Tumour-infiltrating B cells: immunological mechanisms, clinical impact and therapeutic opportunities

Abstract

Although immunotherapy research to date has focused largely on T cells, there is mounting evidence that tumour-infiltrating B cells and plasma cells (collectively referred to as tumour-infiltrating B lymphocytes (TIL-Bs)) have a crucial, synergistic role in tumour control. In many cancers, TIL-Bs have demonstrated strong predictive and prognostic significance in the context of both standard treatments and immune checkpoint blockade, offering the prospect of new therapeutic opportunities that leverage their unique immunological properties. Drawing insights from autoimmunity, we review the molecular phenotypes, architectural contexts, antigen specificities, effector mechanisms and regulatory pathways relevant to TIL-Bs in human cancer. Although the field is young, the emerging picture is that TIL-Bs promote antitumour immunity through their unique mode of antigen presentation to T cells; their role in assembling and perpetuating immunologically ‘hot’ tumour microenvironments involving T cells, myeloid cells and natural killer cells; and their potential to combat immune editing and tumour heterogeneity through the easing of self-tolerance mechanisms. We end by discussing the most promising approaches to enhance TIL-B responses in concert with other immune cell subsets to extend the reach, potency and durability of cancer immunotherapy.

Although the critical role of T cells in antitumour immunity has become indisputable in the era of immune checkpoint blockade and adoptive T cell therapy, these same advances have revealed the many limitations and vulnerabilities of the T cell response, prompting an urgent need to understand and harness orthogonal immunological mechanisms. A flood of new evidence implicates tumour-infiltrating B cells and plasma cells (PCs) (collectively referred to as tumour-infiltrating B lymphocytes (TIL-Bs)) as powerful, multifaceted players in antitumour responses. TIL-Bs are rarely found on their own but, rather, associate intimately with T cells, myeloid cells and other immune cells in the most immunologically ‘hot’ tumours, often substantially exceeding the levels of B-lineage cells in healthy non-lymphoid tissues (FIG. 1; see Supplementary Table 1). Indeed, the fact that exhausted or dysfunctional CD8⁺ and CD4⁺ TILs frequently express the B cell-recruiting C–X–C motif chemokine ligand 13 (CXCL13) suggests they are programmed to solicit B cell help in the face of tumour persistence^1–3. Such interactions can culminate in the formation of tertiary lymphoid structures (TLSs), lymph node-like structures that arise de novo in tumour stroma and appear actively engaged in priming and sustaining adaptive immune responses. Similar to T cells, TIL-Bs are associated with a positive prognostic value in most cancers (FIG. 1; see Supplementary Table 2). Moreover, they can strongly enhance the prognostic impact of CD4⁺ and CD8⁺ TILs, an effect that appears particularly robust in tumours harbouring TLSs⁴. Also similar to T cells, TIL-Bs comprise various phenotypes, including both effector and regulatory B cell (B_reg cell) subsets.

Fig. 1 |. B cells in health and malignancy.

a | Heat maps comparing abundance of B cells and plasma cells (PCs) in normal and tumour tissues. Data from Genotype-Tissue Expression consortium (GTEx v. 8, n = 16,704 samples) and The Cancer Genome Atlas (TCGA) consortium (harmonized dataset, n = 9,922 primary solid tumours) were analysed using xCell¹⁹⁹ to generate cell type enrichment scores for naive B cells (B naive), memory B cells (B memory) and PCs. Scores averaged across samples of the same origin, scaled across columns and ranked by decreasing order. Asterisk indicates an identical normal tissue comparator was not available, so a closely related tissue was used instead. Log-fold change for B cell (average of naive B cells and memory B cells) and PC scores between tumours and their normal counterparts are reported (right). Red, tumour tissue > normal tissue; no colour, tumour tissue ~ normal tissue; blue, tumour tissue < normal tissue. b | Heat map summarizing prognostic associations for intratumoural B cells, PCs, IgG and IgA based on an update of previous literature searches^93,200. Tumour sites are deemed positive, neutral, or negative based on the majority of cohorts. Supplementary Tables 1 and 2 provide supporting data and references. ACC, adrenocortical carcinoma; BLCA, bladder urothelial carcinoma; BRCA, breast invasive carcinoma; CESC, cervical squamous cell carcinoma and endocervical adenocarcinoma; CHOL, cholangiocarcinoma; COAD, colon adenocarcinoma; ESCA, oesophageal carcinoma; GBM, glioblastoma multiforme; HNSC, head and neck squamous cell carcinoma; KICH, kidney chromophobe; KIRC, kidney renal clear cell carcinoma; KIRP, kidney renal papillary cell carcinoma; LGG, brain lower grade glioma; LIHC, liver hepatocellular carcinoma; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; MESO, mesothelioma; OV, ovarian serous cystadeno-carcinoma; PAAD, pancreatic adenocarcinoma; PCPG, pheochromocytoma and paraganglioma; PRAD, prostate adenocarcinoma; READ, rectum adenocarcinoma; SARC, sarcoma; SKCM, skin cutaneous melanoma and squamous cell carcinoma; STAD, stomach adenocarcinoma; TGCT, testicular germ cell tumours; THCA, thyroid carcinoma; THYM, thymoma; UCEC, uterine corpus endometrial carcinoma; UCS, uterine carcinosarcoma; UVM, uveal melanoma.

Exhausted or dysfunctional CD8⁺ and CD4⁺ TILs.

(Exhausted or dysfunctional CD8⁺ and CD4⁺ tumour-infiltrating lymphocytes). Adysfunctional T cell state triggered by chronic antigen exposure and characterized by poor effector functions (for example, cytokine production), sustained expression of inhibitory receptors (for example, PD-1, cytotoxic T lymphocyte-associated protein 4 (CTLA4)) and widespread transcriptomic changes. Exhausted T cells can be placed on a dysfunction continuum (pre-dysfunctional, early dysfunctional and late dysfunctional) based on their clonality (intermediate to high) and their expression of C–X–C motif chemokine ligand 13 (CXCL13) and inhibitory receptors (intermediate to high).

TIL-Bs exhibit the hallmarks of antigen recognition, including somatic hypermutation, class switch recombination, clonal expansion and differentiation into IgG or IgA-producing PCs. Whereas tumour-reactive CD4⁺ and CD8⁺ TILs often recognize mutated proteins (neoantigens)⁵, the target antigens of TIL-Bs identified to date are predominantly self-proteins⁶, suggesting that tumours can disrupt B cell tolerance mechanisms. Indeed, the orthogonal tolerance mechanisms used by B cells and T cells raise fascinating immunological issues and compelling therapeutic opportunities in the setting of cancer. Adding to this theme of complementarity, TIL-Bs have a plethora of effector mechanisms that enable them to enhance not only adaptive immune responses but also innate mechanisms involving complement, myeloid cells and natural killer cells, as well as the myriad possibilities afforded by antibodies.

Although B cells and PCs remain poorly understood in the cancer immunotherapy field, they have received intensive study in autoimmunity, where they use multi-pronged effector mechanisms to mediate potent, persistent and often unstoppable destruction of target tissues, sometimes with remarkable selectivity. Drawing insights from the autoimmunity field, this Review summarizes our current understanding of TIL-Bs in human cancers and highlights the knowledge gaps that require urgent attention. After considering the role of TIL-Bs across the major facets of cancer immunity, we discuss strategies to harness their cell-based and antibody-based effector mechanisms to enable a new generation of cancer immunotherapies.

