularis propria, accompanied by moderate edema, hemorrhage, urothelial changes, and diffuse fibrosis in the lamina propria. Score 3: severe inflammation in the lamina propria and muscularis propria in association with other significant findings including urothelial ulceration, severe chronic edema, hemorrhage, and diffuse fibrosis. For IHC evaluation, paraffin sections of bladders were immunostained using the following primary antibodies: CD11b (Abcam ab133357), F4/80 (Caltag MF48000), and CD3 (Abcam ab5690), an avidin-biotinylated horseradish peroxidase complex (Vectastain Elite ABC Kit), and counterstained with hemalun.

Mice were sacrificed by CO₂ inhalation to collect the bladders. Single-cell suspensions were obtained by mincing in DL-dithiothreitol (Sigma, D9779) and digesting stepwise with 0.5 mg/mL thermolysin (Sigma, T7902) and 1 mg/mL collagenase/dispase (Roche, 11 097 113 01; ref. 12) or by single digestion step with 1 mg/mL collagenase/dispase and 0.1 mg/mL DNAse I (Sigma-Aldrich, D2552) with 20% fetal calf serum (Gibco, 10270). Whole blood was collected in tubes containing heparin-Na 25000 I.E. (Braun, 1718711) from tail vein. Red blood cells were lysed using ammonium–chloride–potassium.

The recovered cells were stained and analyzed by flow cytometry. The following monoclonal antibodies (mAb) to mouse proteins were used: CD3-PE (17A2, 100206), CD3-PerCP/Cy5.5 (17A2, 100218), Ly6G-PE/Cy7 (1A8, 127618), CD11b-APC (M1/70, 101212), Ly6C-AF700 (HK1.4, 128024), Ly6C-APC/Cy7 (HK1.4, 128026), NK1.1-AF700 (PK136, 108730), CD4-AF700 (GK1.5, 100430), CD8-APC/Cy7 (53-6.7, 100714), F4/80-APC-Cy7 (BM8, 123118), CD103-Pcblue (2E7, 121418; BioLegend); CD4-eF450 (GK1.5, 48-0041-82), CD11c-PE (N418, 48-0041), CD11c-PE-eF610 (N418, 61-0014, CD45-PerCP/Cy5.5 (30-F11, 45-0451; eBioscience); CD8-PETXRD (53-6.7, 1550-10; Southern Biotech). Isotype controls were used for CD11c-PE-eF610 [Harmenia Hamster IgG-PE-ef610 (eBio 229 Arm, 61-4888-80) from eBioscience] and for F4/80-APC-Cy7 and CD8-APC/Cy7 [Rat IgG2a kappa-APC/Cy7 (RTK 2758, 400524) from BioLegend]. FMO was used for CD103-Pcblue determination. Dead cells were excluded by a live/dead fixable aqua dead cell stain kit (L34957, Invitrogen, Thermo Fisher Scientific). Cell acquisition and analysis were performed using Gallios Flow Cytometer (Beckman Coulter) and FlowJo software (Tree Star), respectively.

Peripheral blood from healthy volunteers was obtained through the local Swiss blood bank. Fresh peripheral blood mononuclear cells (PBMCs) were purified by density gradient centrifugation and cryopreserved. PBMCs were infected with Ty21a or BCG at the indicated multiplicity of infection (MOI) for 1.5 hours at 37°C. Cells were washed and medium containing 50 μg/mL gentamycin was added. Following 24 hours of incubation at 37°C, cells were harvested and stained for DC identification by flow cytometry. The following mAbs to human proteins were used: CD11c-BV421 (Bu15, 337225), HLA-DR-PE/Cy7 (L243, 307616), and CD14-FITC (HCD14, 325604) from BioLegend; CD3-PE/AF610 (7D6, MHCD0322), CD19-PE/AF610 (SJ25C1, MHCD1922), and CD56-PE/TexasRed (MEM-188, MHCD5617) from Invitrogen. Cells were stained for 20 minutes at 4°C, and an amine reactive dye (aqua live/dead stain kit, from Life Technologies, L34957) was used for dead cell exclusion according to the manufacturer's instructions. Fc-receptor blocking reagent (Miltenyi Biotec, 130-059-901) was used to increase staining specificity. Sample acquisition was performed as described above.

