The immune system and malignant cells interact via a complex network. The importance of immune function in tumour development and control has been acknowledged for decades. As immunotherapy enters clinical practice, these underpinnings have more relevance as we try to identify predictive biomarkers for benefit from new therapies. The seminal papers describing the ‘hallmarks of cancer’ pronounced the capacity to avoid immune destruction as one of the requirements for malignancy (Hanahan and Weinberg, 2011). There is now a vast literature supporting immunosurveillance as a significant contributor to the natural history of malignancy. The interaction between tumours and the immune system has been described in three scenarios of ‘immunoediting’ (Schreiber et al, 2011); specifically: elimination (where immune surveillance successfully eradicates malignant cells); equilibrium (where the immune system exerts control over abnormal cells) and escape (where tumour cells evade immune mechanisms allowing growth and metastasis) (Mittal et al, 2014). In this mini-review, we aim to define the immune infiltrate and its spatial organisation as well as summarising the prognostic value of immune cells in different solid cancers.

Subtypes of immune cell infiltrate

Historically, studies have focused on the interaction between cytotoxic T lymphocytes and cancer cells. The role of other immune cells is also now recognised as contributing to the complex immune response in cancer, some of which promote tumour control and others facilitate cancer progression (Table 1; Figure 1).

Table 1 Cell types in the tumour immune infiltrate
Figure 1
figure 1

Pathways affecting cytotoxic T lymphocyte activity within the tumour microenvironment.

PowerPoint slide

Location and spatial organisation of the immune cell infiltrate

Although the distinction between peritumoural, stromal and intratumoural lymphocytes is made histopathologically (Table 2), this is likely an artificial segregation as this is a dynamic network that allows chemokine-generated cell movement between these areas.

Table 2 Location of the immune infiltrate

A qualitative description of the interplay between tumour and immune cell infiltrate has been termed the ‘immune contexture’, and includes the location of specific immune cells, tertiary lymphoid structures (ectopic lymphoid aggregates that are generated during immune stimulation and exhibit structural characteristics of lymphoid organs), and the chemokines and cytokines involved in this microenvironmental organisation (Fridman et al, 2011). Methods for describing the immune infiltrate are a limitation of current pathological reporting. There remains a lack of consensus regarding reporting of tumour-infiltrating lymphocytes (TILs), including methods to subtype the infiltrating cells and their spatial organisation. International working groups are trying to create and validate reporting guidelines (Salgado et al, 2015). Immunophenotyping of the immune infiltrate by immunohistochemistry or immunofluorescence staining can be performed in tissue samples, or after generation of cell suspensions that are generated by mechanical or enzymatic breakdown of fresh tumour tissue (Stoll et al, 2015). The use of sectioned tissue specimens allows spatial understanding of cell position relative to tumour cells, however, similar to the use of cell suspension, is limited by challenges in antigen retrieval (capacity to bind identifying/specific proteins of interest) and poor standardisation. Novel approaches such as mRNA characterisation of immune co-regulated genes may help to identify and characterise the immune infiltrate (Stoll et al, 2015). Meta-analytical data suggest that in most cancers, the immune infiltrate is heterogeneous and there is limited reproducibility of leukocyte subtypes (Stoll et al, 2015). Although the presence of T cells is clearly important, the interplay between tumour antigens and major histocompatibility complex (MHC) molecules for antigen presentation is critical for efficient T cell activation. High affinity of the targeted peptides for MHC is required for strong stimulation of T cells to secrete cytokines and produce tumour eradication or control (Engels et al, 2013). The specific antigenicity of coding exons in mutated cancer genes is an area of research and the capacity to sequence whole genomes with greater speed and reduced cost is enhancing the capacity to identify potentially antigenic mutations.

Prognostic value of infiltrating immune cells

The prognostic value of lymphocytes in stromal, peritumoural and intratumoural locations remains unclear, with conflicting data from different tumour sites. Peritumoural lymphocytes at the advancing tumour margin and those in direct contact with tumour cells have been purported to carry the most prognostic weight particularly in some disease sites (see below). In general terms, a ‘pro-inflammatory’ tumour microenvironment and infiltrating CD8-expressing T lymphocytes are associated with improved clinical outcomes in a broad range of tumour types. In contrast, the inhibitory function of other immune cells, for example, myeloid-derived suppressor cells and regulatory T cells (Tregs) appear to have a major role in disrupting the capacity for the immune control of cancers and are therefore associated with worse outcome.

