1. Introduction

Non–muscle-invasive bladder cancer (NMIBC) accounts for approximately 75% of newly diagnosed urothelial bladder cancers and is characterised by frequent intravesical recurrences and a meaningful risk of progression to muscle-invasive disease despite complete transurethral resection and adjuvant intravesical therapy^1–3. Risk estimation in daily practice is based on stage, grade, tumour size and number, concomitant carcinoma in situ and prior recurrence pattern, and is summarised in the EORTC and CUETO risk tables and in the updated EAU risk model^4–6. These tools guide the use of BCG, maintenance schedules and early radical cystectomy, but their accuracy for high-grade recurrence and progression is limited and patients within the same risk group often follow divergent clinical courses, particularly in high-risk NMIBC^5,6.

The bladder tumour microenvironment has emerged as a major determinant of clinical course. Most studies have focused on adaptive immunity, including CD8⁺ tumour-infiltrating lymphocytes, regulatory T cells, B cells and tertiary lymphoid structures, and on checkpoint pathways targeted by PD-1/PD-L1 inhibitors^7,8. In contrast, the innate arm in NMIBC has been explored less systematically, although it is directly engaged by intravesical agents and shapes downstream T- and B-cell responses. Over the last few years, however, the field has moved rapidly, driven by single-cell and spatially resolved profiling that has generated NMIBC-specific data beyond conventional immunohistochemistry. Spatial transcriptomics/digital spatial profiling, multiplex immunofluorescence and high-dimensional cytometry are now delineating myeloid states, their spatial niches, and their crosstalk with adaptive immunity in relation to clinicopathological risk and BCG outcomes, strengthening the rationale for an updated synthesis. Neutrophils (tumour-associated and tumour-infiltrating neutrophils; TANs and TINs), tumour-associated macrophages (TAMs), natural killer (NK) cells and dendritic cells (DCs) form the core of this compartment and display marked functional plasticity, with context-dependent anti-tumour and pro-tumour activities^9–11.

Quantitative and translational data indicate that neutrophils and TAMs often adopt pro-tumour, immunosuppressive phenotypes in established disease, associate with higher stage and grade and correlate with increased recurrence, progression and BCG failure, whereas NK cells and functionally competent DCs are required for effective cytotoxic responses and durable immune surveillance^12–16. Systemic indices such as the neutrophil-to-lymphocyte ratio (NLR) and derived inflammatory scores, together with tissue-based measures of tumour-infiltrating neutrophils, M2-skewed TAM profiles, NK-cell density and DC maturation state, have emerged as candidate biomarkers that capture this balance^17–20. At the same time, early clinical work on macrophage-modulating nanoimmunotherapies, DC-targeted strategies and NK-directed approaches suggests that manipulation of innate populations may improve outcomes in high-risk and BCG-unresponsive NMIBC^21–23.

This review summarises current evidence on neutrophils, TAMs, NK cells and DCs in NMIBC, with emphasis on their biology, spatial organisation and interaction with adaptive immunity, and on their impact on clinicopathological features, prognosis and response to BCG. It also discusses how innate-derived biomarkers may complement existing risk models and outlines emerging therapeutic strategies that target innate immunity in high-risk and BCG-unresponsive disease. Because post-BCG endpoints are variably defined across primary studies, we report BCG-related outcomes as stated by each study and avoid re-labelling broad terms ( “BCG failure”) as “BCG-unresponsive” or “non-response” unless explicitly defined.

2. Material and Methods

A focused literature search was performed in PubMed up to December 2025 to identify studies on innate immune cells in non–muscle-invasive bladder cancer (NMIBC), with emphasis on prognostic associations and response to intravesical bacillus Calmette–Guérin (BCG). Search terms were entered as combinations of the following keywords: “non-muscle-invasive bladder cancer”, “NMIBC”, “Ta”, “T1”, “carcinoma in situ”, “CIS”, “Bacillus Calmette–Guérin”, “BCG”, “intravesical BCG”, “neutrophils”, “tumour-associated neutrophils”, “tumour-infiltrating neutrophils”, “neutrophil-to-lymphocyte ratio”, “NLR”, “macrophages”, “tumour-associated macrophages”, “CD68”, “CD163”, “natural killer cells”, “NK cells”, “dendritic cells”, “recurrence”, “progression”, “survival”, “prognosis”, “BCG response”, and “BCG failure”. To maximise capture of relevant evidence, reference lists of included papers and selected relevant reviews were also screened.

We included clinical studies in NMIBC patients reporting tissue- or blood-based measures of neutrophils, macrophages, natural killer cells, or dendritic cells in relation to clinicopathological features and/or outcomes (recurrence, progression, survival, BCG response/failure). We also included NMIBC-focused translational studies using methods such as immunohistochemistry, multiplex immunofluorescence, or transcriptomic profiling, as well as preclinical in vitro or in vivo studies providing mechanistic insight relevant to innate immune regulation and/or BCG-related effects in bladder cancer models. We excluded studies focused on non-urothelial histologies, studies limited to muscle-invasive disease without extractable NMIBC data, case reports and very small case series, and abstract-only records without accessible full text. Non-English publications were excluded when reliable data extraction was not feasible. Titles and abstracts were screened first, followed by full-text assessment of potentially relevant articles.

3. Overview of the innate immune compartment in non–muscle-invasive bladder cancer

The innate immune compartment in NMIBC is dominated by neutrophils, TAMs, NK cells and DCs, which are recruited in response to tumour-derived danger signals, microbial products and intravesical agents such as BCG^10,11,24. These populations show pronounced functional plasticity, oscillating between pro-inflammatory, cytotoxic states and immunoregulatory, tissue-remodelling phenotypes depending on local cytokines, chemokines and stromal cues^9,14. In early or low-risk lesions, innate activation can support immune surveillance, whereas in high-grade and T1 disease, the same compartments frequently become subverted to sustain tumour growth, neoangiogenesis and immune escape^12,13,25.

Neutrophils and monocyte-derived macrophages are usually the numerically dominant innate populations in established NMIBC. Neutrophils accumulate in papillary fronds, tumour–stroma interfaces and perivascular regions and, once polarised towards an N2-like programme, contribute to matrix degradation, vascular endothelial growth factor release, arginase-1–mediated T-cell suppression and COX-2/PGE₂/IDO1-driven immune dysfunction^12,25,26. TAMs, which in NMIBC are predominantly skewed towards an M2-like phenotype, concentrate in tumour-associated stroma, the papillary axis and macrophage-rich carcinoma in situ, where they produce IL-10, TGF-β, VEGF, MMPs and CXCL8, foster angiogenesis and attract regulatory T cells^13,14,27,28. In contrast, NK cells and functionally competent DCs are key effectors of anti-tumour immunity: NK cells provide rapid, MHC-independent cytotoxicity and IFN-γ production, while cDC1 and mature CD83⁺ DCs are required for efficient priming of Th1-polarised CD4⁺ and cytotoxic CD8⁺ T-cell responses to BCG and other intravesical immunotherapies^15,16,23,29. Quantitative or qualitative defects in NK and DC compartments have been linked to exhausted T-cell states, loss of tertiary lymphoid structure organisation and BCG failure in translational and preclinical work^30–32.

Clinically, this innate balance can be captured by a combination of systemic, tissue and molecular readouts. Peripheral indices such as the NLR and related composite scores, such as systemic inflammation response index (SIRI) integrate neutrophilia, monocytosis and lymphopenia and have been repeatedly associated with higher clinicopathological risk, BCG non-response, recurrence, progression and mortality in NMIBC^{18,19,33–35}. Tissue-based measures, including tumour-infiltrating neutrophils, tissue NLR, CD68⁺ and CD163⁺ TAM densities and CD163/CD68 ratios, provide spatially resolved information on neutrophil- and macrophage-rich niches and correlate with high-grade and T1 tumours, diffuse carcinoma in situ and inferior recurrence-free survival^13,14,27,36. More recently, NK- and DC-related signatures derived from multiplex immunohistochemistry, bulk RNA sequencing and single-cell datasets have started to distinguish inflamed but functionally competent microenvironments from myeloid-dominant, “hot but suppressed” states and to inform the design of macrophage-, DC- and NK-directed strategies, including nano-immunotherapies, intravesical CD40 agonists, alternative bacterial vaccines such as Ty21a and adoptive NK-cell approaches^{16,21–23,32}. Key phenotypic and functional features of these innate cell populations in NMIBC are summarised in Table 1.