B cell development and tolerance

B cells arise and mature in bone marrow, where they undergo V(D)J recombination, resulting in cell surface expression of B cell receptors (BCRs) of the IgM isotype (FIG. 2a). Up to 75% of immature B cells are self-reactive⁷ and undergo either clonal deletion or receptor editing, wherein a second immunoglobulin light chain is generated, providing a second chance to reduce self-reactivity (FIG. 2b). Nonetheless, a considerable proportion of self-reactive B cells enter the periphery and comprise about 20% of mature naive circulating B cells in healthy individuals⁸. This reservoir of autoreactive B cells is thought to be critical to avoid large ‘gaps’ in the B cell repertoire that could otherwise serve as havens for pathogens expressing antigens that resemble host proteins⁹. Indeed, self-reactive B cells can participate in responses to infectious agents through the process of clonal redemption^10,11 (FIG. 2b). Peripheral B cell tolerance is maintained by several mechanisms, chief ones being the suppressive effects of CD4⁺ regulatory T (T_reg) cells counterbalanced with the stimulatory effects of CD4⁺ T follicular helper cells (T_FH cells) and other T cell subsets. Indeed, either a decrease in T_reg cells or an increase in T_FH cells is sufficient to generate autoreactive B cells^8,12. Thus, whereas both B cells and T cells are subject to central and peripheral tolerance, the precise mechanisms and underlying antigenic repertoires are different, resulting in complementary, interdependent definitions of self that carry important implications for antitumour immunity. As discussed later, there are numerous examples of TIL-Bs recognizing self-antigens (TABLE 1; see Supplementary Table 3), suggesting that many TIL-B responses do indeed involve disruption of peripheral tolerance mechanisms (FIG. 2b) as seen in other inflammatory contexts such as acute and chronic infection^13,14.

Fig. 2 |. B cell differentiation and tolerance.

a | Stages of B cell activation, clonal expansion, somatic hypermutation, affinity selection, class switch recombination and differentiation to plasma cells (PCs). These events can involve germinal centre reactions or extrafollicular responses, with the former generally being the better source of long-lived PCs. All of these B cell subsets are found in human tumours, with the most commonly reported subsets being activated B cells, memory B cells and PCs. b | Central tolerance takes place in bone marrow and eliminates immature self-reactive B cells via clonal deletion; receptor editing of immunoglobulin light chains (L chains) offers one chance at rescue. Peripheral tolerance maintains weakly and moderately self-reactive B cells in a hypo-responsive or inactive state through several mechanisms. B cell responses to foreign antigens (or potentially neoantigens in cancer) can occur through conventional activation or clonal redemption. Disruption of peripheral tolerance during inflammation, autoimmunity or cancer can license self-reactive B cells and PCs to participate in immune responses. With the exception of human papillomavirus (HPV) antigens and tumour protein p53 (p53) neoantigens (although most p53-specific autoantibodies appear to recognize non-mutated epitopes), studies to date indicate that TIL-Bs generally recognize self-antigens with varying degrees of tumour-specific expression. BCR, B cell receptor; H chain, heavy chain; NA, not applicable; TIL-B, tumour-infiltrating B lymphocyte; T_reg cell, regulatory T cell.

Table 1 |.

Antigens recognized by TIL-Bs

Antigen name	Antigen location (known or predicted)	Cancer of origin	Isotype	Refs
Cancer testis and stem cell antigens
CTAG1B	Intracellular	Lung (NSCLC), breast	IgG > IgA (lung), IgG < IgA (breast)	105,194
CTAG2	Nucleoplasm, nuclear bodies, vesicles	Lung (NSCLC)	IgG > IgA	194
MAGEA3	Intracellular	Lung (NSCLC)	IgG	194
NXF2	Intracellular	Lung (NSCLC)	IgG	194
SPAG8	Intracellular	Breast	IgA, IgG	105
SOX2	Nucleoplasm	Lung (NSCLC)	IgG	194
Viral antigens
E2	Nucleus	Head and neck (HPV + HNSCC)	IgG	53
E6	Nucleus	Head and neck (HPV + HNSCC)	IgG	53
E7	Nucleus	Head and neck (HPV + HNSCC)	IgG	53
Lineage-restricted antigens
BDNF	Secreted	Ovary (HGSC)	IgA	138
MLANA	Membrane	Lung (NSCLC)	IgG	194
Commonly mutated antigens
BRCA2	Nucleoplasm, cytosol	Breast	IgA, IgG	105
TP53	Nucleoplasm	Lung (adenocarcinoma, LCC, NSCLC, SCC)	IgG	194–196
Antigens commonly dysregulated in cancer
ACAT2	Nucleoplasm, cytosol	Lung (LCC)	IgG	195
COPS4	Nuclear speckles	Breast	IgA, IgG	105
GAPDH	Nuclear membrane, vesicles, plasma membrane, cytosol	Lung (LCC)	IgG	195
MUC1	Plasma membrane	Breast	NE	197
RHOC	Intracellular	Lung (NSCLC)	IgG	48
TSPAN7	Plasma membrane	Ovary (HGSC)	IgA	138
Post-translationally modified antigens
ACTB	Cytosol, apoptotic blebs	Breast (MBC)	IgG	88
Hspa4	Nucleoplasm, cytosol, plasma membrane	Breast (4T1)	IgG	149
Self-antigens with no reported aberrancies
GAS8	Plasma membrane, cytosol	Lung (LCC)	IgG	195
MAPK9	Intracellular	Lung (LCC)	IgG	195
PHF3	Nucleoplasm	Lung (LCC)	IgG	195
RPL3	Nucleoli, cytosol	Lung (NSCLC)	IgG	48
Lipid antigens
Ganglioside D3	Plasma membrane	Breast (MBC)	NE	198
Nucleic acid antigens
Mitochondrial DNA	Mitochondria	Lung (LCC)	IgG	195

Known antigens locations are reported in roman font, predicted locations in italics. Data were extracted from the Human Protein Atlas. Cancer subtype indicated in parentheses where available. 4T1, murine BALB/c mammary gland tissue-derived breast cancer cell line; BDNF, brain-derived neurotrophic factor; BRCA2, breast cancer type 2 susceptibility protein; HGSC, high-grade serous ovarian cancer; HNSCC, head and neck squamous cell carcinoma; HPV, human papillomavirus; LCC, large cell carcinoma; MBC, medullary breast carcinoma; MUC1, mucin 1; NE, not evaluated; NSCLC, non-small-cell lung carcinoma; SCC, squamous cell carcinoma; TIL-B, tumour-infiltrating B lymphocyte.

Clonal redemption.

A process by which moderately self-reactive anergic B cells are recruited to the germinal centre, where somatic hypermutation gives them the opportunity to mutate away from self-reactivity.

The emerging range of TIL-B phenotypes

The advent of high-dimensional flow cytometry and single-cell RNA sequencing (scRNA-seq) has advanced our understanding of B cell maturation from nine canonical stages (FIG. 2a) and seven core markers¹⁵ to a plethora of molecular phenotypes that integrate developmental, functional and clonal information. Although the interpretation of such data is evolving, there is evidence for at least 10 B cell subsets in healthy human blood¹⁶, 12–13 in lymphoid tissues^17,18 and 14 in peripheral organs¹⁹. Tissue-resident memory B cells have been reported in skin, gut and lung epithelium^20,21 and, similar to tissue-resident memory T cells²², are characterized by expression of CD45RB and CD69 (REFS^21,23) and provide a first line of defence against reinfection.

In the autoimmunity field, there is strong interest in so-called double-negative B cells that lack expression of IgD and CD27. In particular, the double-negative 2 (DN2) subset of double-negative B cells, which exhibit a T-box transcription factor 21 (TBX21, or T-bet)⁺CD11c⁺ C–X–C motif chemokine receptor 5 (CXCR5)⁻CD21⁻ phenotype, produce autoreactive antibodies, and correlate with more aggressive forms of systemic lupus erythematous and other conditions^24,25. Also of keen interest in autoimmunity are memory PCs that reside in inflamed tissues with support from cytokines such as IL-6, IL-12, tumour necrosis factor (TNF) and type 1 interferon and can induce chronic immunopathology in mouse models²⁶. Finally, B_reg cells have been described to have a role in autoimmunity, transplantation and infection. B_reg cells are more difficult to define than T_reg cells, as they lack a lineage marker analogous to the T_reg cell marker forkhead box protein P3 (FOXP3) and they can arise from any B cell developmental stage²⁷ under the influence of BCR engagement²⁸, Toll-like receptor (TLR)–CD40 co-stimulation^29,30 or cytokine exposure^31–36. B_reg cells are mainly defined by their effector molecules, chief among these being IL-10 as well as IL-35, tumour growth factor-β (TGFβ) and granzyme B (GZMB)^36,37. B_reg cells exert inhibitory effects against CD4⁺ and CD8⁺ T cells, dendritic cells and monocytes, while supporting T_reg cells and invariant natural killer T cells³⁶.

