The immune TME composition differs between CR patients and non-responders (NR)

We first compared the composition of immune TME at baseline, between patients achieving or not the CR after the IP (CRIP and NRIP, respectively). The densities of CD4+ T cells, B lymphocytes (CD20+ cells) and tumor-associated macrophages (TAMs; CD68 + CD163- and CD68 + CD163+ cells) were comparable between the two patient groups, while natural killer cells (CD56+) and neutrophils (neutrophil elastase+) appeared negligible in all patients (Supplementary_Figure_1). Intriguingly, NRIP exhibited a higher intra-tumoral infiltration of CD8+ cytotoxic T lymphocytes (CTLs) as compared to CRIP patients (Fig. 2a). Notwithstanding, in NRIP a higher percentage of such CTL were in proximity to CD4 + FoxP3+ T-regulatory cells (Treg; Fig. 2b) or TAMs (Fig. 2c-e). Additionally, the analysis of the number and type of cell interactions carried out progressively moving away from the tumor margin revealed that CRIP patients had less CTLs in proximity to the tumor edge but their interactions with immunosuppressive subsets (Treg and TAMs) remained constantly low (Fig. 2f). Conversely, CTLs present in NRIP within a distance ranging from 0 to 20 μm from tumor edge were more numerous but also closer to Treg or TAMs, these interactions progressively decreasing only moving away from the tumor edge (Fig. 2f).

Fig. 2

Characterization of the immune infiltrate in bladder CIS at baseline in patients achieving or not a CR after the intensive phase. a Intra-tumoral density (cells/mm2) of CD8+ T cells. b (Left) Representative image of a bladder CIS sample stained with the first mIF panel. In the crop, the proximity between CD8+ cells (magenta staining) and a CD4 + FoxP3+ cell (white and green staining) is highlighted. Original magnification × 20. (Right) Percentage of CD8+ T cells within a radius of 20 μm from CD4 + FoxP3+ Treg cells within the tumor regions. c (Left) Representative image of a bladder CIS sample stained with the first mIF panel. The color code is the same as in (b). In the crop, the proximity between a CD68 + CD163+ macrophage (grey and orange staining) and CD8+ cells (magenta staining) is highlighted. Original magnification × 20. (Right) Percentage of CD163+ M2-polarized macrophages within a radius of 20 μm from CD8+ T lymphocytes. d Percentage of CD68 + CD163- macrophages within a radius of 20 μm from CD8+ T lymphocytes within the tumor regions. e Mean distance (μm) between each CD68 + CD163- macrophage and the nearest CD8+ T lymphocyte. Significantly different data are represented by *p < 0.05. Floating box extends from 25th to 75th percentiles, line through the box indicates median, and bars extend from the smallest to largest values. f Schematic representation of the analysis of the number and type of cell interactions carried out progressively moving away from the tumor margin. Bubble graphs show the percentage of cell-to-cell interactions (dimension of the bubbles), progressively moving away from the tumor margin in CRIP and NRIP patients

In the biopsies collected after the IP, the TME of CRIP patients resulted enriched in CD4 + FoxP3- T cells as compared to NRIP (Fig. 3a,b). Moreover, the mean distance between CD4+ T lymphocytes and CK+ cells (Fig. 3c) or CD8+ T cells (Fig. 3d) was shorter in CRIP patients. On the other hand, NRIP presented more abundant CD68 + CD163+ TAMs in the stromal compartment (Fig. 3e), and closer interactions between macrophages and CK+ cells (Fig. 3f) or CTLs (Fig. 3g), as compared to CRIP patients.