4. Neutrophils in NMIBC

4.1 Biology and phenotypes (N1 vs N2, TANs/TINs, NLR-based indices)

Neutrophils are among the earliest and most abundant innate cells recruited to the bladder after tumour formation or intravesical therapy and can adopt N1 (anti-tumour) or N2 (pro-tumour) phenotypes depending on the local cytokine and chemokine milieu (Jinesh). In bladder cancer, tumour-associated and tumour-infiltrating neutrophils (TANs/TINs) are enriched in biologically aggressive lesions, with high neutrophil density in tumour nests and stroma linked to higher pathological T stage, high grade and increased tissue NLR, while CD8⁺ tumour-infiltrating lymphocytes (TILs) show an inverse correlation with these adverse features^12,25.

Pro-tumour programmes involve matrix metalloproteinase 9–mediated release of vascular endothelial growth factor, arginase 1–dependent T-cell suppression and reactive oxygen species–driven DNA damage, which promote angiogenesis, immune escape and invasion^17,25. TANs can also limit CD8⁺ T-cell immunity via a COX-2/PGE₂/IDO1 axis, in which neutrophil-derived PGE₂ induces IDO1 in tumour cells, activates tryptophan metabolism and reduces interferon-γ production, thereby attenuating responses to immune checkpoint blockade²⁶.

Preclinical BCG models indicate an important early anti-tumour role. In the orthotopic MB49 model, BCG instillation induces a dense neutrophil infiltrate, and antibody-mediated polymorphonuclear depletion abolishes the survival benefit of BCG, showing that neutrophils are required for BCG efficacy in vivo. BCG-activated neutrophils secrete IL-8, GRO-α, MIP-1α and macrophage migration inhibitory factor, recruit monocytes and CD4⁺ T cells into the bladder wall and act as early immunoregulatory cells that organise adaptive responses rather than direct effectors¹⁵. Combination of BCG-stimulated neutrophils with Smac mimetics sensitises bladder cancer cells to neutrophil-derived TNFα and restores tumour cell death, suggesting that N2-like neutrophils can be pharmacologically redirected towards a more cytotoxic, N1-like phenotype⁴².

Systemic indices such as NLR, the systemic immune-inflammation index (SII = platelets×neutrophils/lymphocytes) and the SIRI = neutrophils×monocytes/lymphocytes reflect the balance between neutrophilia or monocytosis and lymphopenia in peripheral blood and are widely used as surrogates of neutrophil-driven systemic inflammation in urothelial carcinoma, including NMIBC^17,35.

4.2 Clinicopathological correlations

Histopathological series show that neutrophil infiltration increases with grade and depth of invasion. TAN counts and tissue NLR are higher in high-grade than low-grade urothelial carcinomas and peak in muscle-invasive disease, while NMIBC lesions display intermediate neutrophil densities with relatively preserved CD8⁺ TIL infiltrates, indicating that NMIBC lies between an inflamed but partially controlled state and neutrophil-dominated, immunosuppressed muscle-invasive cancer^12,25. In the NMIBC subset of one cohort, high TIN density already associates with increased intravesical recurrence, linking neutrophil infiltration to early tumour dynamics rather than muscle invasion alone¹².

Peripheral NLR broadly mirrors these tissue findings. Higher pre-TURBT NLR is reported in T1 versus Ta tumours and in high-grade versus low-grade NMIBC, and associates with multifocality, tumour size >3 cm, concomitant carcinoma in situ (CIS) and higher EORTC risk scores^33,34,43,44. In an EORTC-stratified study, mean NLR increases from low- to intermediate- to high-risk groups for both recurrence and progression, showing that NLR tracks established clinicopathological risk models⁴³. A meta-analysis across urothelial carcinomas confirms that elevated pretreatment NLR is associated with higher pathological T stage, higher grade and lymphovascular invasion, supporting a robust association between neutrophil-dominant systemic inflammation and adverse tumour biology¹⁷.

Prospective and retrospective NMIBC series also link inflammatory indices to early events. Patients who recur within the first year have higher baseline NLR and lower lymphocyte-to-monocyte ratio, and larger tumours, T1 stage and high grade often co-segregate with elevated NLR or related indices^44,45.

4.3 Prognostic and predictive value

4.3.1 Tissue-based markers

In a mixed NMIBC/MIBC study, high TIN density and high tissue NLR are associated with significantly worse overall survival and remain independent adverse prognostic factors alongside pathological T stage, whereas high CD8⁺ TIL density correlates with better survival but loses significance after adjustment. Within the NMIBC subset, high intratumoural neutrophil infiltration predicts shorter relapse-free survival, supporting TINs as a negative tissue biomarker even at non–muscle-invasive stages¹².

4.3.2 Peripheral markers before TURBT

Multiple single-centre series link pre-TURBT NLR with recurrence and progression. Cut-offs in the range of approximately 1.7–3.0 independently predict recurrence and/or progression and are higher in T1, multifocal, CIS-containing and EORTC high-risk disease^33,34,46. Other datasets show that higher NLR associates with recurrence on univariable analysis, but significance is reduced after adjustment for clinicopathological factors^44,45,47.

A urothelial-wide meta-analysis across 32 studies shows that elevated pretreatment NLR predicts worse overall, cancer-specific and progression-free survival in bladder and upper tract urothelial carcinoma across stages and NLR cut-offs¹⁷. A NMIBC-focused meta-analysis pooling six studies with 2,298 patients indicates that high pre-TURBT NLR predicts worse recurrence-free and progression-free survival, including in high-risk NMIBC treated with BCG, supporting NLR as a simple, low-cost prognostic biomarker in this setting¹⁸. A later meta-analysis restricted to BCG-treated NMIBC confirms that elevated pre-BCG NLR associates with higher recurrence and progression risks, although heterogeneity and non-standardised thresholds remain important limitations¹⁹.

Not all systemic indices perform equally well. In a multi-centre study of intermediate- and high-risk NMIBC on BCG, the systemic immune-inflammation index and other composite neutrophil-based scores do not clearly distinguish responders from non-responders or predict recurrence and progression beyond NLR and clinicopathological variables⁴⁸. Large NMIBC series focusing on mortality show that pre-operative NLR independently predicts overall and cancer-specific survival but adds little for recurrence or progression, suggesting that NLR may partly capture global host frailty as well as tumour biology^49,50.

4.4 Role in response to BCG (pro-tumour vs anti-tumour functions)

In preclinical models, neutrophils are required for the anti-tumour effect of intravesical BCG. In the MB49 orthotopic model, BCG instillation induces a dominant neutrophilic infiltrate, and selective polymorphonuclear depletion abolishes the survival benefit, indicating a central role for neutrophils in BCG-mediated tumour control¹⁵. This is not a contradiction as much as a matter of timing and context. In BCG-treated tumours, an early, brisk neutrophil influx appears to be part of the effector cascade required for tumour control, consistent with the loss of benefit after polymorphonuclear depletion in vivo^15,42. In contrast, when neutrophil recruitment is sustained, either systemically (high baseline NLR/SIRI) or intratumourally (high TIN/TAN density), it more often reflects a chronic, myeloid-skewed inflammatory state that associates with non-response and poorer long-term outcomes after BCG^49,51–53. Mechanistically, neutrophil-dominant circuits reported in urothelial carcinoma, including prostaglandin–IDO1-linked immune suppression, provide a plausible basis for reduced durability of BCG-induced cytotoxic T-cell activity and support rational combination strategies^26,42. BCG-activated neutrophils recruit monocytes and CD4⁺ T cells and secrete TNFα, TRAIL and Fas ligand, and these death-ligand signals can be augmented by Smac mimetics to overcome tumour resistance in vitro^15,42.

