The role of B lymphocytes in the immuno-biology of non-small-cell lung cancer

Immune responses within the tumour microenvironment (TME) are important determinants of tumour behaviour, progression, and aggressiveness [1]. The role of T-cell-mediated immune responses in solid tumours is well established and has become the driving force for major therapeutic advances, specifically with the advent of immune checkpoint inhibitors [2]. The success of single agent and combination checkpoint blockade, which enables antigen-experienced microenvironmental T cells, has transformed the treatment paradigm of advanced NSCLC both as a front-line option and in the platinum-refractory setting. As a consequence, our understanding of the biology of these T cells has expanded exponentially. In contrast, our knowledge of B-cell biology in cancer is less well developed. Tumour-infiltrating B cells have been observed in all stages of lung cancer development, with their presence differing according to stage and histological subtype. Given that they play a role in both humoral and cellular immunity B-cell parameters may be important in determining both responsiveness to and toxicity of checkpoint blockade. Manipulation of B-cell biology might offer significant immune-therapeutic opportunities [3]. Thus, understanding B-cell biology in NSCLC is of fundamental importance in informing potential novel multi-faceted approaches to favourably reset the immune contexture of the cancer microenvironment.

In the bone marrow, haematopoietic stem cells serve as the common lymphoid progenitor and continually give rise to B cells throughout life [4]. B-cell development and differentiation are subsequently regulated through the B-cell receptor (BCR). On leaving the bone marrow, B-cell development takes place in B-cell follicles within secondary lymphoid organs (SLOS), where germinal centres (GC) develop in response to antigen encounter. Mature B cells undergo a process known as somatic hypermutation (SHM), where enzyme induced mutations are introduced in the heavy and light chain variable regions to further diversify the immunological repertoire. The same enzyme mediates immunological class switch recombination (CSR), replacing the μ constant region for another heavy chain region to generate IgA, IgE or IgG. The cells with the highest affinity B-cell receptors (BCR) are positively selected (this requires signalling through the BCR for the cell to survive, negative selection occurs when the B-cell antigen receptor binds self-antigen and, therefore, undergoes cell death) and undergo clonal expansion in the germinal centre. These class-switched B cells can then give rise to long-lived plasma cells or memory B cells [5]. Germinal centres, which are normally found in SLOS such as the spleen and lymph nodes, can also occur as tertiary lymphoid structures (TLS) in tumours; this has been reported in colon, breast and NSCLC [6‐8]. The de novo formation of ectopic lymphoid tissue can occur at the site of inflammation in potentially any organ system [9]. There is evidence that the adaptive immune system can be initiated independent of SLOS in these TLS in NSCLC [7].

There appears to be a significant difference in the role of B cells in animal models which represent different stages in the development of cancer. In murine models of pre-malignancy, B cells appear to drive inflammation which in turn induces pre-malignancy.

Table 1 summarises studies that have identified TIL-Bs in tissue and examined their associated prognostic significance. These studies largely focus on NSCLC which is where the correlation between TIL-Bs and disease-specific outcome has been shown to be strongest by comparison with other forms of Lung cancer. However, groups have investigated the immune milieu in large cell carcinoma and Small Cell Lung Carcinoma (SCLC). In the former, Eerola et al. have demonstrated that TIL-Bs do correlate with better overall survival [35], whereas in SCLC, B-cell infiltration is significantly reduced compared with CD8 + T cell and Macrophage infiltration and TIL-B number has not been shown to associate with survival (p = 0.634 by log rank) [36].