CD40.

A molecule central to B cell activation; upon binding to its ligand expressed by activated T cells, CD40 promotes B cell survival and proliferation, germinal centre formation and development of memory B cells and long-lived plasma cells (PCs).

Early studies in human cancer revealed that TIL-B phenotypes include naive, activated and memory B cells, germinal centre B cells, and PCs and their intermediates³⁸. Some studies have described atypical CD27⁻ memory B cells among TIL-Bs^39–42, which could correspond to the double-negative B cells described above. Moreover, an exhausted TIL-B phenotype (CD69⁺CD27⁻CD21⁻) has been described in association with T_reg cells in patients with lung cancer⁴³. Recent scRNA-seq studies of tumour tissues from patients report anywhere from 2 to 13 TIL-B phenotypes spanning the entire continuum from naive B cells to PCs^42,44–57. In general, the TIL-B phenotypes described so far align with canonical B cell subsets identified in healthy donors (FIG. 2a), although TIL-Bs appear to exhibit greater diversity within some of these subsets^44,47. By scRNA-seq, the most heterogeneous TIL-B subsets are memory B cells, which generally cluster according to their class switch and CD27 status⁴⁴, and PCs, which cluster according to immunoglobulin isotype^46,51. Germinal centre B cells (presumably derived from TLSs) may form clusters corresponding to light, transitional and dark zone phenotypes⁴⁷.

B_reg cells have been described in human tumours and, similar to what has been observed in autoimmunity, can arise from memory B cells⁵⁸, plasmablasts⁵⁹ and plasmocytes^60,61. Notably, B_reg cells have been reported in only two scRNA-seq studies so far^44,54. This could, in part, reflect a bioinformatic issue, as IL10, GZMB and PDCD1-expressing TIL-Bs were identified in breast cancer but did not form a distinct phenotypic cluster⁴⁴. It is essential to define more robust markers of B_reg cells in human cancer so that their properties and therapeutic vulnerabilities can be deciphered.

Various factors can influence the phenotype of TIL-Bs, including disease stage and tumour type (FIG. 1). For example, a transition from naive to PC-like B cells was observed in more advanced lung cancers⁴⁸, which underscores the need to account for stage as a confounding factor in prognostic studies. TIL-B phenotypes can differ between spontaneous and transplanted models of the same cancer⁶², an important consideration for murine studies. The antigenic landscape is another important factor. For example, TLSs and germinal centre B cells are enriched in HPV⁺ versus HPV⁻ head and neck cancers^47,55, and an increased tumour mutation burden enhanced both B cell and PC infiltrates in mouse breast cancer models⁶³. The importance of cell–cell interactions was illustrated by a study using a cancer mouse model in which tumour-associated neutrophils induced the recruitment and differentiation of B cells to PCs via cytokines such as TNF and TNF ligand superfamily member 13B (TNFSF13B, or BAFF)⁶⁴. Finally, B_reg cell differentiation has been linked to interaction with semi-mature dendritic cells in the tumour microenvironment (TME) of hepatocellular carcinoma⁶⁵, TLR2 stimulation by tumour-derived autophagosomes in breast, lung and ovarian adenocarcinomas⁶⁶, tumour-derived leukotriene B4 in breast cancer⁶⁷ and IL-21 in the case of GZMB⁺ B_reg cells in breast, cervical, ovarian and colorectal carcinomas⁶⁸.

The increasing understanding of TIL-B phenotypes provided by scRNA-seq is beginning to translate into prognostic studies that go beyond simple markers such as CD20^{42,44,47,48,51,52,54}. For example, a scRNA-seq study of tumour tissue from patients with nasopharyngeal carcinoma led to the identification of double-negative B cells and PCs that carried negative and positive associations with survival, respectively⁴². In contrast, a colorectal cancer scRNA-seq study identified an IgA⁺IGLC2⁺ PC subset that correlated with poor prognosis in The Cancer Genome Atlas (TCGA) cohort⁵¹ and may correspond to a previously described immunosuppressive IgA⁺ cell subset^60,61. Additional studies involving other tumour types are required to elucidate the contributions of newly discovered TIL-B phenotypes to antitumour immunity and clinical outcomes.

TIL-B neighbourhoods

Similar to tumour-infiltrating T cells, TIL-Bs are found in at least three different structural zones of the TME: highly organized TLSs, which resemble lymph nodes; less organized stromal infiltrates involving TIL-Bs, T cells and macrophages, which we refer to here as lympho-myeloid aggregates (LMAs); and intra-epithelial infiltrates, where TIL-Bs, T cells and macrophages come into direct contact with tumour cells (FIG. 3). We will discuss what is known about TIL-Bs in each of these compartments, emphasizing the major gaps and controversies in our understanding.

Fig. 3 |. TIL-B neighbourhoods.

a–d | Multiplex immunofluorescence staining of untreated human high-grade serous ovarian cancer (parts a–c) and medullary breast cancer tissues (part d) showing examples of a secondary follicular tertiary lymphoid structure (part a), astromal lympho-myeloid aggregate (LMA) with infiltration of a neighbouring epithelial region by CD4 and CD8 T cells and B cells (part b), a dense infiltrate of B cells restricted to the stromal compartment despite infiltration of adjacent epithelium by T cells and macrophages (part c) and a dense stromal infiltrate of plasma cells (PCs) accompanied by intra-epithelial B cells (part d). Left column: low-magnification views with relevant structural zones labelled. Right column: high-magnification views of boxed regions from left column showing examples of CD8 T cell (CD8⁺CD3⁺), presumptive CD4 T cell (CD8⁻CD3⁺), dendritic cell (CD201⁺), follicular dendritic cell (FDC; CD21⁺), high endothelial venule (HEV; PNAd⁺), B cell (CD20⁺), PC (CD20⁻CD79A⁺), macrophage (CD68⁺) and regulatory T cells (T_reg cells; CD3⁺FOXP3⁺). Markers and associated colours specified below each image. Epithelium, tumour epithelium; FOXP3, forkhead box protein P3; PanCK, pan cytokeratin; PNAd, peripheral node addressin; stroma, tumour stroma; TIL-B, tumour-infiltrating B lymphocyte. Image courtesy of Katy Milne, Tashi Rastogi and Karanvir Singh, BC Cancer.

Tertiary lymphoid structures

TLSs are lymph node-like structures that arise de novo in the stroma of hot tumours in response to antigens and inflammatory stimuli^4,69. Although the term TLS is used by some to describe any lymphoid aggregate, we recommend a stricter definition that, at a minimum, includes clearly demarcated B cell follicles containing CD21⁺ follicular dendritic cells (FDCs) and adjacent T cell zones containing conventional dendritic cells and high endothelial venules (HEVs) (FIG. 3a). In some TLSs, B cell follicles contain germinal centres defined by features (and markers) such as activated FDCs (CD21, CD23), B cells undergoing proliferation and somatic hypermutation (Ki67, B cell lymphoma 6 protein (BCL6), activation-induced cytidine deaminase (AICDA)), T_FH cells (CD4, BCL6, inducible T cell co-stimulatory (ICOS), PD-1), plasmablasts and PCs⁴. In a recent breast cancer study, the presence of germinal centres was associated with a favourable balance of T_FH cells to follicular T_reg cells in TLSs⁷⁰. Analogous to lymph nodes, TLSs without versus with germinal centres have been referred to as primary follicular TLSs and secondary follicular TLSs, respectively⁷¹.

Despite many similarities, TLSs show structural differences from lymph nodes that may have significant implications for TIL-B responses. TLSs lack a capsule, which likely results in greater exposure to immunoregulatory factors and apoptotic and/or necrotic debris from the TME⁷². The non-capsular structure and proximal location of TLSs may facilitate the uptake and presentation of antigens that are at low abundance or highly context-dependent (for example, post-translationally modified antigens). These features may also afford better access to tissue-specific self-antigens by B cells and other TLS-associated antigen-presenting cells (APCs), thereby facilitating the maintenance of self-tolerance, as discussed below. Similarly, TLSs lack subcapsular sinus macrophages, which play an integral role in antigen presentation to B cells in lymph nodes. It is unclear whether tumour-associated TLSs contain the well-structured conduits that transport soluble antigens in lymph nodes⁷³ or the stromal cell networks that support B cell and T cell interactions, including CXCL12-producing reticular cells in germinal centres and fibroblastic reticular cells in T cell zones⁷⁴. Indeed, tumour-associated germinal centres typically lack well-defined dark and light zones, which in normal lymph nodes are integral to the process of BCR affinity maturation. Thus, it seems reasonable to speculate that, compared with lymph nodes, TLSs may generate TIL-B responses with distinct antigen specificities, as discussed next.

