Loss of Cxcr2 in Myeloid Cells Promotes Tumour Progression and T Cell Infiltration in Invasive Bladder Cancer

Mice and carcinogen treatment

Animal experiments were carried out at the CRUK Beatson Institute, Glasgow, in accordance with the Home Office Animal (Scientific Procedures) Act 1986 in the UK (Project Licence 70/9028), adhered to ARRIVE guidelines, and approved by the University of Glasgow Animal Welfare and Ethical Review Board. Genotyping was performed by Transnetyx (Cordoba, TN, USA). The number of mice used, including gender, is presented (Table 1) and specified in each figure. LysMCre Cxcr2^flox/flox (Cxcr2 flox) mice were generated by crossing LysMCre and Cxcr2^flox/flox mice [28] in C57Bl/6J background. The C57Bl/6J mice (wildtype) purchased from Charles River (Tranent, UK) were used as controls. Mice aged 8-10 weeks were administered with 0.05% v/v N-butyl-N-(4-hydroxybutyl) nitrosamine (OH-BBN, TCI, Oxford, UK) in drinking water for 10 weeks (Fig. 1A). The bladders of wildtype mice aged beyond 20 weeks were monitored using the Vevo3100 ultrasound system with 25–55 MHz transducer (Fujifilm VisualSonics, Toronto, Canada) and culled when they showed clinical signs of bladder tumours. Blood was sampled in EDTA-containing tubes and white blood cell populations were analysed using ProCyte Dx Hematology Analyzer (IDEXX, Westbrook, Maine, USA).

Fig. 1

Bladder tumour phenotype of LysMCre Cxcr2^flox/flox (Cxcr2 flox) mice. (A) Mice were given tobacco carcinogen 0.05% (v/v) OH-BBN in drinking water for 10 weeks, and bladder tissues were examined at 2, 16 and 20 weeks from the start of the carcinogen treatment. (B) An example of a bladder tumour developed at 20 weeks from the start of the carcinogen treatment in a Cxcr2 flox animal, with an invasion to the muscle (arrow) and keratinization (Kr). Representative images of wildtype (C, E) and Cxcr2 flox (D, F) bladders at 16 weeks (C, D), and 2 weeks (E, F). Enhanced tumour pathogenesis of the urothelium, such as carcinoma in situ-like lesions (arrowheads) was observed in Cxcr2 flox bladders at 16 weeks (D). At 2 weeks, the differences between wildtype and Cxcr2 flox bladders were not evident. IHC staining showed infiltrating Cxcr2⁺ myeloid cells in tumours at 20 weeks (G), the urothelium (J) and around the blood vessels in the bladder stroma (L) at 2 weeks in wildtype mice (arrows in G, J, L). In contrast, the staining was absent in Cxcr2 flox tumours (H, I) and in the bladder at 2 weeks’ time point (K, M). IHC staining was performed in at least n = 3 samples in each genotype. Annotations in B-M are; CIS, carcinoma in situ; Kr, keratinization; M, muscle, St, stroma; T, tumour; U, urothelium; V, blood vessel. The scale bar represents 500μm in B, 250μm in C-F, 30μm in G-M. For evaluation of the bladder and tumour phenotype, the scoring criteria from our previous study [35] were refined to optimally capture the phenotype of OH-BBN treated mice under our protocol. The “bladder phenotype” defines the overall state of the urothelial tumorigenesis and progression. The “invasiveness” describes the state of the basement membrane and invasion of tumour cells, while the “squamous transformation” describes the squamous appearance of cells and keratinization in the urothelium or in the tumour. The criterium “minimal changes” of the bladder phenotype was established to indicate the baseline appearance of the urothelium treated with OH-BBN observed [35]. One representative H&E section was scored per animal. The bladders usually showed a mixed phenotype of all criteria. The worst score was used. Bladder phenotype (N, R), invasiveness (O, S) and squamous transformation (P) are expressed as a frequency of observations (%) comparing wildtype (wt, n = 45) and Cxcr2 flox (flox, n = 27) male mice at 20 weeks (N-P) and at 16 weeks (n = 5 per cohort) (R, S). Incidents of tumour in N were presented by numbers in Table 1. The tumour area was measured by QuPath in a representative section per mouse at 20 weeks (n = 15 per cohort) (Q) and carcinoma in situ (CIS) and tumour areas at 16 weeks (n = 5 per cohort) (T). The means with standard deviation are shown in bars (Q, T). Observations in male mice are presented (N-T). The p-values (Mann-Whitney test) are indicated where significant (p < 0.05).