In Table 1, the majority of studies have used a combination of immunofluorescence or immunohistochemistry to identify these cells in tissue. Of the 28 studies identified, 18 have used IHC as a primary method, mRNA sequencing has been employed in 4 studies and 2 have employed qt-PCR [37‐41]. An anti-BCL-6 antibody and an anti-CD21 antibody were utilised to identify Germinal Centre B cells and mature B cells, respectively. Immunohistochemistry was used to determine B-cell density, location and phenotype within NSCLC tissue [8]. Studies have utilised PCR microarray and Mrna-sequencing techniques to identify primarily humoral immunity related gene signatures in NSCLC specimens [42, 43]. CD20 + B-cell infiltration has been shown to be positively prognostic in NSCLC by a number of different groups [38‐43]. Disease-free and overall survival was significantly higher in non-smokers with non-squamous NSCLC [45]. Significantly improved survival has also been shown in large cell carcinomas with higher degrees of CD20 + B-cell infiltration [35]. Associations have been made between TLS (GCs forming as ectopic foci of follicular B cells and clusters of mature DC-Lamp^+ve Dendritic cells and T cells in cancer tissue, in response to antigen stimulation) in NSCLC and improved long-term survival. The presence of both types of antigen-presenting cells and mature dendritic cells in these TLS strongly predicts the outcome of patients [7, 44]. A low density of both follicular B cells and mature dendritic cells allows the identification of patients at high risk of poor survival. A higher prevalence of intra-tumoural GC formation was found in NSCLC stage I tumours compared with higher stage (II–IV) tumours (p < 0.02) [8]. In a recent study [49] the expression of a tumour-induced plasmablast-like B-cell signature (TIPB) was significantly correlated with the expression of CD8a signatures and the density of CD8 + cells. High expression of the TIPB signature was correlated with overall survival in the melanoma TCGA data set. Importantly, a cohort of melanoma patients treated with anti-CD20 antibodies, showed significant on-treatment down-regulation of the TIPB signature: the signature was highly correlated with tumour inflammatory score, interferon gamma and T cell effector signatures all of which significantly decreased on anti-CD20 therapy. There was a marked depletion in both CD4 + and CD8 + cell density at the invasive tumour-stroma margin and a reduction in the TLS area, an effect which was prolonged. In support of this data suggesting the importance of B cells in a successful anti-cancer immune response, long-term follow-up of CD20 depletion with Rituximab in patients with lymphoma, it was shown that CD20 depletion was an independent risk factor for the development of secondary solid tumour malignancy in both univariate and multivariate analyses [50].

Finally, the prognostic impact of follicular B cells was evaluated in two patient cohorts; early stage untreated NSCLC and advanced stage NSCLC treated with neoadjuvant chemotherapy. “Foll-B-Hi” patients had significantly prolonged survival in early stage disease, (97% DFS at 4 years compared with 62% in the “Foll-B-Lo” group), and in advanced stage disease, a benefit was demonstrated albeit not significant (56 month median DFS compared with 23 months in the “Foll-B-Lo” group). The global increase in follicular B-cell density was associated with an overall increase in mature DC density. When the combined immune populations were taken into account and correlated with survival, “Foll-B-Hi/mDC-Lamp^HI” patients had the highest median survival, 100% of early stage patients (p < 0.04) and 55% of advanced disease patients (p = 0.007) were alive after a follow-up of 50 and 60 months, respectively [44]. “Foll-B-Lo/mDC-Lamp^LO” patients had the worst prognosis.

Some studies have not demonstrated a prognostic impact of B-cell density on NSCLC outcomes [48, 51‐54]. However, it may be that the lack of prognostic impact may relate to the high density of Bregs in such studies and it would appear to be essential that Breg density be considered separately [48, 51, 52]. An explicit analysis of whether Breg density is negatively prognostic for outcome, however, has not been performed. Finally, the prognostic impact may be dependant not only on enumeration of the appropriate B-cell subsets (TIMPs, follicular) but also by enumeration of tumour-associated B cells in the appropriate compartment, as opposed to analysis of un-segmented tumoural B-cell density. Most B cells are found at the invasive tumour-stroma margin, and it is here that the cancer cells are likely to polarise B cells to the immune-stimulatory TIPB phenotype.

B cells can exist in a continuum of naïve cells to terminally differentiated plasma cells within the TME and more specifically within the TLS [44]. Determining the ratio between these so-called “anti-tumour” TLS derived TIL-Bs and the “pro-tumour”, inhibitory Bregs is important to understand the biology and long-term outcome from this disease. This balance is likely influenced by the microenvironmental cues which play a role in determining B-cell polarity. CXCL13 and Lymphotoxin have been identified as two factors critical to the formation and development of lymphoid follicles in the gut [84], and in lung cancer, B cells produce CXCL13 and Lymphotoxin via TLR4 signalling which acts as a positive feedback loop to support the formation and high density of TLS [85, 86]. CXCR5 expressing B cells stimulated by CXC13 coupled CpG-ODN can trigger the cytolytic effect of CD8 + T cells leading to the abrogation of metastasis in 4T1.2 tumour-bearing mice [23]. Resveratrol, Lipoxin, Glucosides of Paeony have also inhibited Bregs through STAT3 and/or ERK inactivation leading to a reduction in IL-10 and TGF-β levels thus exerting an anti-tumour effect [87]. B-cell homeostasis and thus polarity will largely be determined by the degree of inflammation within the tumour, factors such as tissue hypoxia, intra-tumoural vascularity, cytokine milieu and cellular infiltration are all factors which are likely to exert control over the pro versus anti-tumour B-cell balance but as yet there is little evidence describing the Breg/B effector ratio in tumour biology, and this is likely due to the transient inducible nature of Bregs.