In addition to their well-known capacity to generate high-affinity antibodies against foreign antigens, germinal centres play an important role in mitigating the dual risks that somatic hypermutation and affinity maturation pose to self-tolerance: new BCR mutations can inadvertently confer autoreactivity, and B cells that are autoreactive at baseline can become activated and undergo somatic hypermutation during the process of clonal redemption^10,11.To maintain tolerance, lymph nodes and germinal centres in healthy tissues have elaborate mechanisms to detect, inhibit, edit and delete autoreactive B cells (FIG. 2b). Critical to reducing the presence and/or activity of autoreactive B cells are tightly regulated interactions between B cells, T_FH cells and follicular T_reg cells, with disruptions in any one being sufficient to promote autoantibody production⁷⁵. Of particular relevance to patients with cancer, the PD-1 and cytotoxic T lymphocyte-associated protein 4 (CTLA4) pathways are also integral to this process⁷⁵. We know very little about how effectively these sophisticated functions are performed by tumour-associated TLSs, where the difference between self and non-self can be as subtle as a single amino acid change, a post-translational modification or, simply, overexpression. On the positive side, TLSs could be a place where self-tolerance mechanisms are sufficiently relaxed to allow the production of autoreactive antibodies that counteract the effects of immune editing of neoantigens and other tumour-specific antigens (FIG. 4). On the negative side, TLSs could give rise to autoreactive B cells and PCs that underlie the autoimmune-mediated toxicities associated with immune checkpoint blockade and, in rare instances, paraneoplastic syndromes⁷⁶. Indeed, an association between TLSs and paraneoplastic syndromes has been reported in ovarian cancer⁷⁷. These risks need to be considered when developing immunotherapies aimed at enhancing TLSs.

Fig. 4 |. Model for intramolecular epitope spreading from neo-epitopes to self-epitopes in cancer.

Top panel: a–h | Antigen released by tumour cells (part a) is taken up by professional antigen presenting cells (APCs) (part b) which then prime neo-epitope-specific CD4 T follicular helper (T_FH) cells and/or T peripheral helper (T_PH) cells as well as CD8 T cells. T_FH cells and/or T_PH cells prime a B cell response against a self-epitope on the same antigen (part c), triggering clonal expansion and plasma cell (PC) differentiation (part d). PC-derived autoantibodies form immune complexes with antigen, which facilitates antigen uptake and presentation by APCs (part e). Additional self-reactive CD4 and CD8 T cells undergo priming, expansion and differentiation (part f). T_FH cells and/or T_PH cells prime and amplify self-reactive B cells (part g), leading to further epitope spreading (part h). Bottom panel: neo-epitope-specific CD8 T cell (cell 1) eliminates neoantigen-expressing tumour cells. Without epitope spreading, this can result in outgrowth of variant tumour cells that no longer express the neoantigen. With epitope spreading, CD4 and CD8 T cells (cell 2, 3 and 4) can additionally target self-epitopes which may drive responses less susceptible to immune editing. Moreover, orthogonal antibody-mediated effector mechanisms can be engaged, including complement-dependent cytotoxicity (CDC), antibody-dependent cellular cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP), resulting in better tumour control. BCR, B cell receptor; MHC, major histocompatibility complex; TCR, T cell receptor.

Somatic hypermutation.

A process by which antigen-activated B cells acquire somatic mutations in their B cell receptor (BCR) sequence, thereby leading to receptor diversification.

Affinity maturation.

A process by which the diverse repertoire of B cell receptors (BCRs) generated by somatic hypermutation is screened to select clones with the highest affinity for their cognate antigen.

Lympho-myeloid aggregates

As mentioned, we propose the generic term ‘lympho-myeloid aggregates’ to describe the various non-follicular aggregates of TIL-Bs, T cells and myeloid cells found in tumours, anticipating that future studies will reveal LMA subtypes with distinct functions (FIG. 3b,c). LMAs are invariably found in TLS-positive tumours, yet they have an even broader distribution that includes virtually all immunologically hot tumours⁷⁸. The following are examples of known or anticipated LMA subclasses that warrant consideration.

Early, immature, stalled or contracting TLSs.

Undoubtedly, some of the LMAs found in tumours correspond to the precursors of bona fide TLSs. Furthermore, as TLSs are transient, antigen-dependent structures⁷³, they presumably undergo a contraction phase if their cognate antigen(s) disappears from the TME. Although mouse models involving experimentally induced TLSs have provided useful insights into these stages⁷³, the field is in need of spontaneous, orthotopic tumour models that engender natural TLS responses, as seen in human cancers.

Extrafollicular response zones.

B cell activation, differentiation and affinity maturation can also occur in so-called extrafollicular responses that take place outside germinal centres in secondary lymphoid organs or inflamed peripheral tissues^72,79. In place of T_FH cells, extrafollicular responses in inflamed tissues involve a CD4⁺ T cell subset known as T peripheral helper cells (T_PH)⁸⁰. Both T_FH cells and T_PH cells are PD-1^high and express CXCL13 and IL-21; however, T_PH cells are generally negative for BCL6 and CXCR5 (REF.⁸¹). Notably, these definitions are fluid, as our understanding of non-canonical T_FH-like cells continues to grow⁸². In autoimmunity, extrafollicular responses have been shown to involve activated B cells, plasmablasts and double-negative B cells alongside T_PH cells^12,83,84. Although extrafollicular responses have not been formally described in cancer, TIL-Bs were found to associate with CXCL13-producing CD4⁺ T cells with a PD-1^highT-bet^highBCL6^lowCXCR5⁻ phenotype in putative early TLSs in breast cancer⁸⁵, and similar findings were recently reported in nasopharyngeal carcinoma⁸⁶.

Secondary lymphoid organs.

Peripheral organs, such as the spleen and lymph nodes, that provide the optimal environment to generate and quench adaptive immune responses.

Antitumour effector zones.

Although antibody-dependent cellular cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP) are widely viewed as important effector mechanisms for TIL-Bs, the architectural setting for these activities remains poorly defined in cancer. In human ovarian cancer, large stromal LMAs have been described in which TIL-Bs, T cells and macrophages congregate and intermix. These are sites of intense PD-L1 expression by macrophages and neighbouring tumour cells, suggesting high T cell activity and associated interferon-γ (IFNγ) production. Given the strong, favourable prognostic significance of these PD-L1⁺ zones⁸⁷, they are compelling candidates for sites of ADCC and/or ADCP fuelled by TIL-B-derived antibodies.

Antibody-dependent cellular cytotoxicity.

(ADCC). A process by which Fc receptor-expressing effector cells, such as natural killer cells and macrophages, recognize antibody-coated cells and orchestrate their destruction through the release of cytotoxic granules.

Antibody-dependent cellular phagocytosis.

(ADCP). A process by which Fc receptor-expressing phagocytes, typically macrophages, recognize antibody-coated cells and destroy them by engulfment.

Plasma cell-rich zones.

In very hot tumours, PCs can dominate the stromal compartment, replacing cancer-associated fibroblasts and other stromal components (FIG. 3d). This is well illustrated in patients with medullary breast cancer, where dense stromal PCs have been shown to produce tumour-reactive antibodies and are associated with an exceptionally favourable prognosis⁸⁸.

Intra-epithelial B cells

Finally, B cells expressing markers such as CD19 and CD20 can be found directly infiltrating tumour epithelium alongside T cells and macrophages (FIG. 3b), a pattern that is prognostically favourable in most cancers (FIG. 1; see Supplementary Table 2). Notably, intra-epithelial B cells are found in only a subset of hot tumours^39,89, suggesting that they have unique entry cues or other requirements relative to T cells. The phenotype, clonality, antigen specificity and function of intra-epithelial B cells is unknown but may soon be revealed by methods such as spatial transcriptomics.