The following mAbs from Bio X Cell were used for cell depletion: CD8 (2.43, BE 0061), CD4 (GK1.5, BE 0003), NK1.1 (PK136, BE 0036), Ly6G (1A8, BE 0075), Ly6C (Monts1, BE 0203), or rat IgG2a (2A3, Roche), or IgG2b (LTF2, BE 0090) as isotype controls. A first i.p. injection of 100 μg (anti-CD8), 200 μg (anti-CD4 or anti-Ly6G), 250 μg (anti-NK1.1), or 400 μg (anti-Ly6C), or isotype controls was given 2 days before tumor implantation. This was followed by i.p. injections every 3 to 5 days, for a period of ca. 40 days, using half of the first dose (except for anti-CD4 and anti-Ly6C that were injected at full dose).

Statistical analyses were performed using Prism 7.00 for Windows (GraphPad software). Single comparisons were performed using Student t test. Multiple comparisons were performed using one-way ANOVA and Dunnet post-test or adjusted log-rank test as indicated in the figure legends.

Using the mouse MB49 orthotopic bladder cancer model (13–15), we determined the dosage range of efficacy of the intravesical Ty21a and BCG treatments (Table 1). After four consecutive instillations (1 week apart, starting 1 day after tumor implantation; ref. 16), Ty21a induced a significantly higher mice survival (80%–90%), as compared with control PBS treatment using doses ranging from 3 × 10⁶ to 3 × 10⁸ CFU/dose (the maximal dose tested, corresponding to 1/10 of the content of the Vivotif capsule). Similar survival in mice was also obtained with BCG at 3 × 10⁷ CFU/dose (the maximal dose tested, corresponding to 1/10 of the Oncotice vial), whereas treatment with the lower BCG dose (3 × 10⁶ CFU) no longer conferred significantly improved survival, compared with controls (Table 1). This shows that multiple intravesical Ty21a instillations can control tumor growth and increase survival in mice with doses as low as 3 × 10⁶ CFU.

Intravesical bladder immunotherapy with BCG or Ty21a is associated with the induction of inflammatory cytokines (7, 17). Here, we show a morphologic analysis of treated bladders to characterize inflammatory features, including the presence of edema, fibrosis, and immune cell infiltration, as summarized by an inflammatory score (ref. 11; Fig. 1A). Single intravesical instillation with either BCG or Ty21a (3 × 10⁷ CFU) induced significant inflammation in the bladder 24 hours later, which was considerably decreased after 7 days (Fig. 1B). Similar inflammation was induced 24 hours after the fourth consecutive instillation with either BCG or Ty21a; however, inflammation persisted for at least 3 weeks (Fig. 1C). When the lower dose (3 × 10⁶ CFU) was used, inflammation was still present upon BCG administration, but not upon Ty21a treatment. These data show that bladder tissue inflammation can be reduced by using low, but still effective (Table 1), doses of Ty21a.