Immune checkpoint blockade antibodies have improved cancer therapy by overcoming the inhibition of T cell effector functions, yet a significant proportion of patients still do not respond to such therapies. In the first study to investigate whether B-cell density impacts outcome with checkpoint blockade, B-cell content was determined in 34 melanoma patients undergoing PD-1 blockade monotherapy and evaluated for response [88]. Dichotomising the patients at the median of CD20 positive cells in at least one histospot there was no difference in response or survival between those with high and low B-cell density. The comments made earlier about enumeration of specific B-cell populations in specific microenvironmental segments made earlier also apply here. B-cell depletion in the MC38 (colon carcinoma) and YUMMER1.7 (melanoma) models did not impact the efficacy of anti-PD-1 treatment. Anti-PD-1 outcomes were similar in muMT mice (mice lacking B cells) and WT mice bearing MC38 tumours [88]. Larger data sets in other cancers need to be interrogated to fully understand whether there is any impact of specific intra-tumoural B-cell populations on the outcome of checkpoint blockade. This is particularly the case given earlier data showing that clinical benefit with ipilimumab was greater in melanoma patients with sero-positivity against NY-ESO [89]: as mentioned above, the CTags appear to be potent immunogens stimulating antibody responses. Furthermore, gene-expression profiling in urothelial carcinoma and melanoma patients undergoing both anti-PD-1/PD-L1 and anti-CTLA4 therapy identified a memory B-cell (MBL) signature which was significantly and reproducibly elevated in patients showing clinical benefit [90]. It significantly outperformed other immune cell signatures and remained significantly associated with outcome when including tumour mutational burden, copy number aberration burden and checkpoint expression. Samples enriched for an innate PD-1 resistance scores had significantly lower levels of MBL scores. The MBL score positively correlated with BCR heavy chain expression and the expression of T cell activation genes, MHC class II and genes responsible for B-cell proliferation and activation within the TME. Finally, high expression of the TIPB signature was associated with improved survival in melanoma patients treated with anti-PD-1, and plasmablast-like and naïve B-cell frequencies were significantly higher in patients responding to immune checkpoint blockade [49]. The use of anti-PD-1 treatment in murine models has shown to increase total IgG and OVA-specific IgG production in OVA-immunised mice [91]. The enhanced humoural response in these mice is thought to be mediated by CD4 + ICOS + T cells which are presumably of the T follicular helper (Tfh) phenotype that go on to augment terminal B-cell differentiation in the germinal centre [91]. PD-1/PD-L1 interactions between Tfh and Bregs control this axis [92, 93] and by blocking this checkpoint, Tfh cells are released from Breg-induced suppression. This demonstrates the importance of heterogeneity of the B-cell repertoire and how checkpoint blockade can impact downstream immune responses by targeting select populations. Importantly, none of the above studies have examined B-cell density as predictive of response to checkpoint blockade in NSCLC or to combination chemo/immunotherapy which has become a 1st line standard of care in this disease.