Initiation and progression of TIL-B neighbourhoods

From the foregoing, one can construct a working model of how different TIL-B neighbourhoods may arise and interrelate. The presence of T cells and myeloid cells in the TME appears to be a precondition for any form of TIL-Bs, as the latter are rarely seen without the former. Indeed, dysfunctional or exhausted T cells are known to produce CXCL13 (REFS^1–3), which along with other factors may initiate the recruitment of activated B cells and PCs primed in secondary lymphoid organs. This would result in simple LMAs that may evolve to become sites of extrafollicular responses, further amplifying TIL-Bs and T cell responses. Progression to primary follicular TLS formation likely requires permissive stroma and a favourable cytokine milieu, whereas secondary follicular TLS formation may additionally require factors such as high antigen load (including neoantigens, cancer testis antigens and viral antigens) and a conducive somatic mutation profile^69,90. Productive secondary follicular TLSs could ‘supercharge’ the TIL-B response by generating long-lived, inflammatory PCs that colonize tumour stroma and release affinity matured antibodies against tumour-associated, tissue-specific and/or self-antigens. A robust, local supply of antibody, in turn, would promote ADCC, ADCP and other effector mechanisms (discussed later), resulting in LMAs corresponding to antitumour effector zones. In line with this model, a recent study provides evidence that TLSs do promote in situ B cell maturation in the TME⁹¹. Applying spatial transcriptomics and immunohistochemistry to clear cell renal cell carcinoma, it was shown that, compared with TLS⁻ tumours, TLS⁺ tumours present with a wider diversity of B cell phenotypes, increased somatic hypermutation and clonal expansion, pronounced infiltration of tumour stroma by plasma cells, and enhanced coating of tumour cells by IgG. BOX 1 discusses the prognostic facets of this model.

Box 1 |. Which lymphoid structures are the strongest drivers of prognosis?

Tertiary lymphoid structures (TLSs) have received considerable attention lately due to their generally favourable association with patient survival and response to both conventional and immune-based therapies⁴. However, TLSs, lympho-myeloid aggregates (LMAs) and intra-epithelial B cells are often found together, underscoring the importance of discerning their independent effects. In a pancreatic cancer study, B cells were only prognostically favourable when organized as TLSs²⁰¹. Moreover, secondary follicular TLSs were found to have a stronger prognostic effect than primary follicular TLSs or LMAs in lung cancer, melanoma and pancreatic cancer^153,202,203. Accordingly, in another melanoma study, favourable prognosis was associated with TLSs containing higher fractions of B cells undergoing somatic hypermutation and decreased fractions of CD21⁺CD20⁺ B cells²⁰⁴. Similarly, across several cancer types, secondary follicular TLSs were a better predictor of clinical response to PD-1–PD-L1 blockade than less mature TLSs, PD-L1 expression or CD8⁺ tumour-infiltrating cell (TIL) density²⁰⁵. A pre-eminent role for secondary follicular TLSs suggests that affinity matured, long-lived PCs may be the most important prognostic and predictive driver. Notably, however, in the autoimmunity and transplantation fields, the importance of secondary follicular TLSs remains unresolved²⁰⁶. In lupus nephritis, LMA-like infiltrates involving B cells and PCs are common and correlate with poor kidney outcome, whereas clearly organized ectopic germinal centres are only present in a minority (6%) of patients⁷². Furthermore, a tolerogenic role for TLSs in transplantation is suggested by their presence in long-surviving heart and kidney allografts, as well as the presence of T_reg cells in many TLSs⁷³. On reflection, one might question why TLSs alone should be a reliable prognostic indicator in any setting given that TLSs are sites where immune responses are initiated, not completed. One might predict that LMAs corresponding to ‘effector zones’, once better defined and objectively measured, will ultimately represent the most robust prognostic and predictive indicators.

Targets of TIL-B-derived antibodies

The landscape of antigens recognized by serum antibodies in patients with cancer has been interrogated extensively over the past 25 years using methods such as SEREX, phage display and protein arrays^92–94, providing a foundation for understanding TIL-B-associated antibody responses. Serum antibodies have been shown to recognize tumour-associated proteins such as tumour protein p53 (REF.⁹⁵) and cancer testis antigen 1B (CTAG1B; also known as NY-ESO-1)^96,97; polypeptides arising from non-canonical reading frames⁹⁸, frameshift mutations⁹⁹ and endogenous retroviruses¹⁰⁰; and post-translational modifications such as aberrantly glycosylated mucin 1 (MUC1)¹⁰¹. Despite examples such as these, the majority of serum antibody targets identified to date are non-mutated self-proteins, which can be nuclear, cytoplasmic, transmembrane or secreted⁶. Moreover, despite concerted efforts to identify serological antigens unique to specific cancers¹⁰², most antigens discovered to date have broad or ubiquitous expression patterns, suggesting that they arise from disrupted peripheral B cell tolerance mechanisms rather than conventional adaptive immune responses. p53 is informative in this regard, as anti-p53 antibodies have been found to primarily recognize wild type rather than mutant epitopes¹⁰³, again suggesting the involvement of self-reactive B cells.

Relatively few antigens recognized by TIL-B-derived antibodies have been identified to date, as experimentally this requires either expanding TIL-Bs from fresh tumours and purifying the antibodies they secrete or molecular cloning of antibodies from TIL-Bs and producing recombinant versions for use in antigen discovery. To date, only 70 TIL-B target antigens have been identified, mainly in patient tissue samples from breast, lung and ovarian tumours (see TABLE 1 and Supplementary Table 3 for antigens, cloning methods and references). These antigens span a broad range of cellular compartments (that is, nuclear, cytoplasmic, extracellular) and molecular composition (for example, nucleic acids, lipids, proteins), although protein antigens have received the most study to date. Many correspond to known T cell antigens, including cancer testis antigens (NY-ESO-1, CT45, CT47), differentiation antigens (melanoma-associated antigen family members and MLANA), overexpressed antigens (MUC1) and mutated antigens (p53, breast cancer type 2 susceptibility protein (BRCA2)). Some are implicated in tumorigenesis (for example, p53, BRCA2, brain-derived neurotrophic factor (BDNF)), whereas others are broadly or ubiquitously expressed, such as heat shock and ribosomal proteins. In HPV-induced head and neck cancer, TIL-B-derived antibodies have been shown to recognize viral antigens (TABLE 1). To date, no studies have reported (or been designed to detect) neoantigen-specific TIL-B-derived antibodies. Thus, the view so far is that TIL-B responses predominantly involve self-antigens, which may reflect a methodological bias arising from the predominant use of wild-type protein libraries and arrays in TIL-B antigen identification efforts so far, a lower propensity for neoantigens to elicit antibody responses, and/or a higher propensity for cancers to disrupt peripheral tolerance mechanisms relevant to B cells (FIG. 2b). With regard to this last point, TIL-B responses to the self-protein matrix metalloproteinase 14 (MMP14) were recently shown to involve both germline-encoded and somatically hypermutated BCRs¹⁰⁴.

Notably, even though serum and TIL-B antibody responses involve similar classes of antigen, they appear uncoupled. In ovarian cancer, autoantibodies to p53 and NY-ESO-1 showed no association with CD20⁺ TIL-Bs³⁹ and persisted even after extensive tumour debulking, ruling out tumour-resident B cells and PCs as their main source. Similarly, in breast cancer, TIL-B-derived and serum-derived autoantibody responses were uncoupled for the majority of antigens¹⁰⁵. These observations suggest that serum antibodies and TIL-B responses can arise and persist independently. As such, it is conceivable that serological responses, hosted primarily in tumour-draining lymph nodes and bone marrow rather than the TME, may represent the major source of antitumour antibodies.

Effector mechanisms used by TIL-Bs

Arguably, TIL-Bs have the most diverse set of effector mechanisms of all immune cells in the TME, enabling them to play central roles in both innate and adaptive responses (FIG. 5). We will review cell-based mechanisms first, followed by those mediated by antibodies.