Histology, scanning and histopathological scoring

The bladders were dissected and gently emptied of urine by pressing the tissue with forceps. The bladders were fixed in 10% neutral-buffered formalin overnight and embedded in paraffin. Four-μm sections were used for haematoxylin and eosin (H&E) staining and immunohistochemistry (IHC). Stained slides were scanned by NanoZoomer S60 and analysed by NDP View 2 (Hamamatsu, Shizuoka, Japan) or QuPath version 0.2.3 [34]. H&E images were used for histopathological analysis of bladder and tumour phenotype. The scoring criteria are described in Fig. 1 legend and Table S1. Scoring was performed by multiple observers (NFBI, NC, MF, TI) in discussion with the pathologists (JMS, JDGL).

2.3Immunohistochemistry

Antigen retrieval (AR) was performed in heat-induced epitope retrieval (HIER) buffer (ThermoFisher Scientific, Waltham, Massachusetts, USA) by microwave to boiling for 1 min at full power, followed by 10 min at 20% power. Sections were incubated with 3% H₂O₂ for 15 minutes and blocked with 2.5% normal goat or horse serum (Vector Laboratories, Peterborough, UK) for 30 minutes. Primary antibody was applied for 1 hour at room temperature. After washing with TBST (VWR, Lutterworth, UK), secondary antibodies (ImmPRESS HRP Polymer Detection Kit, Vector Laboratories) were applied for 30 minutes. Immunoreactivity was detected by 3,3^′-diaminobenzidine (DAB) substrate (Vector Laboratories) and slides were counterstained with haematoxylin. Details of the antibodies used are provided in Table S2.

Image analysis

Manual annotation in QuPath [34] was used to measure the area of the carcinoma in situ (CIS) and tumour. The immune cells stained positively by IHC was quantified in 10 fields of 100×100-μm, randomly placed in the urothelium, stroma, muscle layer, and tumour. The mean value was used to represent each sample. QuPath was also used for quantification of Ki67⁺ and Caspase3⁺ cells. The stain vector was first adjusted by selecting a representative area indicating haematoxylin and DAB. Region of interest (ROI) was drawn manually for area representing tumour. The mean size of the ROIs was 1.42 mm², ranging from 0.18–10.00 mm². Caspase 3⁺ cells were quantified manually using QuPath. For Ki67, positive cell detection command was used. The threshold 1+ on the score compartment “Nucleus: DAB OD mean” was used to categorize positive or negative cells by adjusting the value in each image. For quality control, all images were also manually scored in the test ROIs. Scoring was performed by at least two independent observers (NFBI, AREY, NXC and TI).

Total RNA extraction and cDNA synthesis

At dissection, the bladder tissue was halved at the midline. One half was processed in paraffin for histology, the other half was collected in RNAlater (Qiagen, Manchester, UK) for RNA extraction and stored at –20°C. The total RNA was extracted using the RNeasy Mini Kit (Qiagen) and was quantified using NanoDrop Microvolume Spectrophotometer (ThermoFisher Scientific). Samples with 260/280 absorbance ratio between 1.8 and 2.2 were used. The integrity of the RNA was also evaluated by agarose gel electrophoresis. RNA samples with two sharp bands, corresponding to 28S and 18S ribosomal RNA, and with the intensity ratio of the 28S:18S = 2:1, were considered intact and used for the further analysis. The RNA samples were stored at -20°C or –80°C. For the cDNA synthesis, 0.5μg of total RNA was subjected to the removal of genomic DNA elimination and reverse transcription using the RT² First Strand Kit (Qiagen). Genomic DNA was removed by incubating 0.5μg of the total RNA in Buffer GE with a final volume of 10μl at 42°C for 5 minutes. The reverse-transcription was performed with Buffer BC3, Control P2, RE3 Reverse Transcriptase Mix and RNAse free water added to a final volume of 20μl and at 42°C for 15 minutes. The reaction was terminated by incubation at 95°C for 5 minutes and added with 91μl of RNAse-free water, to achieve the final volume of 111μl. The resultant cDNA was stored at –20°C or used immediately for RNA array.