Fig. 3

The impact of the TME contexture in patient response to ONCOFID-P-B™. a-g Characterization of the immune infiltrate in bladder CIS collected after the intensive phase in patients achieving or not a CR after the intensive phase. h-k Characterization of the immune infiltrate in bladder CIS collected at baseline in patients achieving or not a CR at the end of the 15-month study. Significantly different data are represented by *p < 0.05, **p < 0.01. Floating box extends from 25th to 75th percentiles, line through the box indicates median, and bars extend from the smallest to largest values. l-p Kaplan-Meier survival curves for disease-free survival according to the immune cell composition and cell-to-cell interactions at baseline in ONCOFID-B-P™-treated bladder CIS patients. The median cut-off of each immune variable was used to separate high and low infiltrated groups. Log-rank p values, hazard ratios (HR) and 95% confidence intervals (CI) are reported in each graph

Finally, we compared the TME of the biopsies collected at baseline according to the clinical response reached at the end of the study (CRend and NRend). In this case, both intra-tumoral Treg and macrophages were more abundant in NRend samples (Fig. 3h,i), with macrophages being closer to tumor cells (Fig. 3j) and CTLs (Fig. 3k).

The higher Treg/T cell ratio and the shorter mean distance between CTLs and Treg were associated with a shorter DFS (Fig. 3l,m). Furthermore, a longer DFS was associated with an overall lower density of macrophages (Fig. 3n) and, in particular, of intra-tumoral TAMs (Fig. 3o). Moreover, a higher percentage of tumor cells in close proximity to macrophages was associated with a shorter DFS (Fig. 3p). Collectively, these observations suggest that, rather than the mere presence of CD8+ T cells within the TME, are the interactions between such T lymphocytes and immunosuppressive cells that limit their anti-tumoral activity, to play a key role in bladder CIS progression. Moreover, we identified a key negative predictive role for Treg and TAMs in the patient response to ONCOFID-P-B™.

HA receptors expression differs between responding and non-responding patients

We examined the expression and distribution of the principal HA receptors, namely the CD44 as the standard isoform (CD44s) and its most represented variants (CD44v3, CD44v6, CD44v9), ICAM-1 and RHAMM (Fig. 4). The staining revealed differential expression patterns among HA receptors in bladder CIS: CD44s was preferentially expressed in the basal urothelial cell layer and lamina propria, CD44v6 in the basal urothelial cell layer, while CD44v3 and CD44v9 were strongly evidenced in the basal and intermediate urothelial cell layers (Fig. 4a). ICAM-1 and RHAMM were distributed throughout the urothelium (Fig. 4b).

Fig. 4

Assessment of HA receptors expression and distribution in bladder CIS samples. a Representative 9-color multispectral image of the second mIF panel. Markers and color codes are indicated in the figure. Single markers assessment of the CD44 variants is depicted around the merged image. b Representative 4-color multispectral image of the third mIF panel. Markers and color codes are indicated in the figure. Single markers assessment of the ICAM-1 and RHAMM is depicted on the right of the merged image. Original magnification 20x

In all baseline biopsies we found the expression of at least one CD44 isoform on tumor tissue, with the different variants being often co-expressed by cancer cells and CD44v9 the most represented (Fig. 5a). ICAM-1 and RHAMM were expressed in the majority of tumor cells (Fig. 5a). Only a small proportion of CD44 isoform-expressing cancer cells was also positive for the Ki-67 proliferation marker (Supplementary_Figure_2a).

Fig. 5

Correlation between HA receptors expression at baseline and patient response to ONCOFID-P-B™. a Percentage of tumor cells expressing each HA receptor at baseline. b-c Expression (counts/mm2) of each HA receptor at baseline according to clinical response (b) after the 12-weekly ONCOFID-P-B™ instillation of the intensive phase and (c) at the end of the 15-month study. Significantly different data are represented by *p < 0.05, **p < 0.01. Floating box extends from 25th to 75th percentiles, line through the box indicates median, and bars extend from the smallest to largest values. d Kaplan-Meier survival curves for disease-free survival according to the expression of CD44v6 at baseline in ONCOFID-B-P™-treated bladder CIS patients. The median cut-off of each variable was used to separate high and low groups. Log-rank p values, hazard ratios (HR) and 95% confidence intervals (CI) are reported in the graph. e Correlation between CD44v6 expression and the density of total (Spearman r = 0.5193 95% CI 0.07 to 0.79; p = 0.022) or intra-tumoral CD68+ macrophages (r = 0.54 95% CI 0.09 to 0.80; p = 0.016). f Kaplan-Meier curves for disease-free survival stratifying patients according to the expression of CD44v6 and the intra-tumoral CD68 density. The median value of each variable was used as cut-off to identify high and low subgroups. Log-rank p values, hazard ratios (HR) and 95% confidence intervals (CI) are reported in each graph. g Receiving Operator Curve (ROC) showing the performance of the combined CD44v6 expression and intra-tumoral CD68+ cells to predict patient response to ONCOFID-P-B™