Infiltration of myeloid cells and T cells in the bladder upon Ty21a or BCG instillations was evaluated by IHC (Fig. 2A; Table 2). Myeloid cells (CD11b⁺), mainly of the granulocytic phenotype, were slightly increased within 24 hours following instillation of PBS alone and to a greater extent by BCG or Ty21a; however, within 7 days, myeloid/granulocytes represented again less than 10% of the bladder area similarly to the PBS-treated bladder. In contrast, following multiple BCG instillations (either 3 × 10⁶ or 3 × 10⁷ CFU/dose), a strong (10%–50% of the bladder area; Table 2) myeloid/granulocyte infiltration was maintained for at least 3 weeks. Macrophages (F4/80⁺) were only consistently observed after multiple doses of BCG, whereas T cells (CD3⁺) were increased by Ty21a at early time points (24 hours) and by BCG at later time points (3 weeks). Immune cell infiltration was further characterized and quantified by flow-cytometry (Fig. 2B and C; Supplementary Fig. S1A). The data show a significant infiltration in the bladder of T cells and natural killer (NK) cells 24 hours after a single Ty21a instillation, whereas myeloid cells were significantly induced by either Ty21a or BCG, as compared with PBS-treated and untreated mice (Fig. 2B). A significantly greater number of T cells was maintained for 1 week after Ty21a instillation (Fig. 2B, left), whereas NK cells and myeloid cells returned to the numbers measured in PBS-instilled or untreated mice (Fig. 2B, middle and right). A 10-fold higher dose of Ty21a (3 × 10⁸ CFU) did not induce greater immune cell infiltration (Fig. 2B, dark green bars). After multiple instillations (Fig. 2C), there was a significant increase in T cell, NK, and myeloid cell numbers upon successive BCG instillations as compared with the first dose. In contrast, successive Ty21a doses did not further increase the number of myeloid cells and only to a lesser extent the number of T cells and NK cells (only significant after the third and fourth doses), as compared with the first dose (Fig. 2C). The myeloid cells induced after instillation were further phenotyped by flow cytometry (Supplementary Fig. S1B). They consisted predominantly of Ly6G⁺ neutrophils (40%–60%), followed by Ly6C⁺ monocytes (15%–20%), dendritic cells (DC; 15%–30%), and fewer F4/80⁺ macrophages (3%–7%; Table 3), in agreement with the IHC data (Table 2). Although the proportion of neutrophils was significantly higher upon BCG administration than upon Ty21a treatment, the proportion of DCs was higher upon treatment with Ty21a than with BCG. Further analysis of DCs infiltrating the bladder upon instillation of Ty21a or BCG (Fig. 2D and E) showed that they were mainly Ly6C⁺CD103⁺CD11c⁺ DCs (Fig. 2D) in both type of intravesical instillations, but in significantly higher numbers upon treatment with Ty21a than BCG (Fig. 2E). These DCs thus shared markers of both inflammatory (Ly6C⁺; ref. 18) and cross-presenting (CD103⁺; ref. 19) DCs. Their presence in the bladder, together with the higher number of T cells (Fig. 2E), highlights the potential of Ty21a to promote antitumor T-cell responses locally.

As a surrogate for comparing the effects of BCG and Ty21a instillations on human bladder DC, we examined DC frequency upon ex vivo infection of human PBMCs. Data show that 24 hours after infection, the frequency of DCs (Lineage^negCD11c⁺HLA-DR^high; Fig. 3A) was significantly increased by Ty21a, but not by BCG (both at an MOI = 0.5), as compared with medium alone (Fig. 3B). A significantly higher frequency of DCs was also obtained with Ty21a when high MOI (10) was used. In contrast, infection with BCG at MOI = 10 resulted in a significantly decreased DC frequency as compared with medium (Fig. 3B), suggesting a higher toxicity of BCG toward myeloid cells (see total CD11c⁺ cell frequency in Supplementary Fig. S2), possibly attributed to its greater persistence in PBMC as compared with Ty21a (7). Altogether, these data suggest that intravesical Ty21a may also promote DC infiltration in the human bladder, in contrast to BCG.

We examined the presence of DCs in MB49 bladder tumors following intravesical treatments performed with BCG or Ty21a 1 day or 5 days after tumor instillation. At both time points, significantly higher numbers of CD103⁺Ly6C⁺CD11c⁺ DCs/mg of bladder tumor were obtained 1 day after intravesical Ty21a as compared with BCG (Fig. 4A), similar to the data obtained in the absence of tumor (Fig. 2E). To highlight the potential of CD103⁺Ly6C⁺CD11c⁺ DCs to present tumor antigen, we used MB49 cells expressing GFP and examined whether these DCs had engulfed GFP from the tumors. Compared with MB49-luc tumors, ca. 5% of CD103⁺Ly6C⁺CD11c⁺ DCs were GFP⁺ at day 6 in MB49-GFP tumors (Fig. 4B). Comparison between intravesical BCG and Ty21a treatments in this setting further confirmed that a significantly higher number of CD103⁺Ly6C⁺CD11c⁺ GFP⁺ DCs/mg of tumor was obtained after Ty21a than after BCG (Fig. 4C). This suggests that CD103⁺Ly6C⁺CD11c⁺ DCs may be key for the Ty21a treatment.