Perhaps counter-intuitively, favourable outcomes have also been observed in tumours infiltrated by inhibitory immune cells, for example, forkhead box P3-positive regulatory T cells (FOXP3) cells in colorectal cancer. This may represent a feedback loop in the context of an existing anti-tumour immune response and thus actually indicate increased tumour immunogenicity (Gajewski et al, 2013). Myeloid-derived suppressor cells (MDSC) and tumour-associated macrophages are both capable of negative regulation of innate and adaptive immune pathways. MDSCs have a role in tumour growth and metastasis via promotion of immune privilege (ability to tolerate the introduction of antigens without eliciting an inflammatory immune response), tumour microenvironment remodelling, establishment of a pre-metastatic niche (a scenario where non-cancer cells promote future metastasis) and interaction with tumour to promote differentiation, invasion and angiogenesis (Marvel and Gabrilovich, 2015). There is evidence that MDSC expansion is associated with more advanced stages of malignancy in multiple cancer types and also correlates with poor prognosis independent of tumour burden (Ugel et al, 2015). Paradoxically, anti-tumour immunity also leads to selective pressure on malignant cells, which ultimately leads to survival of tumour cells with reduced immunogenicity (Shankaran, 2001). There are also data supporting the hypothesis that tumour-infiltrating immune cells can promote invasion and metastases (Man et al, 2013), which may in part explain the heterogeneity of results between studies examining this topic.

Tumour-specific prognostic value

Breast cancer

In breast cancer, the presence of TILs is associated with improved prognosis in human epidermal growth factor receptor 2 (HER2) positive and triple negative breast cancers (TNBC), but not in luminal subtypes. In addition, the recognition of the prognostic value of the immune infiltrate has been the basis for establishing a breast cancer immunological grade (Salgado et al, 2015).

Independent of other clinicopathological prognostic factors or chemotherapy regimens, multiple studies have confirmed stromal TILs are associated with higher rates of pathological complete response (pCR) to neoadjuvant chemotherapy in all subgroups evaluated (including ER positive, HER2-positive tumours) (Dushyanthen et al, 2015). However, these differences in response only appear to translate into improved longer term outcomes in non-luminal tumours.

A meta-analysis of 25 published studies comprising over 22 000 patients, failed to show that immune infiltrates are associated with overall survival (OS) in unselected breast cancer patients, but did find such an association in TNBC (hazard ratio (HR): 0.79; 95% confidence interval (CI): 0.71–0.87). CD8-expressing lymphocytes were associated with improved disease-free survival (DFS; HR: 0.69; 95% CI: 0.56–0.84) and breast cancer-specific survival (HR: 0.78; 95% CI: 0.71–0.86) in the overall population, whereas the FOXP3-expressing lymphocytes were associated with worse DFS (HR: 1.47; 95% CI: 1.06–2.05) and OS (HR: 1.50; 95% CI: 1.15–1.97, P=0.004) (Mao et al, 2016).

Clinical trials have not reported an association between TIL, nuclear grade or histopathological grade in TNBC with most making the assumption that TNBC are high grade (Adams et al, 2014). It remains uncertain whether this association may be explained partly by response to chemotherapy; lower grade luminal tumours have lesser response to cytotoxic therapy and are less frequently associated with infiltrating immune cells.

A Th1 immune phenotype and mRNA profiles consistent with immune activation have also been associated with response to neoadjuvant chemotherapy (Denkert et al, 2015). There is more variability in results seen in trials reporting outcome for CD4-expressing T lymphocytes and FOXP3-expressing Tregs. The presence of Tregs prior to chemotherapy is associated with higher probability of attaining a pathological complete response (pCR), which probably reflects their association with a higher number of CD8-expressing cells. A high ratio of CD8:FOXP3 cells and a lower proportion of FOXP3 at the end of neoadjuvant chemotherapy may have a more meaningful prognostic value (Dushyanthen et al, 2015).

The current working group have recommended semi-quantitative assessment of stromal TILs and at this stage do not advocate for sub-classification of lymphocytes (Salgado et al, 2015). This is due to both the greater reproducibility of stromal TIL measurement compared with intratumoural TILs, which are difficult to distinguish from malignant cells in standard H&E sections, and the fact that in TNBC and HER2-positive breast cancer, the prognostic power of TILs persists among all subtypes of infiltrating immune cells (Salgado et al, 2015).