Among the CD44 isoforms considered independently, NRIP turned out to express significantly higher levels of CD44v6 as compared to CRIP patients (Fig. 5b). Moreover, NRIP showed a higher expression of both ICAM-1 and RHAMM, as compared to CRIP patients (Fig. 5b). This trend was maintained in CRend and NRend patients (Fig. 5c). Accordingly, the only variable with a predictive value was CD44v6, as patients with a high CD44v6 expression had a shorter DFS as compared to patients with a low expression of the isoform (Fig. 5d and Supplementary_Figure_2b).

Moreover, a direct correlation existed between CD44v6 expression and the density of CD68+ macrophages (Spearman r = 0.5193 95% CI 0.07 to 0.79; p = 0.022), and in particular with intra-tumoral macrophages (Spearman r = 0.54 95% CI 0.09 to 0.80; p = 0.016; Fig. 5e). Thus, patients were divided in two groups (CD44v6low/CD68low versus CD44v6high/CD68high) depending on CD44v6 expression level and intra-tumoral CD68+ macrophages density, to be thereafter correlated with DFS. CD44v6high/CD68high patients had a significantly worst prognosis as compared to CD44v6low/CD68low patients (Fig. 5f). The derived integrated score had an AUC of 0.85 (95%CI 0.68–1.00) for patient response prediction (Fig. 5g). Therefore, the combined evaluation of both CD44v6 expression and intra-tumoral macrophage density provided a biomarker with increased predictive value for patient response to ONCOFID-P-B™.

ONCOFID-P-B™ modulates immune cell populations and HA receptors expression

We then analyzed the changes in immune subsets and HA receptors expression on tumor cells induced by ONCOFID-P-B™ treatment. CTLs, Treg and B lymphocytes densities were minimally modified by ONCOFID-P-B™ treatment both in CRend and NRend patients (Fig. 6a-c). On the other hand, CD4 + FoxP3- T cells increased after the IP only in responding patients, to return thereafter to the basal level after the MP (Fig. 6d). Conversely, while ONCOFID-P-B™ treatment induced very limited variations in macrophage densities, we observed a trend for an increase in TAMs when patients relapsed during the MP (NRMP) (Fig. 6e,f).

Fig. 6

Changes in immune subsets and HA receptors expression on tumor cells induced by ONCOFID-P-B™ treatment. Density (cells/mm2) of (a-f) immune cell populations and (g-l) tumor cells expressing each HA receptor at baseline, after the intensive phase (IP) and during or after the maintenance phase (MP) in responding and non-responding patients. Significantly different data are represented by *p < 0.05 and **p < 0.01. Floating box extends from 25th to 75th percentiles, line through the box indicates median, and bars extend from the smallest to largest values

Regarding HA receptors expression, all CD44 isoforms were affected by the treatment in either patient groups, likely a feature reflecting a direct interaction between ONCOFID-P-B™ and HA receptors. Indeed, the density of tumor cells expressing CD44s appeared significantly increased in CRend patients after MP (Fig. 6g), while CD44v3 progressively decreased throughout the treatment protocol (Fig. 6h). Moreover, CD44v6 and CD44v9 in either patient groups appeared increased at the end of the treatment, albeit not significantly (Fig. 6i,j). Differently, the changes in ICAM-1 and RHAMM expression induced by ONCOFID-P-B™ were very limited both in responding and non-responding patients (Fig. 6k,l).