Clinical data in NMIBC link neutrophil-based studies to BCG response and long-term outcome. In a large single-centre study on a one-year BCG schedule, SIRI independently predicts BCG non-response and shorter recurrence-free and progression-free survival, whereas NLR and platelet-to-lymphocyte ratio lose significance in multivariable models, suggesting that combined neutrophil–monocyte activation with lymphopenia better captures myeloid-driven resistance³⁵. In intermediate- and high-risk NMIBC receiving BCG induction only, higher pre-operative NLR is independently associated with non-complete response at first post-induction cystoscopy and with shorter subsequent recurrence-free and progression-free survival, with a cut-off around 2.4 identifying higher-risk patients⁵⁴.

In cohorts receiving induction plus maintenance BCG, NLR thresholds around 2.3–2.5 independently predict recurrence, progression and BCG failure and improve the discriminative performance of CUETO risk models when combined, while high pretreatment NLR and derived NLR associate with worse overall and cancer-specific survival and higher cystectomy rates, even when recurrence or progression are not consistently predicted^49,51–53. Meta-analyses restricted to BCG-treated NMIBC confirm that elevated pre-BCG NLR is associated with increased recurrence and progression risk, but heterogeneity and non-standardised cut-offs limit routine use^18,19. Multi-centre analyses of SII and related indices have not shown clear associations with BCG response, indicating that not all neutrophil-containing metrics are clinically informative⁴⁸.

Neutrophil-based indices may also capture susceptibility to BCG-related toxicity. In a prospective cohort of men with high- or very high-risk NMIBC, baseline NLR ≥3 independently associates with greater worsening of lower urinary tract symptom–related quality of life during BCG induction, suggesting that elevated NLR identifies patients at risk of both oncologic failure and functional intolerance⁵⁵. Mechanistic work in urothelial carcinoma shows that neutrophil-dominated inflammatory circuits, such as PGE₂–IDO1 signalling can suppress CD8⁺ T-cell function and may constrain the durability of BCG-induced anti-tumour immunity, supporting combinations targeting these pathways^26,42. Τhe shift towards N2-like neutrophil programs in NMIBC is most plausibly promoted by persistence of inflammation rather than its acute initiation during effective BCG. Candidate drivers include TGF-β and IL-6–dominant signalling, together with COX-2/PGE₂-linked immunosuppressive circuits that can constrain effector responses and favour chronic myeloid inflammation. These mechanisms provide a coherent explanation for why neutrophils may be required early for BCG efficacy yet correlate with resistance and progression when chronic inflammatory pressure predominates ^56,57.

Overall, available data support a dual role for neutrophils in NMIBC: BCG-activated neutrophils are needed to initiate effective local anti-tumour immunity, whereas sustained neutrophil-dominant systemic and intratumoural inflammation, reflected by high TIN/TAN densities and elevated NLR-based indices, is consistently associated with higher grade and stage, increased recurrence and progression, BCG failure and worse survival. This pro-tumour neutrophil biology is also linked to metastatic dissemination. In bladder cancer models, tumour-cell ETV4 signalling increased CXCL1/8-driven TAN recruitment, and TAN-derived VEGFA/MMP9 promoted lymphangiogenesis and lymph node metastasis, while neutrophil depletion reduced both processes⁵⁸. Beyond NLR, clinically feasible ways to approximate neutrophil programs in NMIBC include tissue quantification of neutrophil infiltrates (CD66b/MPO/NE) combined with selected adjunct markers linked to suppressive myeloid features, as well as urine-based immunomonitoring during BCG using flow cytometry of urinary immune cells. Functional readouts such as NET-associated markers (MPO–DNA complexes and H3Cit) can provide complementary information, although assay standardisation and tumour specificity remain important limitations^59,60. Key data on neutrophil biology, inflammatory indices and clinical impact in NMIBC are summarised in Table 2.

5. Tumor-associated macrophages

Tumour-associated macrophages (TAMs) are major components of the NMIBC tumour microenvironment. They derive from circulating monocytes and tissue-resident macrophages and exhibit functional plasticity along an M1–M2 spectrum, from classically activated, pro-inflammatory M1 to alternatively activated, immunosuppressive M2 states^9–11. M1-like macrophages express CD80, CD86 and iNOS and produce IL-12 and TNFα, whereas M2-like macrophages express CD163, CD204 and CD206 and support tissue remodelling, angiogenesis and immune suppression^9,10. In NMIBC, TAMs are predominantly M2-skewed and contribute to tumour progression, neoangiogenesis and an immunosuppressive microenvironment^9,10,14,61. CD68 is widely used to identify macrophages but does not distinguish M1 from M2, so combined panels (for example CD68 with CD163 or CD204) and spatially resolved analyses are required for functional interpretation^10,13.

5.1 Density, localization and outcome in NMIBC

Across several cohorts, high CD68⁺ TAM density correlates with higher pathological stage, high grade, increased recurrence risk and shorter RFS^13,14. TAMs accumulate along the papillary axis, in tumour-associated stroma and peritumoural lymphoid aggregates, with particularly high density at the tumour margin, while intratumoural nests show more variable infiltration. In Ayari’s de novo pTa/pT1 series without early BCG, CD68⁺ macrophages were present in almost all papillary and stromal areas; although high versus low TAM density did not reach significance for progression, none of the low-TAM tumours progressed to muscle-invasive disease, suggesting that low macrophage burden may mark a subset with very favourable course²⁷.

M2-skewed populations provide additional prognostic detail. CD163⁺ macrophages preferentially localise to stromal and perivascular regions and associate with higher recurrence risk and inferior RFS in NMIBC, particularly when the CD163/CD68 ratio is elevated¹³. In primary T1 high-grade disease, high CD163⁺ TAM infiltration independently predicts poorer recurrence-free and disease-specific survival, indicating that M2-dominant macrophage networks are tightly linked to aggressive T1 biology²⁸.

5.2 Association with stage, grade and clinical course

Taken together, these data converge on a consistent association between M2-skewed TAM profiles and more advanced disease. High CD68⁺ and CD163⁺ densities correlate with high-grade and pT1 tumors, diffuse carcinoma in situ (CIS), larger lesions and multifocality^9,13,14. In Ayari’s papillary NMIBC cohort, mature CD83⁺ dendritic cells (TIDCs) were the strongest independent predictor of progression, but the absence of progression events in low-TAM tumors suggests that macrophage burden may act as a necessary cofactor in progression to muscle invasion in at least a subset of cases²⁷.

5.3 TAMs and BCG therapy

Macrophages are central to the mechanism of action of intravesical BCG,while they internalise mycobacteria, produce cytokines such as IL-1β and TNFα and help shape downstream Th1 responses. However, an excessive, M2-leaning baseline TAM infiltrate is repeatedly associated with suboptimal BCG efficacy. In high-risk NMIBC treated with TURBT plus BCG, high CD68⁺ macrophage density independently predicted higher recurrence risk. The study further reported an adverse subgroup characterised by high TAM density together with prominent CD68 immunoreactivity within tumour areas, with three-year recurrence approaching two-thirds in the high-TAM/high-CD68 group versus about one-third in all others. In a separate low-risk, non-BCG cohort, TAM density did not correlate with RFS, supporting a specific interaction between baseline macrophage burden and BCG response rather than a pure stage surrogate³⁶.

In CIS, increased TAM infiltration within the carcinoma is associated with poorer BCG response and higher recurrence and progression rates, indicating that macrophage-rich CIS is particularly refractory to standard intravesical immunotherapy³⁷. Macrophage phenotyping in NMIBC further shows that sustained M2 (CD163⁺) infiltration significantly limits BCG benefit, whereas M1-like polarisation aligns with better recurrence control, supporting the concept that a pre-existing M2-heavy landscape can buffer BCG-induced inflammation and favour immune escape⁶².