Colorectal cancer

Several scoring systems have been proposed for quantifying the inflammatory response in colorectal cancer. These include the Jass score, the Immunoscore and the Klintrup–Mäkinen grade of overall peritumoural inflammation (Park et al, 2014). There is evidence that TILs are associated with greater prognostic value than the American Joint Committee on Cancer TNM stage (Jochems and Schlom, 2011). In a meta-analysis of nine trials examining tumour inflammation in colorectal cancer, the pooled HR confirmed an OS benefit for patients with prominent TILs compared with those without, with a HR of 0.59 (95% CI: 0.48–0.72, P<0.001) and a HR for cancer-specific survival of 0.40 (95% CI: 0.27–0.61, P<0.001). There were differences between all the studies in the thresholds used to determine TIL positivity of tumours, for example, some used mean or median cut offs, others used high vs low scores of Klintrup–Mäkinen or Jass scores (Mei et al, 2014). The evaluation of T cell subsets and specific location of lymphocytic infiltrate did not show strong prognostic value, specifically CD3, CD8, FOXP3 and at different sites (tumour centre, peritumoural stroma and invasive tumour margin) were examined. CD3-positive cells at the invasive margin had OR for DFS of 0.4 (95% CI: 0.35–0.68) and for OS of 0.63 (95% CI: 0.42–0.93). This analysis was limited by significant inter-study heterogeneity (Mei et al, 2014). This contrasts to earlier individual study data showing statistically significant association between the type of immune cell density at the centre of the tumour or the infiltrating margin and patient outcome (Jochems and Schlom, 2011).

Ovarian cancer

In a meta-analysis of 10 studies comprising 1815 patients with treated ovarian carcinoma (Hwang et al, 2012), presence of intra-epithelial T lymphocytes was associated with improved OS (pooled HR for death 0.45, 95% CI: 0.34–0.58, P<0.001). CD3- and CD8-expressing lymphocytes were both examined, and both conferred a survival advantage; CD8 was examined more frequently and demonstrated a larger magnitude of effect on OS than CD3 (pooled HR: 0.46 and 0.57, respectively) (Hwang et al, 2012).

This positive association between CD8-expressing lymphocytes and clinical outcome is also observed in the assessment of patients before treatment and following neoadjuvant chemotherapy. Data on CD3-expressing lymphocytes, B cells and NK cells are less clear (Santoiemma and Powell, 2015). There are conflicting data regarding FOXP3-positive Tregs, with a few studies demonstrating superior outcome, but most studies suggesting a negative impact on survival outcomes through inhibition of cytotoxic T cell activity (Santoiemma and Powell, 2015). The measured absolute number of infiltrating cells may not be as important as the proportion of CD8-expressing cells relative to all infiltrating cells. The prognostic value of intratumoural CD8-positive lymphocytes appears superior even to the adequacy of surgical debulking in prognosticating for both progression free survival and OS (Zhang, 2003).

Non-small cell lung cancer

In a meta-analysis of 29 trials with over 86 000 patients, high levels of CD8-expressing cells infiltrating the tumour or in the tumour stroma of non-small cell lung cancer (NSCLC) specimens were associated with better OS (HR: 0.76 and 0.80, respectively) compared with tumours without lymphocytes present. CD3 expression also demonstrated similar findings; pooled HR for OS 0.65 (95% CI: 0.50–0.84, P=0.001) for stromal CD3 cells and 0.66 (95% CI: 0.45–0.97, P=0.03) for intratumoural CD3 cells. Presence of intratumoural CD4-expressing cells between the tumour cells resulted in improved OS (HR: 0.65; 95% CI: 0.46–0.91, P=0.01). Despite a higher effect size, a significant association between stromal CD4-expressing cells and outcome was not observed (HR 0.43; 95% CI: 0.07–2.61, P=0.36), likely due to greater heterogeneity. FOXP3-expressing T cells in the tumour stroma had association with worse progression-free and OS (HR: 2.14; 95% CI: 1.68–2.72; P<0.001) and 2.67 (95% CI: 1.74–4.08; P<0.001, respectively) (Geng et al, 2015).


Checkpoint inhibitors were first approved in melanoma after a long history of interest in the immune response to these tumours after observation of spontaneous responses (Mihm and Mule, 2015). One histopathological definition of the immune response in melanoma categorised the immune infiltrating response as ‘brisk’, a scenario where lymphocytes are demonstrated in the entire tumour mass or along the advancing edge; ‘non-brisk’, where lymphocytes are seen focally in the centre of the tumour or along part of the invasive margin; or ‘absent’ with no tumoural lymphocytes at all or lymphocytes seen, but not interacting with melanoma cells. These subgroups provide prognostic information in historical studies. In one study, melanoma-specific death was 30 and 50% lower in the non-brisk and brisk groups, respectively, compared with the absent group (Mihm and Mule, 2015). In contrast, studies report no survival advantage with lymphocytic infiltrate particularly with respect to tumours of earlier stage and not in the radial growth phase (Ladanyi, 2015). However, overall, there is a large body of evidence documenting the prognostic value of the immune infiltrate in melanoma (see summary in Table 3).