To functionally investigate the immune cells involved in the Ty21a-mediated tumor regression, we performed a single Ty21a intravesical instillation (3 × 10⁷ CFU) 1 day after MB49 bladder tumor implantation, a treatment protocol that we previously showed to significantly increase mice survival (7). This allowed antibody-mediated immune cell depletion to be performed without interfering with successive Ty21a treatments. Depletion of CD4⁺ T cells, CD8⁺ T cells, NK cells, and Ly6G⁺ neutrophils was highly effective and maintained for more than 1 month, as detected in the blood (Supplementary Fig. S3A). In contrast, our attempt to deplete CD103⁺Ly6C⁺CD11c⁺ DC with an Ly6C-depleting antibody only resulted in partial depletion (ca. 50%) of the Ly6C⁺ myeloid cells in blood (Supplementary Fig. S3B) and of the Ly6C⁺CD11c⁺ DCs in the bladder (Supplementary Fig. S3C). Tumor luminescence, as a surrogate for tumor volume, was not significantly different 10 days after tumor implantation in the different treatment groups, whereas 1 week later, significantly larger tumors were found in CD4-, CD8-, and Ly6C-depleted Ty21a-treated mice compared with mice receiving intravesical Ty21a with isotype control antibodies (Fig. 4D). In contrast, neutrophil- or NK-depleted Ty21a-treated tumors were similar to those in control Ty21a-treated mice, suggesting that neutrophils and NK cells are not necessary for an effective Ty21a treatment. Indeed, mouse survival was not affected by the long-term depletion of neutrophils and NK cells (35–40 days) after intravesical Ty21a treatment (80%–90% survival; Fig. 4E). In contrast, CD4⁺ or CD8⁺ T-cell depletion rapidly decreased mouse survival, and all mice were euthanized before day 40 (Fig. 4E), demonstrating that these T cells were essential immune cells for Ty21a treatment efficacy. The survival of the Ly6C⁺-depleted group was not significantly affected, with only 2 of 10 mice dying at earlier time points than the control group (Fig. 4E). This suggests that the partial depletion of CD103⁺Ly6C⁺CD11c⁺ DCs was not sufficient for long-term interference with the Ty21a treatment. Altogether, these data demonstrate the requirement for T cells (CD4⁺ and CD8⁺) and suggest the importance of CD103⁺Ly6C⁺CD11c⁺ DCs for the intravesical Ty21a treatment of bladder tumors, whereas, in contrast to BCG (17), neutrophils and NK cells are not necessary.

Our data show that significantly improved survival of MB49 bladder tumor–bearing mice could be achieved by intravesical instillations of Ty21a at doses that induce only low inflammation in the bladder. With a single instillation, Ty21a induced significant infiltration of T cells and DCs in the bladder at numbers that required multiple instillations with BCG. In contrast to BCG, for which mechanisms of action involve induction of both innate (mainly neutrophils and NK cells) and adaptive immune responses (17), the therapeutic effect of Ty21a relied on the presence of T cells and DC, but not of neutrophils and NK cells.