The interplay between TAMs, regulatory T cells and cytokines adds another layer. In a series of 154 NMIBC patients, high FOXP3⁺ Treg and CD204⁺ TAM densities together with elevated IL-6 expression were associated with unfavourable features, and in the BCG-treated subset both high Tregs and high CD204⁺ TAMs independently predicted shorter RFS, consistent with a Treg–TAM–IL-6 immunosuppressive circuit that dampens BCG-induced Th1 responses⁶³. In a high-risk CIS–T1 study undergoing en bloc resection followed by BCG, high stromal CD68⁺ infiltration combined with TERT C228T and KDM6A alterations was associated with inferior RFS, suggesting that TAM burden interacts with tumour genomics to shape BCG response⁶⁴. Single-cell RNA sequencing in primary and recurrent NMIBC shows enrichment of immunosuppressive, pro-angiogenic macrophage subclusters and altered myeloid–T-cell communication in recurrent tumours, providing a mechanistic link between TAM reprogramming and BCG failure or post-BCG recurrence^65,66.

5.4 Prognostic roles of M2-heavy microenvironments

Collectively, these studies support an “M2-heavy” microenvironment as an unfavorable signal in NMIBC, particularly under BCG pressure. However, the strength of these associations is not uniform across NMIBC clinical subgroups, reflecting differences in risk distribution (including CIS), treatment exposure (BCG-treated vs BCG-naïve) and the degree to which a given marker captures macrophage abundance versus polarization. Accordingly, CD68 alone should be interpreted primarily as a density marker, whereas combined panels (e.g., CD68 with CD163 and related ratios), ideally with spatial context, are better suited to distinguish prognostically relevant macrophage programs. High CD68⁺ and CD163⁺ densities, elevated CD163/CD68 ratios and co-enrichment of FOXP3⁺ Tregs and IL-6 associate with high-risk pathology, reduced BCG efficacy and shorter RFS^{13,36,62,63,67}. In T1 high-grade disease, CD163⁺ TAMs independently predict poor survival, while in papillary NMIBC without initial BCG, low TAM densities mark a subgroup with negligible progression risk^27,28. At the systemic level, plasma proteomic profiling has identified MMP12 as a macrophage-derived factor whose elevated levels track with poor prognosis and more advanced urothelial bladder cancer, indicating that TAM activity may be measurable in blood and could serve as a non-invasive biomarker for risk stratification and therapy monitoring. Mechanistically, MMP12 is a macrophage-enriched metalloproteinase, so elevated plasma levels are biologically compatible with increased myeloid/TAM activity. However, current clinical evidence is largely associative and does not establish MMP12 as a direct surrogate of intratumoural M2 polarisation. Key limitations are the lack of tumour specificity, potential contribution from other myeloid compartments, and variability related to sampling context and assay platform, so tissue correlation and prospective validation are required²⁰. 

5.5 Macrophage modulation by intravesical nanoimmunotherapy

Translational data in BCG-unresponsive NMIBC illustrate that pharmacologic modulation of TAMs can reshape the bladder TIME and improve outcomes. In a retrospective study of patients treated with intravesical OncoTherad® nanoimmunotherapy after BCG failure, paired biopsies before and after treatment showed a significant reduction in MAO-B expression and CD163⁺ M2 macrophages, alongside decreased FOXP3 immunoreactivity and reduced CD163⁺ macrophage counts in the lamina propria. In parallel, IFN-γ expression increased, CD8⁺ T-cell infiltration rose and CX3CR1⁺ effector T cells accumulated in urothelial and stromal compartments. A composite immune score, defined as the ratio of effector (CD8/CX3CR1) to suppressive (CD163/FOXP3) cells, shifted from intermediate to high categories after therapy and aligned with prolonged RFS. These findings provide proof-of-principle that selectively reducing M2/Treg-like suppression and enhancing cytotoxic T-cell recruitment can restore immune control in high-risk, BCG-unresponsive NMIBC²¹. Key features of tumour-associated macrophage phenotypes, spatial patterns and prognostic relevance in NMIBC are summarised in Table 3.

5.6 Targeting macrophages

Given their central role in NMIBC progression and treatment resistance, tumour-associated macrophages are attractive therapeutic targets in bladder cancer^9–11.

5.6.1 TAM depletion. 

Several agents used in other solid tumours can deplete macrophages or reduce their survival. Nitrogen-containing bisphosphonates and trabectedin induce apoptosis preferentially in monocytes/macrophages and have been proposed to reduce pro-tumour TAM pools, while colony-stimulating factor 1 receptor inhibitors target CSF1–CSF1R signalling to deplete or remodel TAMs^10,69. These strategies have not yet been systematically tested intravesically in NMIBC but provide a framework for macrophage-directed therapies.

5.6.2 Inhibition of recruitment. 

Blocking chemokine axes that drive monocyte trafficking into the bladder microenvironment, particularly CCL2–CCR2, can reduce TAM accumulation and restore anti-tumour immunity in preclinical models and pan-cancer analyses^10,69. This approach is considered a plausible future direction for bladder cancer, especially in combination with BCG or systemic immunotherapy.

5.6.3 Reprogramming towards an M1 phenotype. 

Other strategies aim to repolarise TAMs from an M2-like, immunosuppressive state towards a pro-inflammatory, M1-like phenotype. Toll-like receptor agonists, phosphoinositide-3-kinase inhibitors and histone deacetylase inhibitors can promote M1 skewing and enhance antigen presentation and effector T-cell priming in preclinical models^10,11,69. In NMIBC, intravesical nano-immunotherapies such as OncoTherad® provide proof of principle: in BCG-unresponsive patients, treatment was associated with reduced CD163⁺ M2 macrophages and FOXP3⁺ Tregs, increased IFN-γ and CD8⁺/CX3CR1⁺ T-cell infiltration and a shift towards a more favourable effector:suppressor immune score²¹.

5.6.4 Targeting TAM-derived mediators. 

TAMs in bladder cancer produce soluble factors that support tumour growth and immune evasion. Macrophage-derived MMP12 has been identified as a circulating marker of poor prognosis in urothelial bladder cancer, with plasma MMP12 levels reflecting an M2-skewed, pro-tumour microenvironment²⁰. Chemokines such as CXCL8/IL-8, produced by TAMs, are implicated in invasion, migration and angiogenesis and represent additional pharmacological targets²⁶.

5.6.5 Combination with immune checkpoint inhibition and intravesical strategies. T

AMs express and induce immune checkpoints, including PD-L1, and thereby contribute to T-cell dysfunction. Mechanistic and clinical data suggest that TAM-high tumours may be relatively resistant to BCG alone but could benefit from combinations that include checkpoint blockade^10,11,70. Integrating CSF1R or CCL2–CCR2 inhibitors, M2 to M1 reprogramming agents, or intravesical macrophage-modulating nano-immunotherapies with BCG and PD-1/PD-L1 inhibitors is a logical next step in trial design for high-risk and BCG-unresponsive NMIBC.

6. Dendritic cells

6.1 Dendritic cells and antigen-presenting networks in NMIBC

Dendritic cells are the main professional antigen-presenting cells (APCs) in the bladder tumour microenvironment and link intravesical therapy to adaptive immunity. The DC compartment in urothelial cancer includes conventional myeloid DCs (cDCs), plasmacytoid DCs (pDCs) and monocyte-derived inflammatory DCs^22,36. Cross-presenting cDC1 are particularly important, as they present exogenous antigens on MHC class I and license CD8⁺ T cells; Batf3-deficient mice lacking cDC1 lose both spontaneous immune surveillance and responses to DC-targeted agonists in orthotopic model¹⁶.