Table 3 Studies examining the prognostic impact of infiltrating immune cells in melanoma

Renal cell carcinoma

There is contradictory evidence regarding the role of the immune cell infiltrate in renal cell carcinoma. Multiple studies have demonstrated a worse outcome in patients with a neutrophilic, and/or lymphocytic infiltrate (Jochems and Schlom, 2011), a finding which appears reproducible (Table 4). The reasons for this are not clear.

Table 4 Immune cells in renal cell carcinoma

Head and neck cancer

Several clinical trials have demonstrated that tumour infiltration by CD3- and CD8-expressing T cells correlates with improved disease outcome in chemoradiotherapy-treated patients with head and neck cancer. This positive prognosis holds true regardless of the human papilloma virus (HPV) DNA status (Balermpas et al, 2016a). Smoking-associated tumours with higher degrees of genomic instability and higher antigenicity would be expected to have increased potential to activate an immune response; however, this is not supported by clinical evidence. There is conflicting information regarding differences in the immune infiltrate in HPV-positive vs negative status (Wansom et al, 2012; Partlova et al, 2015); see Table 5.

Table 5 Studies of TILs in head and neck squamous cell carcinoma

Urothelial cancers

The approval of immunotherapy in the treatment of advanced urothelial malignancy suggests the relevance of the immune system. This is supported by most studies demonstrating the positive prognostic value of CD3, CD4 and CD8 T cells, and the negative association of FOXP3-positive T cells with survival, see Table 6.

Table 6 Studies examining prognostic impact of immune cells in bladder cancer

Hepatocellular carcinoma

Several studies have examined the role of the intratumoural and peritumoural (parenchymal) infiltrate in hepatocellular carcinoma (HCC) (Table 7). High levels of FOXP3 Tregs are associated with worse DFS and OS. Two large meta-analyses performed in 2014 demonstrate the importance of FOXP3 in both the development and prognosis of HCC (Huang et al, 2014; Zhao et al, 2014). Gabrielson et al, 2016 applied the Galon Immunoscore (Galon et al, 2014) to HCC and confirmed its prognostic value, CD3 and CD8 cell densities predicted recurrence with ORs of 5.8 (95% CI: 1.6–21.8) and 3.9 (95% CI: 1.1–14.2), respectively. PDL1 staining was positively correlated with high CD3 and CD8 density and predicted a lower rate of recurrence (Gabrielson et al, 2016). The applicability of these tools remains limited by routine access to technology to subtype these T cells.

Table 7 Studies examining prognostic value in HCC

Other tumour types

The prognostic role of the immune infiltrate in less common malignancies is summarised in the Online Appendix.

Brain metastases

Although the central nervous system (CNS) has been purported to be an ‘immune privileged’ site, there is an increasing evidence supporting the role of immune infiltrating cells in brain tumours. In a study by Harter et al, TILs in brain metastases from different tumour types were quantified and associated with outcome. This was then validated in a breast cancer only brain metastases cohort. Carcinomas demonstrated more frequent stromal infiltration, whereas TILs in melanoma were more often diffusely infiltrative. High TILs level, high-programed cell death protein (PD)1+/CD8+ and programed death ligand (PDL)-1 staining were associated with smaller tumours but there was no significant association with survival demonstrated (Harter et al, 2015). In contrast, Bienkowski and Preusser, 2015 provide a review of the literature in which they concluded that tumour-infiltrating lymphocyte density in CNS metastases were strongly associated with improved OS.


Broadly speaking, the immune infiltrate can be classified as a ‘pro-inflammatory’ phenotype with infiltrating T cells and a cytokine profile consistent with immune activation. Immune control of tumours can occur spontaneously, and the presence of an immune infiltrate is generally a good prognostic sign. However, the immune infiltrate has variable effect in prognostic models depending on the tumour type, location of the cells and state of activation; the complexity of immune networks are likely oversimplified in current measurement models. Tumour evasion through inhibitory mechanisms may serve as a predictive marker for benefit from immunotherapy, which inhibits negative regulators of the immune system. Alternatively, the microenvironment may lack immune cell infiltration, and tumour resistance is likely through immune system ignorance (Gajewski et al, 2013), and therefore promoting immune activation is less likely to be successful in this setting.

For tumour-infiltrating immune cells to live up to the ‘hype’ of inducing and promoting long-term tumour control and contribute as valuable prognostic markers, their subtype (especially activated antigen specific cytotoxic T lymphocytes) and position (organised spatial response) need to be defined and measured in a standardised manner. Successful inclusion of immune cell markers in prognostic clinical models is becoming a realistic hope in some cancers.