RNA array

RNA array experiment was performed using RT² Profiler Mouse Cancer Inflammation & Immunity Crosstalk PCR Array (Qiagen). The PCR component mix was prepared with 1350μl of 2X RT² SYBR Green Mastermix, 102μl of cDNA synthesis reaction and RNase-free water added to the final volume of 2700μl. For each well of the RT² Profiler PCR array, 25μl of the PCR components mix was added. PCR reaction was performed using the StepOnePlus Real-Time PCR System (ThermoFisher Scientific). Thermal cycle conditions were 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. Fold changes between samples were determined using the 2^–ΔΔCT method. Total RNA from wildtype (n = 3) and Cxcr2 flox (n = 2) bladders at 20 weeks were used. The ΔCT values of house-keeping gene was subtracted from those of each gene, then the mean value for wildtype and Cxcr2 flox was determined. The ΔΔCT value was determined by ΔCT (Cxcr2 flox mean) - ΔCT (wildtype mean), then converted to log2 fold by 2^–ΔΔCT.

TCGA data analysis

The clinical information, genomic alternations, mRNA expression, and MIBC molecular Subtype (mRNA Cluster) of human bladder cancer (TCGA, Cell 2017) (n = 412) were downloaded from cBioPortal (https://www.cbioportal.org/public-portal/index.do) and from the reference [32].

Statistics

GraphPad Prism 9 (GraphPad Software, San Diego, CA, USA) was used. Gaussian distribution of the data was first evaluated. The differences of the means between two data groups were evaluated using the Mann-Whitney test for non-parametric distributions. Kruskal-Wallis test was used to evaluate the differences of more than three data groups. Pearson (parametric) or Spearman (non-parametric) tests were used for evaluation of correlation. Overall survival was estimated using the Kaplan–Meier method and analysed using the log-rank test. A p value of < 0.05 was considered significant.

Cxcr2 deletion enhanced urothelial tumorigenesis

Using the previously established 10-weeks on, 10-weeks off protocol of OH-BBN induction of invasive urothelial carcinoma [35], we have compared the bladder pathogenesis in mice deleted with Cxcr2 in the myeloid cell lineage, LysMCre Cxcr2^flox/flox (Cxcr2 flox) and wildtype mice at the timepoints of 2, 16 and 20 weeks from the start of carcinogen treatment (Table 1, Fig. 1A). Human Protein Atlas data showed that CXCR2 expression can be detected in many immune cells and is enriched particularly in granulocyte lineage and neutrophils (Protein Atlas version 21.0, https://www.proteinatlas.org/ENSG00000180871-CXCR2) [36]. LysMCre mice were previously shown to drive Cre-lox recombination in the majority of neutrophils and macrophages, as well as dendritic cells to some extent [33]. The current model is therefore designed to allow evaluation of the effects of Cxcr2 loss in recruitment of myeloid cells to the bladder tissue, where inflammation was induced by OH-BBN.

Evaluating the tissue histopathology, Cxcr2 flox mice showed more frequent occurrence of bladder tumours (p = 0.005), higher invasiveness of tumour cells (p = 0.025) and squamous transformation (p = 0.001) compared to wildtype mice at 20 weeks from the start of OH-BBN treatment (Fig. 1B, N-P). The tumour area was similar (mean; wildtype 9.2 mm², Cxcr2 flox 10.4 mm², n = 15 per genotype) (Fig. 1Q). Cxcr2 flox bladders showed formation of carcinoma in situ (CIS) and increased invasiveness at 16 weeks, while wildtype bladders remained unchanged (Fig. 1C, D, R-T). At 2 weeks, the urothelium of Cxcr2 flox and wildtype was morphologically indistinguishable (Fig. 1E, F). Absence of Cxcr2⁺ myeloid cells were evident in tumours, the urothelium and around the blood vessels in Cxcr2 flox mice (Fig. 1G-M). Female Cxcr2 flox mice also showed more advanced bladder phenotype compared to wildtype females (Fig. S1). The phenotype was overall more advanced in males than female mice (Fig. 1, S1). Male mice were used in further investigations.

Altogether, Cxcr2 deletion in myeloid cells led to a significant increase in bladder tumorigenesis, with an earlier and more frequent occurrence of tumours with morphological appearances indicative of enhanced progression.