Fig. 5 |. TIL-B effector mechanisms.

a–e | Five major categories of effector mechanism. ADCC, antibody-dependent cellular cytotoxicity; ADCP, antibody-dependent cellular phagocytosis; ADO, adenosine; 4–1BBL, 4–1BB ligand; CCL, C–C motif chemokine ligand; CDC, complement-dependent cytotoxicity; FASLG, FAS ligand; GABA, γ-aminobutyric acid; GZMB, granzyme B; ICD, immunogenic cell death; ICOS, inducible T cell co-stimulatory; ICOSL, ICOS ligand; IFNγ, interferon-γ; LMA, lympho-myeloid aggregate; LTαβ, lymphotoxin-αβ; mAb, monoclonal antibody; MAC, membrane attack complex; MHC, major histocompatibility complex; NK cell, natural killer cell; OX40L, OX40 ligand; PD-L1, programmed cell death 1 ligand 1; pIgR, polymeric immunoglobulin receptor; TCR, T cell receptor; T_FH cell, T follicular helper cell; TGFβ, tumour growth factor-β; T_H1 cell, T helper 1 cell; TIL-B, tumour-infiltrating B lymphocyte; TLS, tertiary lymphoid structure; TNF, tumour necrosis factor; T_PH cell, T peripheral helper cell; TRAIL (or TNFSF10), TNF superfamily member 10; T_reg cell, regulatory T cell; TRIM21, tripartite motif containing 21.

Cytokine production and lymphoid organization

Normal B cells produce various immunostimulatory cytokines, including IL-6, IFNγ, TNF, C–C motif chemokine ligand 3 (CCL3), IL-2 and colony-stimulating factor 2 (CSF2; also known as GM-CSF)¹⁰⁶. Breast cancer-derived TIL-Bs were found to express elevated levels of IFNG, TNFA and IL4 transcripts, as well as elevated transcript levels of the immunosuppressive cytokines IL-10 and TGFβ¹⁰⁷. In metastatic ovarian tumours, TIL-Bs secreted GM-CSF, IFNγ, IL-12p40, CXCL10 and IL-7, which can stimulate macrophages, dendritic cells, T cells and natural killer cells⁴¹. By scRNA-seq, plasmablast-like TIL-Bs from patients with ovarian cancer and melanoma expressed higher levels of transcripts encoding IFNγ and chemokines that attract T cells, macrophages and natural killer cells (for example, CCL3, CCL4 and CCL5) compared with other TIL-B subsets, and indeed were associated with higher T cell infiltration^50,57. Accordingly, anti-CD20-antibody-mediated B cell depletion in patients with melanoma led to reduced T cell and macrophage tumour infiltrates⁵⁷. B cells can also contribute to TLS formation and maintenance through secretion of lymphotoxin α1β2¹⁰⁶, a role that proved essential for reducing tumour growth in a mouse melanoma model¹⁰⁸. Conversely, B cell-derived lymphotoxin promoted the outgrowth of androgen-independent prostate tumours in mice¹⁰⁹, underscoring the context dependence of lymphotoxin signalling.

Antigen presentation and spreading

Along with dendritic cells and macrophages, B cells serve as professional APCs that can process and present major histocompatibility complex (MHC) class I and II epitopes to CD8⁺ and CD4⁺ T cells, allowing them to induce, shape and amplify T cell responses¹¹⁰ (FIG. 5b). Whereas dendritic cells and macrophages use phagocytosis, pinocytosis and Fc receptor-mediated endocytosis to sample antigens from their milieu, B cells use their surface immunoglobulin molecules to take up cognate antigen, even using physical extraction via repetitive pulling and, if necessary, lysosome-mediated enzymatic digestion to do so^111,112. Consequently, antigen-specific B cells are more efficient than dendritic cells and macrophages at presenting low-abundance antigens to T cells¹¹³. Moreover, B cells are the only professional APCs capable of clonal expansion, enabling them to skew the immune response towards specific antigens in a manner unavailable to other APCs. Classic studies demonstrated that autoreactive B cells are sufficient to prime autoreactive CD4⁺ T cells¹¹⁴. Similarly, the APC function of B cells is critical in mouse models of autoimmune diseases^115–118, human coeliac disease¹¹⁹ and multiple sclerosis¹²⁰. TIL-Bs from human tumours have a phenotype consistent with an APC role, including expression of MHC class I and II⁴³ and co-stimulatory molecules such as CD80, CD86 and ICOS ligand (ICOSL)^39,54,121 (FIG. 5b). However, direct evidence that TIL-Bs present tumour antigens acquired in vivo has been provided by only one study to date, where CD69⁺CD21⁺CD27⁺ TIL-Bs isolated directly from human lung cancer samples were shown to stimulate autologous CD4⁺ TILs in vitro⁴³. A recent murine study using model antigens showed evidence of antigen presentation by tumour-specific B cells, leading to enhanced T_FH cell and CD8⁺ T cell responses against tumours¹²². Thus, further studies are warranted to elucidate this important aspect of TIL-B biology.

An APC role for TIL-Bs has fundamental implications for epitope spreading, a process in which an initial immune response to epitope A spreads to epitopes B, C and so on (FIG. 4). Epitope spreading is well documented in autoimmunity and is considered a desirable outcome of cancer immunotherapy, owing to its potential to mitigate the challenge of tumour heterogeneity. Indeed, epitope spreading from mutant to wild type antigens may underlie successful anti-PD-1 antibody therapy in melanoma¹²³. B cells have an integral role in epitope spreading through their ability to acquire cognate antigen via surface immunoglobulin, and process and present associated MHC class II epitopes to CD4⁺ T cells. In turn, CD4⁺ T cells can provide help to B cells specific for other epitopes on the same antigen (intramolecular spreading) or on physically associated antigens (intermolecular spreading). In cancer, such a mechanism could enable an initial CD4⁺ T cell response to a mutated neoantigen to trigger B cell and T cell responses to wild type flanking epitopes, which could mitigate the effects of neoantigen loss through immune editing. Evidence supporting such a mechanism was obtained in the setting of scleroderma (an autoimmune connective tissue disease) with coincident malignancy, where patients were found to harbour both CD4⁺ T cells specific for a mutated autoantigen and serum autoantibodies to the wild type autoantigen¹²⁴. Similar findings were reported for the Yo autoantigen in ovarian cancer⁷⁷. Such a process may explain the favourable prognostic impact of certain autoimmune conditions in cancer¹²⁵ and, in extreme cases, the development of the runaway autoimmune responses seen in paraneoplastic syndromes⁷⁶.

Regulatory B cells

The most commonly reported B_reg cell effector mechanism in human cancer is secretion of IL-10 (REF.¹²⁶) (FIG. 5e). Although IL-10-expressing B_reg cells are rare or absent in scRNA-seq datasets, IL-10⁺ B_reg cells have been detected by immunohistochemistry or flow cytometry in various human cancers^127–129 and after anti-PD-1 antibody treatment in patients with hepatocellular carcinoma¹³⁰. Most studies report a positive correlation between B_reg cells and T_reg cells, suggesting interaction between these subsets. IL-35 is another key regulatory cytokine produced by B_reg cells, as shown in a mouse pancreatic cancer model and further supported by bioinformatic analysis performed on human tumour data from TCGA¹³¹. GZMB⁺ B_reg cells have been detected in multiple human cancers and can inhibit T cell proliferation in vitro⁶⁸. Paradoxically, however, GZMB expression is also associated with ‘killer B cells’, which can have a cytolytic effect against tumour cells in vitro¹³², similar to Fas ligand (FASLG)⁺ and TNF superfamily member 10 (TNFSF10; also known as TRAIL)⁺ B cells^133,134. Finally, other potential B_reg cell effector molecules include adenosine (ADO)¹³⁵, TGFβ¹²⁸ and γ-aminobutyric acid (GABA)¹³⁶, although more studies are needed in human cancer.