Our results also indicate that a particular type of DC expressing Ly6C, CD11c, and CD103 infiltrated the bladder and the MB49 bladder tumor upon Ty21a or BCG instillation. Ty21a instillation induced ca. 4-fold more of these DCs in the bladder than BCG. Similar Ly6C⁺ DCs were reported to differentiate from monocytic precursors upon inflammatory signals activating p53 within tumors undergoing immunogenic chemotherapies (20). Whether Ty21a may use a similar process to promote such an immunogenic microenvironment within the treated tumor deserves further investigation. Experiments with human PBMC further confirmed an increased DC frequency upon Ty21a infection, in contrast to BCG. In the context of antituberculosis vaccines, the relatively poor immunogenicity of BCG was previously associated with DC impairments, such as a poor maturation capacity (21, 22). The presence of DCs able to efficiently cross-present tumor antigen for activating antitumor T-cell immunity is crucial for an effective immunotherapy (23). In the absence of a specific tumor antigen, we could not assess the cross-presenting potential of the Ly6C⁺CD11c⁺CD103⁺ DCs but demonstrated their ability to engulf tumor antigen (GFP). It was previously reported that Salmonella enterica serovar Typhi can induce maturation of human DC capable of cross-presenting bacterial antigen and priming Typhi-specific CD8⁺ T cells (24). Given that Ty21a can rapidly induce urothelial tumor-cell death (7), this in combination with Ty21a-induced DCs may enable efficient cross-presentation of the released tumor antigens and activation of antitumor T cells either locally and/or in the draining lymph nodes. This process may explain why a single intravesical instillation of Ty21a is more efficient than BCG to induce regression of established bladder tumors and improve survival of treated mice (7).

Here, we present a thorough analysis of the inflammatory effects of single and successive intravesical instillations of BCG on morphology and immune infiltration in the bladder of mice. Our analysis is in line with previously reported data from NMIBC patients undergoing intravesical BCG immunotherapy, which resulted in augmented immune responses, including increased cytokine concentrations and immune cells in the urine upon successive BCG instillations (25–27). In contrast, a single Ty21a instillation was sufficient to induce DCs and T cells capable of controlling tumor growth. Successive instillation of Ty21a, compared with BCG, did not result in a similar profile of enhanced innate immune cell infiltration and inflammation of the mouse bladder, particularly when the lower doses were used. Repeated exposure to lipopolysaccharide of gram-negative bacteria was shown to transiently silence inflammatory cytokines to prevent excessive inflammation (28, 29), which may partly explain our finding.

Immunotherapy of NMIBC with BCG is multifactorial, relies on an intact immune system, as the inflammatory response is crucial, with neutrophils playing a primary role (14). This overt inflammation, however, often leads to adverse events that can greatly reduce the patient's compliance. Although attempts to reduce the dose of BCG to one third (30) or one half (31) resulted in decreased toxicity, full efficacy could not be guaranteed, and the standard BCG regimen is still recommended (32).

The finding that Ty21a-mediated immunotherapy can operate in the absence of NK cells, neutrophils, and overt inflammation therefore holds promise for the implementation of this type of intravesical treatment in NMBIC patients, and a phase I trial is currently recruiting patients in our hospital (NCT 03421236: IVES Ty21a).

No potential conflicts of interest were disclosed.

Conception and design: S. Domingos-Pereira, J.A. Haefliger, D. Nardelli-Haefliger

Development of methodology: S. Domingos-Pereira, K. Sathiyanadan, P. Martel, L. Derré, J.A. Haefliger

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Domingos-Pereira, K. Sathiyanadan, L. Polák, M.F. Chevalier, P. Martel, R. Hojeij, D. Nardelli-Haefliger

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Domingos-Pereira, K. Sathiyanadan, S. La Rosa, L. Polák, M.F. Chevalier, P. Martel, L. Derré, D. Nardelli-Haefliger

Writing, review, and/or revision of the manuscript: S. Domingos-Pereira, K. Sathiyanadan, S. La Rosa, L. Polák, M.F. Chevalier, L. Derré, J.A. Haefliger, D. Nardelli-Haefliger

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Domingos-Pereira, S. La Rosa, J.A. Haefliger

Study supervision: J.A. Haefliger, P. Jichlinski, D. Nardelli-Haefliger

We thank the mouse pathology facility of the UNIL/CHUV for paraffin section and IHC staining. The study was funded by Swiss National Funds (#32153201 and #CRII3 160742 to D. Nardelli-Haefliger); Swiss Cancer League #2808-08-2011 and KFS 4103022017 to D. Nardelli-Haefliger and KFS 3796022016 to J.A. Haefliger; and InnoSTEP grant to D. Nardelli-Haefliger.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.