6.2 Tumour-infiltrating DCs as biomarkers

Studies of NMIBC resections show marked inter-patient heterogeneity in CD1a⁺ and CD83⁺ tumour-infiltrating DCs (TIDCs) in tumour epithelium and peritumoural stroma³⁶. In a separate study, high intraepithelial and stromal CD83⁺ TIDC density at diagnosis was independently associated with progression to muscle-invasive disease. Taken together, these data suggest that DC-based biomarkers in NMIBC should consider subset, maturation and spatial distribution, and that abundant mature TIDCs may indicate an immune-edited, aggressive phenotype rather than a uniformly favourable response²⁷

6.3 DCs in BCG-induced immunity and failure patterns

In orthotopic models, intravesical BCG rapidly recruits Ly6C⁺CD11c⁺ DCs to the bladder, where they take up BCG and tumour antigens, migrate to draining lymph nodes and prime Th1-polarised CD4⁺ and CD8⁺ T cells²². In human NMIBC, TIDCs often show low CD80/CD86/CD83 expression, consistent with incomplete maturation^27,36. DCs from urothelial carcinoma patients also overexpress inhibitory receptors such as BTLA and TIM-3, and ligation of these receptors reduces DC cytokine secretion, indicating checkpoint-mediated functional restraint³⁰. In addition, IL-10, TGF-β and indoleamine-2,3-dioxygenase in the tumour microenvironment can drive DCs towards a low–IL-12, tolerogenic phenotype and promote regulatory T-cell induction^22,31. These findings support a model in which BCG failure is frequently related to qualitative defects in DC activation and licensing under dominant immunosuppression and inhibitory signalling, rather than simple absence of DCs^22,30,31. Taken together, these data support DC dysfunction (incomplete maturation and inhibitory signalling) as a plausible mechanistic bottleneck that can blunt Th1 priming and CD8⁺ licensing despite BCG-driven inflammation, thereby contributing to clinical BCG failure^{22,27,30,31,36}.

6.4 DC-targeted therapies in NMIBC

6.4.1 α-type-1-polarised DC vaccines (αDC1)

Ex vivo DC vaccination is one approach to overcome in situ dysfunction. In male NMIBC patients, αDC1 generated from monocytes with IL-1β, TNF-α, IFN-α, IFN-γ and poly(I:C) showed higher CD86 expression and produced high, stable IL-12p70 with low IL-10, including after loading with UVB-irradiated allogeneic T24 bladder cancer cells and CD40 ligand stimulation. T24-loaded αDC1 induced autologous bladder cancer–specific cytotoxic T lymphocytes with stronger IFN-γ responses against T24 and autologous primary tumour cells than CTLs primed by standard mature DCs, with clear MHC class I/II restriction²⁹. Earlier preclinical work with Ag85A-engineered DCs in bladder cancer models similarly showed that gene-modified DC vaccines can enhance anti-tumour immunity⁷¹. Clinically, a HER2-targeted autologous DC vaccine has entered phase I testing and includes high-risk bladder cancer among other solid tumours, although data remain limited and not NMIBC-specific⁷². Overall, these studies indicate that DC vaccination is feasible in bladder cancer and that αDC1 provide a rational platform for ex vivo DC trials in BCG-unresponsive NMIBC.

6.4.2 Intravesical CD40 agonists

Agonistic CD40 antibodies offer a way to activate DCs in situ. In orthotopic models, intravesical CD40 agonist antibody reduced tumour burden, decreased PD-1⁺LAG-3⁺ exhausted CD8⁺ T cells and improved survival. These effects were lost after CD8⁺ T-cell depletion or in Batf3-deficient mice, confirming a cDC1–CD8⁺ mechanism. A fully human Fc-engineered CD40 agonist delivered intravesically produced durable local control, generated systemic memory against rechallenge and avoided the systemic toxicities seen with intravenous CD40 agonists; in BCG-unresponsive tumours, switching to 2141-V11 restored disease control, whereas continued BCG did not¹⁶. These findings support intravesical CD40 agonists, alone or combined with IL-15, as DC-centred candidates for salvage immunotherapy in high-risk NMIBC.

6.4.3 Ty21a: DC- and T-cell-oriented bacterial immunotherapy

The live-attenuated Salmonella Typhi Ty21a has been evaluated as an intravesical agent with a DC-focused mechanism. In MB49 orthotopic models, Ty21a achieved tumour control comparable to or better than BCG, with less neutrophil-dominated inflammation and relative enrichment of CD11c⁺Ly6C⁺CD103⁺ DCs; efficacy depended on these DCs and on CD4⁺ and CD8⁺ T cells, but not on neutrophils or NK cells²². A phase I trial in low/intermediate-risk NMIBC established 1×10⁸ CFU weekly as the maximum tolerated dose, with mainly grade 1 flu-like and cystitis-like adverse events and rare, low-level Ty21a persistence in urine⁷³. Immune monitoring showed a Th1-biased but moderate urinary cytokine profile and induction of systemic Ty21a-specific CD4⁺ and CD8⁺ T-cell responses, anti–LPS-Typhi IgG and strong Vδ2 γδ T-cell activation. Ty21a also increased total DCs, cDC2 and cDC1 in urine, whereas matched BCG-treated controls did not show similar DC enrichment⁷⁴. These data support Ty21a as a DC- and T cell–oriented intravesical immunotherapy with a different immune footprint from BCG^22,73,74.

6.4.4 Gene-modified DCs and future directions

Work on gene-modified DCs provides a framework for next-generation NMIBC strategies. Approaches such as mRNA electroporation, viral vectors and CRISPR/Cas9 can enhance antigen loading, enforce maturation and co-stimulation, promote CCR7-dependent migration and inhibit IL-10, TGF-β, PD-L1 or indoleamine-2,3-dioxygenase. Among these, PD-L1-silenced DCs loaded with bladder tumour antigens showed improved T-cell priming in vitro, illustrating a direct bladder-specific application³¹. In NMIBC, such concepts could be implemented via ex vivo DC products or intravesical tools designed to re-programme resident DCs along similar signalling axes^16,31,74.

6.4.5 Implications

Current evidence supports several working hypotheses. Qualitative DC features, cDC1 presence, maturation status, inhibitory receptor expression and resistance to local immunosuppression, seem to influence outcome and may add to classical risk factors, but require validation in prospective, biomarker-driven NMIBC cohorts^22,27,30. DC-focused interventions such as αDC1 vaccines, intravesical CD40 agonists and Ty21a are biologically plausible options for BCG-unresponsive disease, yet current data are preclinical or early phase and should be extended in prospective trials with integrated DC profiling^16,29,73,74. Combination approaches that pair DC-centred intravesical therapies with modulation of IL-10, TGF-β, PD-L1, IDO or IL-15 are conceptually attractive but remain experimental and should, at present, be regarded as hypotheses for translational studies rather than established therapeutic strategies^16,31. Principal data on dendritic-cell subsets, dysfunction and DC-centred therapeutic approaches in NMIBC are summarised in Table 4.

7. Natural killer cells in non–muscle-invasive bladder cancer

NK cells, mostly CD56⁺ cells, are key innate lymphocytes that recognise and kill transformed cells without prior sensitisation, providing early anti-tumour defence^7,23,32. Cytotoxicity is driven by activating receptors such as NKG2D and DNAX accessory molecule-1 and restrained by inhibitory killer-cell immunoglobulin-like receptors and sialic acid-binding immunoglobulin-like lectins (Siglecs)^23,32,40,75. In non–muscle-invasive bladder cancer (NMIBC), NK cells act in blood, tumour and urine, participate in the response to intravesical Bacillus Calmette–Guérin (BCG) and are emerging therapeutic targets^41,76.

7.1 NK subsets, education and checkpoints

Intratumoural profiling shows three NK subsets, CD56^bright, CD56^dim and CD56⁻, with CD56^bright cells displaying higher IFN-γ production, stronger cytotoxicity and favourable association with cancer-specific and overall survival, in contrast to CD56^dim and CD56⁻ cells³². Transcriptomic deconvolution of The Cancer Genome Atlas bladder cohort identifies a dominant CD56^bright-like NK signature that co-segregates with mature myeloid dendritic cells and effector-memory CD8⁺ T cells and correlates with improved survival, whereas a CD56^dim signature is not clearly prognostic⁷⁶.