Antibody-dependent cell-mediated phagocytosis

ADCP is mediated by myeloid cells upon Fc receptor-mediated recognition of antibody–antigen complexes, leading to phagocytosis of the target cell. ADCP has garnered increased attention lately due to the arrival of therapeutic antibodies that block the CD47-signal regulatory protein-α (SIRPα) ‘don’t eat me’ pathway¹³⁷. Although very few studies have investigated the induction of ADCP by TIL-B-derived antibodies (FIG. 5c), Biswas et al. recently described a monoclonal IgA against tetraspanin 7, which in association with CD351⁺ myeloid cells mediated an antitumour effect in an ovarian cancer mouse model¹³⁸.

Antibody-dependent cell-mediated cytotoxicity

ADCC is mediated by natural killer cells and myeloid cells upon Fc receptor-mediated recognition of antibody–antigen complexes, leading to target cell lysis. Although natural killer cells and myeloid cells are abundant in multiple cancer types¹³⁹ and therapeutic antibodies such as rituximab can induce ADCC¹⁴⁰, there is no direct evidence to date of ADCC induction by TIL-B-derived antibodies (FIG. 5c). Nonetheless, in a bioinformatic analysis of TCGA data, B cell signatures were positively associated with expression of GZMB and Fc fragment of IgG receptor IIIa (FCGR3A), suggesting natural killer cell-mediated ADCC¹⁴¹. Moreover, around 20% of tumours showed increased expression of genes encoding metalloproteinases that cleave the natural killer cell-stimulating ligands MHC class I polypeptide-related sequence A and B¹⁴¹, consistent with a previously described mechanism of natural killer cell evasion¹⁴².

Complement-based mechanisms

The complement cascade can be triggered by the classical, lectin or alternative pathways, culminating in the formation of the membrane attack complex (MAC) and target cell lysis¹⁴³. In general, complement-dependent cytotoxicity (CDC) is a difficult mechanism to profile in tumour tissues, as it involves more than 30 protein components that are regulated primarily through post-translational mechanisms. Nonetheless, the finding that most cancer types express high levels of complement inhibitory receptors, such as CD46, CD55 and CD59, strongly implicates CDC as important in human cancer¹⁴³ (FIG. 5c). Indeed, neoadjuvant chemotherapy in patients with breast cancer induced an ICOSL⁺ B cell population via the complement–complement C3d receptor 2 axis, which correlated with an increased T effector cell to T_reg cell ratio and tumour clearance⁵⁴. This effect was antagonized by tumour cells expressing the complement inhibitory receptor CD55. By contrast, in clear cell renal cell carcinoma, classical pathway activation (as evidenced by co-localization of C1q and IgG on tumour cells) was associated with CD163⁺ M2 macrophages, T cell exhaustion and poor prognosis¹⁴⁴, in accord with the immunosuppressive properties of C1q (REF.¹⁴⁵). Similarly, IgM-dependent activation of the classical pathway was reported in human lung cancer, and was shown to inhibit T cell responses in a corresponding mouse model¹⁴⁶.

Whereas ADCP, ADCC and CDC are generally directed towards cell surface antigens, a significant proportion of TIL-B-derived antibodies recognize intracellular antigens (TABLE 1; see Supplementary Table 3), raising the question of how such responses might contribute to antitumour immunity. The following effector mechanisms, although understudied in cancer so far, represent potential means by which TIL-Bs could impact both extracellular and intracellular processes in cancer cells, representing a new avenue that warrants further study.

Antibody-dependent intracellular neutralization

Antibody-dependent intracellular neutralization is a process by which antibodies induce the proteasomal degradation of intracellular proteins^147,148. Although antibody-dependent intracellular neutralization has been studied primarily in the context of viral infection, TIL-B-derived IgG from lung cancer was shown to promote degradation of the tumour protein Ras homologue family member C by a mechanism resembling antibody-dependent intracellular neutralization⁴⁸ (FIG. 5d).

Antibody-mediated signalling interference

Upon binding to their cognate antigen, antibodies can modulate signalling pathways that influence tumour growth or behaviour (FIG. 5d). For example, in a mouse model, IgG against glycosylated heat shock protein A4 induced CXCR4 expression on tumour cells which facilitated lymph node metastasis¹⁴⁹. Moreover, TIL-Bs from a patient with ovarian cancer were found to produce antibody that antagonized the tumour-promoting effects of BDNF¹³⁸.

Transcytosis

The above-mentioned study also introduced IgA transcytosis as a new TIL-B-mediated antitumour mechanism¹³⁸ (FIG. 5d). Mucosal epithelial cells express the polymeric immunoglobulin receptor (pIgR) on their surface, which binds to PC-derived IgAs and is internalized. This internalized complex is then transported from the basolateral to the luminal space. Remarkably, in vitro exposure of pIgR⁺ ovarian tumour lines to IgA (either tumour antigen-specific or nonspecific) inhibited oncogenic signalling and promoted T cell-mediated killing by a pIgR-dependent mechanism. Similarly, IgA infusions slowed the growth of pIgR⁺ human ovarian tumours engrafted in mice¹³⁸, suggesting a provocative new therapeutic concept with potential relevance to many human cancers.

Therapeutic manipulation of TIL-Bs

Few cancer therapies have been designed to intentionally induce TIL-B responses; however, our evolving understanding of the myriad cellular and antibody-based mechanisms used by TIL-Bs offers many exciting opportunities (FIG. 5). Below, we provide an overview of recent insights and advances.

Chemotherapy and targeted agents

Baseline TIL-B density is positively associated with response to chemotherapy in humans and mice^54,63. Furthermore, chemotherapy can further increase TIL-B density^41,150 and expression of co-stimulatory molecules^54,151, as well as induce TLSs¹⁵². Notably, the use of corticosteroids during chemotherapy can be detrimental to germinal centre formation in TLSs¹⁵³. Compared with cytotoxic chemotherapy, targeted agents may interact differently with TIL-Bs, as a negative association between TIL-Bs and response to BRAF and MEK inhibitors was seen in patients with melanoma¹⁵⁴. STING agonists can promote the formation of immature TLSs in a mouse melanoma model and may prove useful in combination with agents that induce TLS maturation¹⁵⁵. Finally, a triple regimen involving low-dose chemotherapy, oncolytic virus and dual checkpoint blockade led to enhanced TIL-Bs and increased tumour regression in a breast cancer mouse model¹⁵⁶. Building on these early studies, further work is warranted to identify optimal strategies for using chemotherapeutic and targeted agents to amplify TIL-B responses.

Immune checkpoint blockade

The presence of PD-1⁺ and PD-L1⁺ TIL-Bs has been reported in multiple human cancers^{130,157–159}, providing an initial suggestion that TIL-Bs may be directly influenced by PD-1 and/or PD-L1 targeting immune checkpoint blockade; analogous data for CTLA4 expression have not yet been reported. Substantial evidence is accumulating for a strong, positive effect of anti-CTLA4 and anti-PD-1 antibody blockade on TIL-Bs and TLSs, reflecting the aforementioned roles played by these pathways in regulating interactions between B cells, T_FH cells and follicular T_reg cells. Baseline TIL-B densities predict response to PD-1 and CTLA4 blockade in patients with multiple cancers^89,160–162 and, indeed, are augmented by immune checkpoint blockade^160,161,163. In patients with non-small-cell lung cancer treated with neoadjuvant anti-PD-1 antibody, regressing tumours were associated with dense tumour immune infiltrates, including the presence of PCs and TLSs, whereas unstructured lymphoid aggregates in tumours showed no association with response¹⁶⁴. In patients with urothelial cancer, the presence of stromal B cells and immature TLSs in tumours was associated with a lack of response to combined anti-CTLA4 and anti-PD-1 antibody treatment, yet mature TLSs were seen in responding tumours after treatment¹⁶⁵, suggesting a substantial, favourable transformation of the TIL-B response.

Preclinical models provide direct mechanistic evidence for a favourable influence of immune checkpoint blockade on TIL-Bs and TLSs. B cells were required for response to CTLA4 and PD-1 blockade in a mouse breast cancer model⁶³, and addition of an anti-PD-L1 antibody to radiation therapy converted a B_reg cell response to an effector B cell response associated with improved tumour control in a squamous cell carcinoma model¹⁶⁶. Similarly, in a mouse melanoma model, immune checkpoint blockade led to quantitative and qualitative improvements in TLSs and increased efficacy in a melanoma model¹⁰⁸. Finally, in a lung cancer model, anti-PD-1 antibody treatment increased the number of TLSs and the production of IgG by TLSs, along with improved tumour control¹⁶⁷. Thus, immune checkpoint blockade provides a powerful means to enhance both T cell and B cell responses in cancer.