Effects of Cxcr2 deletion in tissue infiltration of immune cells

We next investigated how Cxcr2 deletion influenced the tissue infiltration of immune cells. At 2 weeks timepoint, acute inflammation is normally observed in response to OH-BBN treatment [35]. As expected, infiltration of neutrophils, assessed by the canonical IHC markers, including Ly6G, NIMP and S100 calcium-binding protein A9 (S100A9), was suppressed in the Cxcr2 flox bladder stroma compared to those in the wildtype (p = 0.004) (Fig. 2A, B). Similarly, infiltration of F4/80⁺ macrophages were also downregulated as expected (p = 0.004) (Fig. 2D). Furthermore, TILs, including CD3⁺, Granzyme B⁺ (GrB⁺), CD4⁺, and FoxP3⁺ cells, was significantly reduced in the Cxcr2 flox bladder stroma (Fig. 2F). Similar changes were observed throughout the bladder tissue, including the urothelium and the bladder muscle (Fig. S2).

At 16 weeks, reduced infiltration of neutrophils and T cells was observed (Fig. 2A, F, S2), likely to be in reflection of the cessation of carcinogen at 10-week time point. Interestingly, however, a significant increase in macrophages was observed in the Cxcr2 flox bladder stroma compared to wildtype (p = 0.029) (Fig. 2D, E). At 20 weeks, infiltration of Ly6G⁺ and S100A9⁺ neutrophils was continued to be suppressed in Cxcr2 flox tumours compared to wildtype (p = 0.039, 0.048, respectively, Fig. 2A, C), while levels of macrophages remained low (Fig. 2D). Unexpectedly, overall levels of TILs were higher in Cxcr2 flox mice compared to wildtype (CD3⁺, p = 0.049. Fig. 2F). TILs were mostly observed in the tumour stroma, as shown in the example of CD3⁺ cells (Fig. 2G).

We then investigated how genotype and tumour phenotype are correlated with immune cell levels. Cross-examining the relationship between genotype, phenotype and immune cell infiltrations observed in each sample (Fig. S3), genotype was associated with infiltrating immune cells at 2 weeks as expected. At 20 weeks, association between genotype remained with Ly6G and CD3 levels, with little association between tumour phenotype and TILs, indicating that Cxcr2 deletion is more likely to have influenced the changes in the level of CD3⁺ T cells at 20 weeks rather than the tumour phenotype.

Finally, to see how changes in the immune cell levels have influenced the tumour burden, we analysed the status of cell proliferation and apoptosis in tumour tissues at 20 weeks (Fig. 2H-K). Ki67⁺ proliferating cells were observed mostly in the basal edge of CIS and in tumours (Fig. 2H). Levels of Ki67⁺ cells were similar in wildtype and Cxcr2 flox tumours (Fig. 2I), indicating that the increased tumour burden in Cxcr2 flox may have originated at an earlier stage. Caspase 3⁺ cells were relatively infrequent in tumours of both genotypes (Fig. 2J). No significant difference was observed in the density of Caspase 3⁺ cells in wildtype and Cxcr2 flox tumours (Fig. 2K). No obvious similarity was found in the location of Caspase3⁺ cells with those of Ly6G⁺ neutrophils or TILs (Fig. 2C, G, J).

Taken together, following a transient increase of macrophages at the outset of tumour formation, an increase in T cell infiltration was observed in Cxcr2 flox tumours. However, the elevated level of TILs was not associated with changes in cell proliferation or cell death in the absence of Cxcr2. Therefore, the increased tumour burden upon Cxcr2 deletion may have resulted from a suppression of acute inflammation in the bladder at the time of tumour initiation.

Levels of blood immune cells did not always correlate with those in the bladder tissue

The effects of Cxcr2 deletion on blood leukocyte populations were compared to those in the tissues (Fig. S4). At 2 weeks from the start of OH-BBN treatment, the level of neutrophils in the blood appeared lower in Cxcr2 flox mice compared to wildtype, as expected, while basophils appeared to have increased (Fig. S4A, B, not statistically significant). At 16 weeks, the blood levels of monocytes were higher in Cxcr2 flox mice (p = 0.016) (Fig. S4B), similar to the observations in the tissue (Fig. 2B). In addition, the lower population of neutrophils within the whole blood (p = 0.024, Fig. S4A), and higher blood concentration of lymphocytes were observed (p = 0.032, Fig. S4B), leading to the lower NLR value (p = 0.032, Fig. S4A, B). While this was not detected in the tissue level at 16 weeks, it resembled the observations in the tumour tissue at 20 weeks (Fig. 2A, F).