In CRend patients, a normal bladder immune contexture is re-established after the maintenance phase

Four normal bladder samples were also collected, stained and analyzed to compare their immune infiltrate and HA receptors expression pattern with those observed in bladder CIS samples. Baseline bladder CIS specimens had a higher density of Treg, B lymphocytes and macrophages as compared to normal bladder (Fig. 7a). Normal epithelial cells stained moderately for CD44v9 and RHAMM, low for CD44s and ICAM-1, and negligibly for CD44v3 and CD44v6 isoforms (Fig. 7a).

Fig. 7

Comparison of immune TME and HA receptors expression between bladder CIS and normal bladder samples. a Density (cells/mm2) of immune cell populations and tumor cells expressing each HA receptors in bladder CIS collected at baseline and normal bladders. b Density (cells/mm2) of immune cell populations and tumor cells expressing each HA receptors in bladder CIS samples collected after the IP and in normal bladders. Bladder CIS patients were grouped according to the clinical response (NRIP: patients who did not respond to ONCOFID-P-B™ treatment after the intensive phase; CRend: patients with a complete pathological response at the end of the 15-month study; NRMP: patients who relapsed during the maintenance phase). c Density (cells/mm2) of immune cell populations and tumor cells expressing each HA receptors in bladder CIS samples collected after the MP and in normal bladders. Bladder CIS patients were grouped according to the clinical response (CRend: patients with a complete pathological response at the end of the 15-month study; NRMP: patients who relapsed during the maintenance phase). Significantly different data are represented by *p < 0.05 and **p < 0.01. Floating box extends from 25th to 75th percentiles, line through the box indicates median, and bars extend from the smallest to largest values

The immune contexture of bladder CIS samples collected after the IP appeared similar to what observed in NB (Fig. 7b). Similarly, in CRend and NRMP patients, the differences in the expression of HA receptors tended to smooth as compared to normal bladders, with the exception of CD44v3 that remained still elevated (Fig. 7b). However, in NRIP the expression of CD44v3, CD44v9, ICAM-1 and RHAMM was more elevated as compared to CRend patients and normal bladders (Fig. 7b).

Finally, in biopsies collected during or after the MP, the density of infiltrating immune cells in CRend patients was comparable to normal bladders (Fig. 7c). Conversely, in NRMP, Treg and macrophages resulted still elevated as compared to normal bladders (Fig. 7c). Moreover, we observed a trend for a higher expression of HA receptors in CRend and NRMP patients as compared to normal bladders (Fig. 7c).

Gene-based cell types and signatures are differentially expressed between CRend and NRend patients, and between patients with high or low CD44v6 expression

We investigated potential differences in gene expression in baseline biopsies between CRend and NRend patients. Due to the very limited tumor tissues available, only 3 CRend and 3 NRend successfully passed the quality controls, and therefore were considered for the subsequent gene expression analysis. In CRend patients, a trend for a higher expression of ALCAM, ITGAE and CXCL16 genes was observed (Supplementary_Figure_3a). Based on the expression of cell-type and signature-associated predefined genes present in the panel, we found a trend for a higher expression of genes associated to CD8 T cells, Th1 cells, Treg cells, Exhausted CD8 cells and macrophages in NRend patients as compared to CRend (Supplementary_Figure_3b). Moreover, several gene signatures were found differentially regulated between CRend and NRend patients (Supplementary_Figure_3c). To validate these results, the TME of these selected patients was analysed in terms of immune cell infiltration and spatial distribution, and HA receptors expression. A trend for a higher infiltration of Treg cells and TAMs was found in NRend patients as compared to CRend (Supplementary_Figure_3d), as well as a higher percentage of CD8+ T cells in close proximity to Treg cells and TAMs (Supplementary_Figure_3e). Moreover, NRend patients disclosed a trend for a higher expression of HA receptors except for CD44v9 (Supplementary_Figure_3f).