Combined analysis of KIR–human leukocyte antigen genotypes and DNAM-1/CD226 expression on peripheral NK cells defines low-, intermediate- and high-risk “immunogenetic” groups with distinct 10-year progression-free and overall survival, independent of TNM stage. High-risk constellations, such as KIR2DL5 with HLA-C*16 plus low CD226, are associated with poorer outcomes, including in T1 NMIBC receiving BCG, indicating that suboptimal NK licensing and reduced DNAM-1 (CD226) signalling may compromise anti-tumour immunosurveillance. Functionally, inhibitory KIR engagement with self-HLA during NK development “licenses” NK cells, enabling stronger responses against tumour targets that downregulate HLA (“missing-self”). Unfavourable KIR–HLA combinations, particularly when coupled with low CD226/DNAM-1 expression, are expected to yield less responsive NK cells with weaker activating synapse signalling and reduced degranulation/cytokine output within the tumour microenvironment⁷⁵. Inhibitory Siglec-7 is highly expressed on peripheral, tumour-infiltrating and urinary NK cells, and high SIGLEC7/sialyltransferase expression in tumours associates with inferior survival; desialylation of bladder-cancer cells enhances NK degranulation and cytokine release, implicating a functional sialic acid–Siglec-7 checkpoint in urothelial cancer⁴⁰.

7.2 NK cells in NMIBC tissue and predictive value

Histology in low-grade, BCG-naïve NMIBC shows that higher stromal CD56⁺ density in primary Ta/T1 tumours treated by transurethral resection of bladder tumour (TURBT) alone associates with earlier recurrence, especially in single, small (<3 cm) Ta lesions, whereas intraepithelial CD56⁺ cells are not discriminative (Krpina 2014). In solitary low-grade NMIBC without intravesical therapy, smaller tumours (≤3 cm) contain a higher proportion of CD56⁺ cells in the inflammatory infiltrate than larger tumours, while NK density does not differ between Ta and T1; tumour-associated CD68⁺ macrophages are significantly enriched in T1, suggesting a shift towards a macrophage-dominated microenvironment with invasion⁷⁷.

In BCG-treated intermediate- and high-risk NMIBC, higher pre-treatment CD56⁺ cell density in the urothelial compartment associates with increased probability of BCG response and improved recurrence-free survival, and retains significance in multivariable models, supporting baseline tumour-infiltrating NK cells as a favourable predictive marker under BCG²⁴. Peripheral data show that higher baseline NK cytotoxic activity combined with a positive purified protein derivative (PPD) skin test associates with reduced recurrence risk during BCG, suggesting that simple functional NK assays may provide non-invasive prediction of BCG efficacy³⁹. During BCG induction, urinary NK cells consistently express Siglec-7 and Siglec-6, with stable expression across instillations, indicating that glyco-immune checkpoints remain active on trafficking NK cells throughout therapy⁴⁰.

7.3 NK cells, BCG and cytotoxic function

Comparisons of NK cells from healthy donors and patients with high-grade Ta NMIBC show that IL-2, IL-15 activated allogeneic NK cells from donors efficiently kill both bulk bladder-cancer cells and sphere-derived cancer stem-like cells, with cytotoxicity largely dependent on NKG2D and DNAM-1 and partly on Fas–Fas ligand. Conditioned media from these activated NK cells downregulate stemness and drug-resistance genes and increase cisplatin sensitivity in residual stem-like cells, whereas patient-derived NK cells display an immature, functionally skewed phenotype with reduced natural cytotoxicity receptors and poor killing of stem-like cells, consistent with endogenous NK dysfunction in NMIBC⁷⁵.

BCG stimulation of peripheral blood mononuclear cells expands a CD56^bright/high NK population with high perforin/granzyme B content and strong degranulation against bladder-cancer lines. Heat-killed BCG and selected mycobacterial fractions are sufficient to trigger this response, implying pattern-recognition-driven NK activation and potential strain-dependent bias between NK- and T-cell activation³⁸.  A two-step protocol combining one week of BCG stimulation with low-dose IL-12, IL-15 and IL-21, plus a short IL-18 pulse, expands CD56^high NK and Vγ9Vδ2 T-cell populations up to approximately 100-fold while preserving cytotoxicity, providing proof-of-concept for BCG-primed innate effector products⁷⁸.

7.4 Therapeutic modulation of NK cells

Cytokine-based and cellular approaches aim to exploit NK biology in high-risk and BCG-unresponsive NMIBC. Intravesical nogapendekin alfa inbakicept (N-803), an IL-15/IL-15 receptor α–Fc superagonist, combined with BCG achieves complete response rates of about 70% in BCG-unresponsive carcinoma in situ with or without papillary disease, with many responses durable beyond two years and low cystectomy rates, whereas N-803 monotherapy shows limited activity, consistent with IL-15-driven expansion of NK and effector/memory CD8⁺ T cells on a BCG-primed background⁷⁹. Orthotopic models show that intravesical IL-2/IL-15-activated allogeneic NK cells induce major tumour regression, frequent complete remissions and loss of stem-cell markers, supporting intravesical adoptive NK transfer as a potential adjunct in high-risk NMIBC²³. BCG-primed CD56^high NK/Vγ9Vδ2 T-cell products expanded ex vivo with the Esteso protocol reach clinically relevant numbers and represent candidates for off-the-shelf innate cell therapies⁷⁸.

Pharmacological NK-directed immunotherapies include antibodies against inhibitory receptors such as NKG2A, agonists of IL-2/IL-15 pathways and engineered vectors designed to improve NK recognition, trafficking and persistence in the tumour microenvironment⁴¹. The prominence of Siglec-7 on bladder-cancer NK cells and the association of high Siglec7 expression with poor outcome provide additional rationale for targeting glyco-immune and Siglec pathways, potentially in combination with BCG or cytokine-based strategies⁴⁰.

Overall, converging evidence from histology^77,80, pre-treatment immunohistochemistry in BCG-treated NMIBC²⁴, peripheral functional assays³⁹, phenotypic and transcriptomic analyses^{7,32,40,75,76}, mechanistic studies^23,38,78 and clinical/translational work^41,81 supports the view that NK cells are integral to NMIBC immune surveillance and BCG response, with clinical impact driven more by subset composition, licensing context and inhibitory-receptor landscape than by absolute NK-cell numbers. Main findings on NK-cell subsets, biomarkers and therapeutic modulation in NMIBC are summarised in Table 5.

8. Integrated view: innate signatures, spatial context and interaction with adaptive immunity

Data across neutrophils, macrophages, NK cells and dendritic cells support a model in which the NMIBC tumour microenvironment is organised along an innate immune axis that both shapes and is shaped by adaptive responses. Neutrophils and monocyte-derived macrophages are early, abundant populations that respond to tumour-derived and therapy-induced danger signals, while NK cells and DCs integrate these cues and influence whether effector T- and B-cell responses will be sustained or blunted^9–11,15,42. Within this framework, NMIBC lies between a partially controlled, inflamed state and a myeloid-dominated, immunosuppressed muscle-invasive phenotype, with innate cell composition and activation status acting as key determinants of trajectory^12,25.

Functionally, these innate compartments do not act in isolation but form a coupled signalling network that can either support productive priming or reinforce chronic suppression. Neutrophil-dominant inflammation can amplify recruitment and activation of monocyte-derived macrophages and DCs, while macrophage polarisation in turn “tunes” DC maturation and antigen-presenting capacity, shaping the quality of downstream CD8⁺ T-cell and B-cell responses^15,42. Convergent suppressive axes, including IL-10/TGF-β–conditioned myeloid states and COX-2/PGE₂ and IDO-related circuits, may restrain durable effector function^27,30,31,36, whereas a more balanced myeloid context with effective DC maturation and preserved NK activity is more compatible with sustained adaptive immunity and structured TLS^32,76. This bidirectional crosstalk provides a mechanistic bridge between the spatial patterns described below and the observed links to prognosis and BCG outcomes^9–12,25.