In contrast, the lower densities of tissue-infiltrating neutrophil and the higher TILs observed in Cxcr2 flox compared to wildtype (Fig. 2A, F) were not detected in the blood in tumour-bearing mice at 20 weeks (Fig. S4A, B). Instead, the higher monocyte and lower lymphocyte concentrations were observed in the blood in non-tumour bearing Cxcr2 flox mice compared to wildtype (Fig. S4B).

Finally, tumour-bearing mice showed higher levels of neutrophils, NLR, and eosinophils, compared to non-tumour bearing mice in wildtype (p = 0.028, 0.049, 0.025, respectively) (Fig. S4B). This indicated that, in contrast to the findings in the tissue, the presence of tumours had more effects on granulocyte and lymphocyte levels in circulation, and that it confirms the association of higher NLR values with worse prognosis in terms of tumour presence.

Taken together, the levels of immune cells in the blood did not always correspond with those in the bladder tissue. Interestingly, changes may have occurred in a specific population of granulocytes depending on the stage of tumour pathogenesis.

Both low and high CXCR2 expression ranges are associated with a poor prognosis in human bladder cancer

To elucidate the role of CXCR2 in human bladder cancer, we analysed the data available from TCGA Bladder Cancer (Cell 2017) cohort (n = 412) [32]. The OncoPrinter function in the cBioPortal revealed that CXCR2 deletions occurred in 8 samples (n = 8/404, 2.0%) together with deletions of CXCR1 (n = 8/8) and those of chemokine receptor CCR2 (n = 2/8) (Fig. S5A, B). Interestingly, deletions of CXCR2 occurred together with those of T cell inhibitors, PD-1 (n = 6) and CTLA4 (n = 3) (Fig. S5C), however mostly mutually exclusive with genomic alterations common in bladder cancer, such as FGFR3 and TP53 (Fig. S5D). By Spearman correlation analysis, we found that the level of CXCR2 mRNA was inversely correlated with age, disease stage, metastatic stage, histological grade and mutation count (Fig. S6A). CXCR2 was not associated with specific MIBC molecular subtype, unlike the example of FGFR3 that is associated with luminal papillary subtype [32, 37] (Fig. S6B).

Kaplan-Meier analysis showed a worse overall survival associated with a lower level of CXCR2 mRNA expression, with a median survival of 32.0 months, in contrast to 92.9 months in patients with higher expression (log-rank, p = 0.044) (Fig. 3A). Previous studies, however, mostly reported the pro-tumour role associated with an increase of CXCR2 signalling [25, 26, 28, 38, 39], including those in bladder cancer [29–31]. Investigating the highest level of CXCR2 mRNA (≥80 log2 expression), we found that the median survival was 46.8 months, while 97.0 months in the medium range of CXCR2 mRNA (≥40, < 80 log2 expression) (Fig. S7A). CXCR2 expression was positively associated with the levels of CXC ligands, including CXCL1, CXCL6 and CXCL8, in both CXCR2^low (< 40 log2 expression) and CXCR2^high (≥80 log2 expression) ranges (Figs. 3B, S7B). Interestingly, in the CXCR2^low group, PD-L1 expression level was inversely correlated with CXCR2 expression (Spearman r -0.124, p = 0.041). In contrast, in the CXCR2^high group, the correlation was positive (Spearman r 0.281, p = 0.023) (Figs. 3C, S7), which is in line with the previous studies where increased CXCR2 expression was associated with an immune suppressive, pro-tumour role.