In addition, we stratified the 6 patients according to the expression of CD44v6 (higher or lower than the median), and found that patients with a higher expression of CD44v6 had a lower ratio between TILs-related and exhausted CD8-related genes (Supplementary_Figure_3g). Moreover, in patients with higher CD44v6 levels, genes related with regulation, chemokines, macrophage functions and T cell functions were overexpressed as compared to patients with lower CD44v6 levels. Conversely, genes related to transporter functions, tumor-inflammation signature, cytotoxicity, antigen processing and adhesion were downregulated in patients with higher expression of CD44v6 (Supplementary_Figure_3h).

The combined evaluation of CD44v6 coding transcript and estimated macrophage infiltration is an independent prognostic biomarker in the TCGA bladder cancer cohort.

We considered the whole TCGA BLCA dataset that includes both clinical and RNA-seq data of 404 patients, to correlate the expression of the transcript encoding the CD44v6 isoform and the CD68+ cell fraction in tumor samples with clinical features of patients (Fig. 8a). Of note, the bulk RNA-seq sample deconvolution analysis using CIBERSORTx allowed to estimate the fraction of 22 immune cell types, including three subsets of CD68+ macrophages (M0, M1, and M2).

Fig. 8

The prognostic value of CD44v6 transcript and estimated intra-tumoral macrophages in the TCGA Bladder Cancer (BLCA) cohort. a Flowchart of the analysis performed on the TCGA BLCA clinical and transcript expression data. b, c Kaplan-Meier survival curves for overall survival of the TCGA BLCA samples considering combined expression levels of CD44v6 transcript and b) estimated total fraction of CD68+ cells or c) the M0 macrophage fraction. High and low levels were computed according to the median in the dataset

In the TCGA cohort, stage I tumors were quite rare (5 cases, 1.2%) and therefore we considered grouping with stage II cancers (overall 131 cases, 32.4%), while stages III and IV accounted for 34.7 and 32.9%, respectively (Supplementary_Table_2).

Patient stratification according to the median expression value of the CD44v6 isoform-coding transcript, disclosed that the high expression of CD44v6 was associated with an increased risk (HR 1.22, 95% CI 0.91-1.65, p = 0.179; Supplementary_Table_2). Moreover, we stratified patients according to the median value of all estimated CD68+ macrophages in the tumor samples (M0 + M1 + M2). In univariate analysis, CD68high cases had a significantly worse prognosis as compared to CD68low patients (HR 1.61, 95% CI 1.19-2.18, p = 0.002) (Supplementary_Table_2). Notably, the combination of the two factors evidenced that patients with CD44v6high/CD68high had a significantly worse prognosis than CD44v6low/CD68low patients (HR 2.02, 95% CI 1.30-3.15, p = 0.002; Fig. 8b and Supplementary_Table_2), with a median survival of 19.4 months versus 86.8 months, respectively. The prognostic value of the combination remained significant in multivariate analysis that also considered tumor stage and patient age (HR 1.85, range 1.18-2.89, p = 0.007), both significant predictors of outcome [20] (Supplementary_Table_2 and Supplementary_Figure_4).

Since the deconvolution analysis allowed to estimate the fraction of three distinct macrophage populations (M0, M1 and M2), we investigated the prognostic role of such subtypes more in detail. We observed that the M0 population was linked with an increased risk (HR = 2.07, 95% CI 1.52-2.82, p < 0.001), whereas the M1 and the rarest M2 macrophage fractions did not have prognostic relevance (Supplementary_Table_3). Combining the evaluation of both CD44v6 transcript expression and M0 macrophages, patients with CD44v6high/M0high had a significantly increased risk than patients with CD44v6low/M0low (HR 2.53, 95% CI 1.64-3.92, p < 0.001; Fig. 8c and Supplementary_Table_3), which was even higher than considering the CD68+ cells altogether. This risk remained significantly higher also in multivariate analysis with age and stage (2.21, 95% CI 1.42-3.46, p < 0.001) (Supplementary_Table_3 and Supplementary_Figure_5).