Histological and spatial analyses indicate that neutrophils accumulate in papillary fronds, tumour–stroma interfaces and perivascular regions, with tissue NLR rising from low-grade NMIBC to muscle-invasive disease and intratumoural neutrophil density already associating with early intravesical relapse in the NMIBC subset^12,25. TAMs show a complementary distribution, with CD68⁺ and CD163⁺ macrophages concentrated in tumour-associated stroma, the papillary axis and peritumoural lymphoid aggregates, particularly at the invasive front, while intranest infiltration is more variable^13,14,36,68. Mature CD83⁺ DCs are enriched in intraepithelial and stromal niches in high-risk NMIBC but often display incomplete maturation and inhibitory receptor expression, consistent with functional restraint under local IL-10, TGF-β and IDO^27,30,31,36. NK cells are typically scattered within the lamina propria and at tumour margins, and transcriptional signatures of CD56^bright-like NK cells cluster with inflamed, T cell–rich phenotypes, whereas NK-poor tumours align with myeloid-heavy, immunosuppressed profiles^32,76.

Beyond descriptive co-localization, spatial proximity is likely to have direct functional consequences by creating short-range suppressive niches. Close apposition of M2-skewed TAMs with FOXP3⁺ Tregs and exhausted CD8⁺ T cells can intensify local exposure to macrophage- and Treg-derived inhibitory mediators (IL-10 and TGF-β), reinforce checkpoint signalling (PD-1/PD-L1 interactions), and promote metabolically suppressive microdomains (arginine depletion and other myeloid-linked metabolic constraints), collectively reducing effective antigen presentation and dampening effector function⁸². These mechanisms are difficult to infer from bulk measurements alone and motivate spatially resolved approaches. In NMIBC, multiplex IHC/IF and highly multiplexed tissue imaging platforms such as CODEX can quantify cell–cell distances and “cellular neighbourhoods” (TAM–Treg–Tex clusters versus DC-rich niches), enabling hypothesis-driven links between spatial architecture and functional immune restraint or productive priming^61,83.

These spatial relationships translate into distinct innate–adaptive constellations. Macrophage-rich regions often co-localise with FOXP3⁺ Tregs and IL-6 expression, forming an immunosuppressive triad that dampens Th1 polarisation and favours recurrence under BCG pressure⁶³. High CD163/CD68 ratios and co-enrichment of CD163⁺ TAMs, FOXP3⁺ Tregs and diffuse CIS mark “hot but suppressed” microenvironments in which CD8⁺ T cells and B-cell aggregates are present but functionally constrained^13,14,28. By contrast, lesions with moderate myeloid infiltration, preserved CD8⁺ TIL density and organised tertiary lymphoid structures enriched for CD20⁺ B cells, germinal centre markers and proliferating Tfh cells show more favourable recurrence-free survival and appear better able to maintain immune surveillance^7,8,14.

Neutrophils and macrophages interface with these lymphoid structures at several levels. BCG-activated neutrophils produce chemokines such as IL-8, GRO-α and MIP-1α that recruit monocytes and CD4⁺ T cells into the bladder wall, seeding sites where DCs and macrophages can prime or tolerise T cells depending on their activation state^15,42. M2-skewed TAMs produce VEGF, MMPs, CXCL8 and immunosuppressive cytokines, promoting angiogenesis, matrix remodelling and T-cell dysfunction, and align with macrophage-enriched transcriptional clusters associated with high-risk molecular subtypes^20,70. DCs occupy a central position in this network: cross-presenting cDC1 are required for effective CD8⁺ priming and responses to intravesical CD40 agonists, whereas tolerogenic DC phenotypes expand Tregs and reinforce suppression^16,22,29,31. NK cells in turn sense and modulate this landscape; higher baseline NK-cell infiltration and preserved NK activity associate with better BCG responses, while up-regulation of inhibitory receptors such as Siglec-7 limits NK-mediated cytotoxicity and may facilitate immune escape^24,39–41.

Overall, these observations support an “inflamed” versus “cold” spectrum in NMIBC with important nuances. Truly cold tumours show low TIL and B-cell density with sparse DCs and NK cells and may require de novo immune priming. More commonly, high-risk NMIBC lesions display mixed inflamed–suppressed phenotypes, with abundant myeloid cells, Tregs and exhausted TILs coexisting with TLS-like aggregates and residual effector activity. In such settings, innate signatures, neutrophil or monocyte predominance, M2-heavy TAM profiles, cDC1 depletion or dysfunction and impaired NK activity appear to influence whether the balance shifts towards durable tumour control or progression under surgery, BCG and subsequent therapies. To summarise this framework, Figure 1 depicts the functional spectrum of the NMIBC tumour immune microenvironment, contrasting an immune-active state that supports effective surveillance and BCG responsiveness with a myeloid-skewed, immunosuppressive state that is more often linked to adverse outcomes.

Figure 1. Conceptual balance of the innate immune microenvironment in NMIBC.
Schematic overview illustrating two ends of a functional spectrum in the NMIBC tumour immune microenvironment: a permissive, anti-tumour state that supports durable immune surveillance and response to intravesical therapy, versus a myeloid-dominant, immunosuppressive state associated with impaired effector function and unfavourable outcomes.

The left panel summarizes features commonly linked to a more effective immune context, including N1-like neutrophil activity, M1-leaning macrophage programs, mature antigen-presenting DC phenotypes, and functional NK-cell activity, alongside structured TLS that support adaptive priming and maintenance. The right panel depicts an opposing pattern characterized by N2-like neutrophil programs, M2-skewed TAM dominance, incompletely matured or inhibited DC states, and restrained NK-cell function, often accompanied by regulatory T-cell–enriched suppressionand reduced quality of effector responses. The central arrows indicate the net functional shift (“balance”) of the local microenvironment toward either immune control or immune escape under tumour and therapy-driven pressures.

Abbreviations: NMIBC, non–muscle-invasive bladder cancer; DC, dendritic cell; NK, natural killer; TLS, tertiary lymphoid structure; TAM, tumour-associated macrophage. 

9. Clinical translation of innate immune signatures in NMIBC

Innate-derived biomarkers have consistent associations with prognosis in NMIBC and may refine existing clinicopathological risk models. At the systemic level, elevated pre-TURBT and pre-BCG neutrophil-to-lymphocyte ratio and related indices such as the systemic inflammation response index are repeatedly linked to higher stage and grade, larger and multifocal tumours, concomitant carcinoma in situ, higher EORTC risk scores and shorter recurrence-free and progression-free survival, as summarised in the neutrophil section and in Table 2^{18,19,33–35,43,44,46,72}. Tissue-based markers provide complementary information: high tumour-infiltrating neutrophil density and elevated tissue NLR associate with inferior overall and relapse-free survival, while M2-skewed tumour-associated macrophage profiles characterised by high CD68⁺ and CD163⁺ densities, increased CD163/CD68 ratios and macrophage-enriched transcriptional clusters correlate with high-grade and pT1 tumours, diffuse CIS, larger tumour size and shorter recurrence-free survival^{9,12–14,27,28}. Soluble macrophage-derived factors such as circulating MMP12 and integrated effector. Suppressor scores that combine CD8⁺/CX3CR1⁺ T cells with CD163⁺ TAMs and FOXP3⁺ Tregs, as shown in the OncoTherad series, illustrate how myeloid activation can be monitored non-invasively and integrated into composite prognostic tools^20,21. NK- and DC-related signatures, including CD56^bright-like NK profiles, cDC1 abundance and expression of DC inhibitory receptors such as BTLA and TIM-3, further distinguish inflamed from suppressed microenvironments but remain exploratory and require validation in dedicated NMIBC studies^30,32,76. At present, these innate biomarkers appear most useful as adjuncts to standard clinical and pathological risk models rather than standalone decision tools.