Fig. 3

Low CXCR2 mRNA expression is associated with poor overall survival in human bladder cancer. (A) Kaplan-Meier survival analysis was performed using a bladder cancer dataset (TCGA, Cell 2017) with information on CXCR2 mRNA expression and overall survival (n = 400), categorised according to low (< 40 log2 mRNA expression, n = 269), and high (≥40 log2 mRNA expression, n = 131) CXCR2 mRNA expression (log2). (B, C) Using the MIBC cases of the Bladder Cancer, TCGA, Cell 2017 cohort (> T2, n = 398), correlation between CXCR2 and CXC ligands, and that of PD-1/PD-L1 was analysed. The mRNA expression of CXCR2 and CXCL1, CXCL6 and CXCL8 were positively correlated (B). The mRNA expression of CXCR2 and PD-L1 were inversely correlated in the CXCR2 “low” range (< 40 log2 mRNA expression, n = 269), while positively correlated in the CXCR2 “high” range (≥80 log2 mRNA expression, n = 65) (C). (D) Mice aged beyond 20 weeks (21–43 weeks) were monitored by ultrasound imaging for the presence of bladder tumours and culled when mice showed clinical signs of bladder tumour, such as haematuria and weight loss. (E) Relationship between the timepoint of tumour development and tumour area size, as well as immune cell density was evaluated by Spearman correlation in wildtype mice that developed tumour at 20 weeks (n = 12) and beyond 20 weeks (n = 10) from the start of OH-BBN treatment. wildtype mice were treated with 0.05% OH-BBN for 10 weeks. Mice aged beyond 20 weeks (21–43 weeks) were monitored by ultrasound imaging for the presence of bladder tumours and culled when mice showed clinical signs of bladder tumour, such as haematuria and weight loss. (F) The status of immune cell infiltration was evaluated using Spearman correlation between Ly6G⁺ neutrophil density and T cell populations positive with CD3⁺, CD8⁺, Granzyme B⁺, and FoxP3⁺ in tumours as in E.

Status of immune suppression in the OH-BBN treated mice

To better understand how the above observations in human bladder cancer reflect in our mouse model, we examined the expression of regulatory genes in inflammation and immunity using an RNA array approach. Upregulation was observed in both immunostimulatory and immunosuppressive factors in Cxcr2 flox tumours comparing to wildtype tumours at 20 weeks (n = 3 wildtype, n = 2 Cxcr2 flox) (Table S3), including that of a pro-inflammatory factor, interferon γ (Ifng), as well as immunosuppressive factors produced by regulatory T cells, Transforming growth factor β (Tgf β) and interleukin 10 (Il10). Pcdc-1/Pd-1 and Cd274/Pd-L1 were also upregulated (4.55 and 2.84 log2 fold, respectively) in the absence of Cxcr2. Interestingly, Ccr2, a chemokine receptor that plays a major role in monocyte chemotaxis [40], and its ligand Ccl2, were highly upregulated in Cxcr2 flox tumours (6.68 and 3.16 log2 fold, respectively).

As our current OH-BBN model is optimised for the evaluation of tumour formation and initial tumour pathogenesis, it is possible that tumours had not been allowed to develop long enough to acquire a strong immunosuppressive tumour microenvironment. To assess whether immune suppression could be developed in the tumour microenvironment of our OH-BBN model, we quantified the density of Ly6G⁺ neutrophils, F4/80⁺ macrophages and TILs in tumour-bearing wildtype mice, at later timepoints, up to 43 weeks from the start of the carcinogen treatment (Fig. 3D). The majority of F4/80⁺ macrophages were observed in the stroma and rarely observed in tumours at 20 weeks and the timepoints beyond (Fig. 2B and data not shown). Longer timepoints were associated with fewer Ly6G⁺, CD3⁺ and CD8⁺ cells, but not with tumour area size (Fig. 3E). Tumour area size was not corelated with immune cell densities (Fig. S8A). In contrast, a higher density of Ly6G⁺ cells was associated with a higher overall populations of TILs (Fig. 3F), indicating that Ly6G⁺ neutrophils remained to bear an anti-tumour role, being populated together with T cells in tumours.

Taken together, our OH-BBN-induced Cxcr2 flox mouse model may reflect human bladder cancer whereby CXCR2 expression is low, in terms of worse disease prognosis, and an elevated level of immunosuppressive factors, such as Pd-1/Pd-L1. The status of immune suppression in Cxcr2 flox mice remained ambiguous, as upregulation was found in both immunostimulatory and immunosuppressive factors. However, it is unlikely that tumours in our current OH-BBN model developed an immunosuppressive tumour microenvironment, in which myeloid cells turned pro-tumour and suppressed T cell functions.