Agonistic anti-CD40 antibodies

Agonistic anti-CD40 antibodies are under clinical evaluation for several cancers¹⁶⁸ and are expected to strongly influence TIL-B responses, given the central role of CD40 in B cell biology¹⁶⁹. Indeed, in a mesothelioma model, anti-CD40 antibody treatment enhanced TIL-B density and B cell-dependent tumour control¹⁷⁰. In mouse models and patients with cancer, anti-CD40 antibody treatment induced a sharp decrease in the levels of circulating B cells, while increasing expression of co-stimulatory molecules, possibly reflecting activation-induced extravasation from blood^171,172. In a glioma model, anti-CD40 antibody treatment induced TLSs and other lymphoid aggregates, but these were associated with hypofunctional T cells, suppressive CD11b⁺ B cells and impaired response to immune checkpoint blockade¹⁷³. Thus, further work is required to determine how best to modulate the CD40 pathway in favour of TIL-B responses.

Cytokine therapy

Cytokines such as IL-2, IL-15 and IL-21 not only support the differentiation and proliferation of T cells and natural killer cells but also promote B cell development and isotype switching¹⁷⁴. Indeed, IL-21 was required for curative responses to immune checkpoint therapy in breast cancer models⁶³. Although IL-2 is commonly used to stimulate cytolytic T cells, it inhibits T_FH cell differentiation and promotes T_reg cell responses, and hence may be detrimental to TIL-B responses¹⁷⁵. Other cytokines that could potentially be deployed to promote TIL-B responses and TLS formation include BAFF (a B cell maturation and survival factor) and CXCL13 (a B cell chemoattractant), both of which are generally prognostically favourable in human cancer^176,177. Notably, however, CXCL13 can also mediate pro-tumour effects¹⁷⁷. Finally, a drug targeting TNFSF14 to tumour vessels through linkage to a vascular targeting peptide induced TLS formation and normalized tumour vasculature, leading to enhanced T cell infiltration and therapeutic synergy with a tumour vaccine¹⁷⁸. Thus, cytokines offer several promising avenues for promoting TIL-B responses, although as with all cytokine therapies, the challenge is to develop strategies that specifically enhance antitumour immunity while avoiding unwanted immunological effects.

Vaccines

Various cancer vaccines have been shown to induce TLS formation in patients, including a heterologous DNA viral vaccine against HPV antigens in cervical intra-epithelial neoplasia¹⁷⁹ and an allogeneic tumour cell-based vaccine in pancreatic cancer¹⁸⁰. Furthermore, antigen-loaded, CD40-activated B cells are being investigated as cell-based vaccines and proved equivalent to dendritic cells in models of melanoma and lymphoma¹⁸¹; this approach was further enhanced by co-transfer of tumour antigen-specific PCs¹⁸². Recently, TNFSF9⁺ (also known as 4–1BB ligand, 4–1BBL) B cells stimulated with CD40 agonist and IFNγ showed impressive efficacy in combination with radiation, CD8⁺ T cells and PD-L1 blockade in a glioblastoma model when compared with the same combination minus stimulated B cells¹⁸³. There are no reports so far about B cell or antibody responses to personalized neoantigen-targeting vaccines¹⁸⁴, but these may be expected based on known mechanisms of antigen presentation and epitope spreading (FIG. 4).

T_reg cell and B_reg cell depletion

Given the essential role of T_reg cells in preventing systemic autoantibody responses^8,12, T_reg cell depletion is a compelling approach to enhance TIL-B responses. Indeed, in a lung cancer model, T_reg cell depletion led to expansion of TLSs that supported improved T cell proliferation and dendritic cell function, leading to tumour destruction¹⁸⁵. Although B_reg cell depletion is challenging given the lack of unique markers, B_reg cell differentiation and function can be impacted by certain chemotherapeutic agents^54,166,186 and inhibitors of STAT3 (REF.¹⁸⁷), MEK¹⁸⁸ and BTK¹⁸⁹. Furthermore, T_reg cell and B_reg cell activity can potentially be inhibited by antagonism of IL-10, IL-35, TGFβ and other suppressive factors.

Antibody-related therapies

Finally, there are several approaches to leverage the antibody-mediated effects of TIL-Bs. Transformative methods such as scRNA-seq enable high-throughput cloning of TIL-B-derived antibodies, so the challenge now is to identify which antibodies and cognate antigens are best suited for safe and effective therapeutic targeting, as recently reviewed⁹³. An important consideration is that some antibodies can have pro-tumour effects¹⁴⁹ and carry the risk of autoimmune sequelae⁷⁶. That said, high-affinity, autoreactive antibodies do not necessarily induce unacceptable immunopathology, as demonstrated by the use of autoreactive, broadly neutralizing antibodies against HIV-1 (REF.⁹). Once antibodies with appropriate specificities are identified, they can be further engineered or used in combination therapies to enhance desired antitumour properties¹⁹⁰. Of particular relevance to TIL-B effector mechanisms, Heemsherk et al. recently described a bispecific antibody that induced tumour regression in vivo by triggering neutrophil-mediated cytotoxicity, ADCC and ADCP¹⁹¹. Other promising approaches include CD47 blockade to promote ADCP¹⁹² or blockade of CD55 or CD46 to promote complement-mediated killing mechanisms¹⁹³.

Conclusion

There is now abundant evidence that TIL-Bs and the antibodies they produce are prominent features of the immune response to human cancer. TIL-Bs are found alongside T cells, natural killer cells, myeloid cells and others in the TMEs of the ‘hottest’ tumours and show hallmarks of active antigen recognition and diverse effector functions. They carry strong prognostic and predictive significance in many indications and treatment scenarios, including CTLA4-directed and PD-1-directed immune checkpoint blockade. Moreover, their numbers and functional attributes can be modulated by standard treatments, immune checkpoint blockade, therapeutic vaccines and various other experimental approaches. Beginning with pioneering studies two decades ago, the target antigens of TIL-Bs are being elucidated, so far revealing a preponderance of self-proteins with varying degrees of tumour-specific features. In this and other ways, TIL-Bs share many features with the pathogenic B cells and PCs that mediate autoimmune responses. If TIL-B specificities are indeed skewed towards self-reactivity, one can envision this providing an effective means for the immune system to contend with intratumoural heterogeneity through the recognition of non-mutated antigens with more ‘truncal’ expression patterns within the tumour phylogeny.

Having garnered intense interest in the past few years, the TIL-B field faces exciting new challenges. Most, if not all, TIL-B effector functions depend on the nature of their cognate antigens; for example, ADCC requires surface antigens, whereas signal interference requires antigens that drive tumour cell fitness. Therefore, to fully understand the mechanisms of action of TIL-Bs, we require a broader, unbiased view of their antigenic landscape, which will require the use of multiple, orthogonal discovery methods. scRNA-seq is enabling the first mapping of TIL-B clonotypes to phenotypes, ushering in an era when recombinant antibodies from TIL-B phenotypes of interest can be readily produced and used to probe antigen specificities and effector mechanisms on an unprecedented scale. In parallel, we need to decipher and more precisely describe the various cellular neighbourhoods in which TIL-Bs are found, moving beyond TLSs to include the other fascinating structures revealed by multiplex immunohistological methods. To this end, spatial transcriptomics with matched BCR and TCR sequencing data promise to be transformative. Finally, we need to leverage insights from the above endeavours to design a new generation of immunotherapies purpose-built to leverage TIL-B effector mechanisms in concert with T cells and other immune cells. Again, the autoimmunity field provides rich examples of powerful, multipronged, tissue-restricted and unrelenting immune responses against many of the same peripheral tissues that challenge us as malignancies. To develop this next generation of immunotherapies, new preclinical models are urgently needed that recapitulate the complex cellular ecosystems in which TIL-Bs operate. Investment in the above objectives will pay dividends in the form of new target antigens, therapeutic antibodies and integrated immunotherapies that can contend with the extensive heterogeneity and unremitting evolution of metastatic cancers.