Innate profiles also shape response to intravesical BCG and point towards rational combination strategies. Experimental models show that effective BCG immunotherapy depends on early recruitment and activation of neutrophils and Ly6C⁺CD11c⁺ DCs, whereas clinical data associate unfavourable outcomes with high baseline NLR or SIRI, dense M2-biased TAM infiltrates, enrichment of FOXP3⁺ Tregs and IL-6 and myeloid-driven exhaustion of effector T cells, particularly in CIS and high-risk NMIBC under BCG pressure^{15,19,22,35,37,51,63–66}. In contrast, more permissive profiles combine moderate neutrophil and macrophage infiltration with adequate cDC1 numbers and maturation, preserved NK activity and structured tertiary lymphoid aggregates, and are associated with better BCG responses and longer recurrence-free survival in translational and clinical series^{7,8,14,23,24,27,38,39}. These patterns support development of biomarker-driven strategies that (i) use simple systemic indices and selected tissue markers to flag patients at high risk of BCG failure for early intensification or switch to alternative intravesical or systemic therapies and (ii) combine macrophage-targeted agents (CSF1R or CCL2–CCR2 blockade, M2 to M1 reprogramming, nano-immunotherapies such as OncoTherad), DC-centred approaches (αDC1 vaccination, intravesical CD40 agonists, Ty21a, gene-modified DCs) and NK-directed interventions (IL-15-based agonists, N-803, cytokine-activated or adoptive NK products, Siglec-7 blockade) with standard TURBT, BCG and checkpoint inhibition as outlined in the cell-specific sections^{10,11,16,21,29,40,41,69,71–74}. Evidence for these combinations is still largely preclinical or early phase, so they should currently be regarded as hypotheses and priorities for prospective, biomarker-integrated trials rather than routine options in everyday practice.

From a clinical translation perspective, the strength of evidence differs across biomarker classes. Systemic indices such as NLR and SIRI are supported by multiple retrospective series and BCG-focused meta-analyses, but their implementation is limited by non-standardised cut-offs, variable sampling windows (pre-TURBT vs pre-BCG), and confounding by intercurrent inflammation. Tissue macrophage markers (CD68/CD163 and ratios) and spatial myeloid patterns are biologically coherent but remain largely retrospective and are constrained by assay and scoring variability (antibodies, platforms, thresholds, inter-observer reproducibility) and by within-tumour spatial heterogeneity. Higher-dimensional NK/DC signatures and transcriptional clusters are still exploratory, with additional hurdles including platform dependency, batch effects and the need for clinically practical, standardised assays. Across all classes, progress towards clinical use will require prospective, protocolised sampling, pre-specified cut-offs, harmonised pathology pipelines (ideally with digital quantification), and external validation in well-defined BCG-treated NMIBC cohorts.

Recent multicentre prospective/registrational evidence in BCG-unresponsive NMIBC provides a stronger clinical anchor for parts of the therapeutic landscape discussed here. Intravesical nadofaragene firadenovec has been evaluated in a pivotal phase III setting with durable complete responses reported in CIS-containing disease, supporting bladder-preserving management in selected patients⁸⁴. In parallel, nogapendekin alfa inbakicept (N-803) plus BCG received regulatory approval based on a multicentre registrational dataset (QUILT-3.032), highlighting the translational relevance of NK- and T-cell activation pathways in BCG-unresponsive CIS⁸⁵. In BCG-naïve high-risk NMIBC, a randomised phase III trial evaluating durvalumab in combination with BCG versus BCG alone (POTOMAC) has reported outcomes that further contextualise the role of immune-modulating combinations in contemporary practice⁸⁶. Finally, additional phase III programmes are ongoing, including checkpoint-plus-BCG combinations and intravesical immunotherapies, which will be important to define where these strategies can be integrated into routine care⁸⁷.

Although innate-targeted strategies are attractive, several translational pitfalls warrant caution. Systemic depletion or broad reprogramming of myeloid cells can have off-tumour consequences, given the central role of macrophages and related innate compartments in host defence and tissue homeostasis. Clinical and preclinical experience with CSF1R blockade illustrates that macrophage depletion may be accompanied by on-target toxicities such as oedema and broader physiological effects, underscoring the need for careful dosing and monitoring⁸⁸. Likewise, potent innate agonism (eCD40 agonists) can be limited by systemic inflammatory toxicities (including cytokine release and hepatic effects), which is one reason why local delivery approaches have been pursued in bladder cancer¹⁶. For emerging intravesical platforms that combine tumour killing with microenvironmental remodeling (including nanomedicine-based ferroptosis induction and gene-circuit reprogramming), additional challenges include biodistribution beyond the bladder, durability and controllability of immune activation, interpatient heterogeneity, and the risk that strong immune stimulation may not be uniformly tumour-selective^89,90. Overall, these considerations support biomarker-guided patient selection, preference for locoregional delivery when feasible, and prospective safety-focused validation alongside efficacy endpoints.

10. Future directions

Key areas for future research are the validation of innate biomarkers and the testing of innate-targeted interventions. Prospective multi-centre NMIBC cohorts with predefined sampling time points and harmonised assays are needed to evaluate pre-TURBT and pre-BCG NLR, SIRI and related indices alongside EORTC or CUETO scores and molecular classifiers. Tissue-based measures such as TIN/TAN density, CD68⁺ and CD163⁺ TAM burden, M2-skewed spatial patterns and DC/NK signatures should be assessed with standardised panels, digital pathology and, where feasible, spatial transcriptomics.

Innate signatures need to be integrated with genomic and transcriptomic data, linking myeloid-rich or NK/DC-rich clusters to mutational profiles and NMIBC molecular subtypes. Larger single-cell and multiplex imaging studies in treatment-naïve, BCG-responding and BCG-unresponsive disease can clarify how innate cell states change over time and under different intravesical regimens. Interventional trials that combine macrophage-directed agents, intravesical nano-immunotherapies, DC-based approaches and NK-directed strategies with BCG and checkpoint inhibition in well-defined high-risk and BCG-unresponsive NMIBC, with embedded correlative work, are a logical next step.

To move from biomarker association to clinical utility, several pragmatic, biomarker-integrated trial concepts follow directly from the evidence reviewed here. First, a Phase II study in BCG-unresponsive NMIBC could test intravesical CD40 agonism with pre-specified stratification by baseline CD163⁺ TAM burden (or CD163/CD68 ratio) and a cDC1 signature, coupled with paired pre-/on-treatment tissue for digital pathology and spatial immune profiling; this design is supported by strong preclinical rationale for intravesical Fc-enhanced CD40 agonists and by an ongoing first-in-human intravesical program^16,91. Second, given the established activity of intravesical IL-15 agonism (nogapendekin alfa inbakicept, N-803) plus BCG in BCG-unresponsive CIS with/without papillary disease, a biomarker-driven Phase II/III strategy could prospectively enrich or stratify patients by baseline NK functional/transcriptional signatures to determine whether “NK-low” tumours derive the largest incremental benefit and to link on-treatment immune restoration with durable response⁸¹. Third, for high-risk NMIBC at increased risk of BCG failure, a randomised Phase II trial could evaluate a macrophage-modulating add-on to standard BCG (intravesical nano-immunotherapy platforms with TAM-reprogramming readouts), restricting enrolment or stratifying by M2-skewed macrophage profiles, with recurrence-free survival and predefined immune reprogramming endpoints⁹². 

11. Conclusion

Innate immune cells are central components of the NMIBC microenvironment. Neutrophils and TAMs consistently associate with higher stage and grade, adverse histology, increased recurrence and progression and BCG failure, whereas sufficient and functionally competent NK cells and DCs appear to support effective antitumour responses. Peripheral indices such as pre-operative and pre-BCG NLR, tissue-based measures of TIN/TAN density, TAM burden and phenotype, macrophage-derived soluble factors and composite effector:suppressor scores offer biologically plausible, low-cost tools to refine risk assessment. At present, these biomarkers are best viewed as adjuncts to established clinical and pathological risk models rather than replacements. Further prospective and interventional studies are required to define robust cut-offs, clarify stage-specific roles and determine whether targeting neutrophil, macrophage, NK and DC networks can improve outcomes and support bladder-sparing strategies in high-risk NMIBC.