Given that PD-1 is expressed on B cells and can limit B-cell responsiveness [94, 95], and furthermore, the certain autoimmune conditions are mediated via auto-antibody formation, the association of B-cell sub-populations with checkpoint blockade toxicity has become a focus of investigation. Das et al. demonstrated in melanoma patients, a detectable decline in circulating B-cell numbers together with an increase in CD21^LO B cells and plasmablasts after the first cycle of combination checkpoint blockade therapy [96]. These treatment induced changes in B-cell numbers preceded and correlated with both the frequency and timing of immune related adverse events (irAE). Early B-cell changes correlated with a higher rate of grade 3 or higher irAEs 6 months after starting treatment. Contrastingly, several groups have shown through case report series’ that B-cell depletion therapy using Rituximab successfully treated B-cell-mediated irAEs in NSCLC [97, 98], SCLC [99], Melanoma [100] and Urothelial Carinoma [101]. This recapitulates the idea of B-cell enumeration and selective targeting of certain microenvironmental B-cell populations. Bregs which by nature are immunosuppressive and dampen down inflammation would limit anti-tumour activity in response to PD-1 blockade but effector B-cell populations with robust humoural responses and T cell activation mechanisms are driving autoimmunity and irAE in response to PD-1 blockade. These intriguing analyses need to be validated, extended and applied to other cancers amenable to checkpoint blockade and the mechanisms underlying these observations identified.

Studies in mouse models of pre-malignancy suggest that B-cell-mediated inflammation may be important in promoting the progression to invasive malignancy. Given the huge promise of reversing the pre-malignant phenotype to reduce the cancer burden, there is an urgent need to understand the role of B cells in human metaplasia, dysplasia and in situ cancer and how they mediate progression through these stages to decide whether B-cell-directed strategies may be of value in reducing the progression of pre-malignancy.

Studies examining B cells with a regulatory phenotype (Bregs) consistently suggest that Breg infiltration may enhance tumour progression. The factors that induce Bregs in human malignancy need to be defined. Specifically are there particular microbes, TLR ligands or cancer cell produced cytokines in the TME that polarise B cells to a Breg phenotype [14, 102]. Currently used B-cell depleting antibodies cannot distinguish between effector and regulatory B-cell subsets; therefore, meticulous phenotypic characterisation and study of this subset in the TME [14, 102] is required to identify Breg specific targets that can be exploited to selectively deplete Breg populations but more fundamentally to fully understand the role of Bregs in human cancer. There are some current potential anti-Breg strategies. In vivo murine studies have displayed selective Breg depletion using LXA4 without affecting conventional B-cell proliferation, differentiation and germinal centre formation thus promoting anti-tumour responses [87]. An alternate to Breg depletion would be repolarisation of this subpopulation into B effector cells, as has been shown with TLR9 ligands in vitro [22, 23]. Adoptive transfer of CpG-pulsed B cells with effector phenotypes into patients with established cancer could be employed to shift the balance in favour of an anti-tumour B-cell response within the TME.

More work is needed to understand the anti-tumour impact of antibodies against tumour associate antigens, particularly CTags which appear to be strong immunogens, and to identify new humoural immunity targets. The disappointing results of the MAGRIT trial vaccinating NSCLC patients in the adjuvant setting [103] should not be taken as suggesting that harnessing the anti-tumour antibody response should be deprioritised: mono-epitopic vaccination as cancer therapy has a long history of failure. Multi-valent vaccines, preferably against personalised B-cell antigens, are one option. Building on the model of the chimaeric antigen receptor T cells (CART), highly specific B-cell receptors to critical tumour antigens could be cloned into autologous B cells and transferred into patients with resultant high specificity and high affinity anti-tumour Ig production. Alternatively, antibodies could be produced ex vivo and adoptively transferred. Given the role of B-cell PD-1 expression in mediating B-cell hypo-responsiveness, the role of PD-1 blockade in augmenting these strategies should be explored, as a research priority. Understanding B-cell biology will help to refine the understanding behind the effects of checkpoint blockade on the immune milieu. Toxicity from these therapies is the Achilles heel of this treatment strategy. As was alluded to earlier, work in mice and humans has demonstrated that PD-L1^hi Bregs play a role in the suppression of humoral immunity through Tfh cell regulation; moreover, these cells are resistant to classical anti-CD20 therapy [93]. Firmly understanding the ontogeny of these B cells and their relationship to other B-cell subsets, including other Breg phenotypes is of paramount importance if we hope to be able to refine therapeutic strategies so as to augment anti-tumour protective immunity and dampen down autoimmune and hence toxic responses.

Finally, large scale prospective and careful B-cell sub-type specific and microenvironment segment specific analyses are required in lung cancer and in other cancers to clarify the role of B cells in modulating the responsiveness to checkpoint blockade and in mediating the toxicity to these therapies. These studies will define the role of B-cell-targeted strategies in augmenting the activity of, reducing resistance to and the ameliorating toxicity of this crucial class of anti-cancer agents.