Anti-cancer immune responses start by the capture and processing of tumor-associated antigens (TAA) by antigen-presenting cells such as dendritic cells (DCs) ^[1][5]. TAAs can be varied in nature, and they range from, for example, viral proteins, overexpressed proteins, embryonic antigens and neoantigens. Hence, the nature of such TAAs can affect therapeutic activities and toxicities, as extensively revised in ^[1][5]. These TAAs are released by cancer cells dying either by immunogenic cell death or following attack by natural killer cells (NK) through the establishment of an initial inflammatory response. TAA-loaded DCs migrate to secondary lymphoid organs such as lymph nodes, where they prime both CD4 and CD8 T cells specific for these TAAs. As CD8 T cells possess strong cytotoxic activities, this T cell subset has been classically considered as the main effector in anti-tumor immunity through direct tumor killing ^{[2][3][4][5][6]}[51,52,53,54,55]. Following antigen presentation, CD8 T cells expand exponentially in peripheral blood and differentiate into cytotoxic T cells (CTLs). These primed T cells infiltrate tumors where they recognise cancer cells bearing TAAs and exert their cytotoxic activities ^{[4][7][8][9][10]}[7,16,53,56,57]. It is believed that this process of TAA recognition and initial immunological attack over tumors takes place right at the beginning. Thus, most cancers that progress end up evading this initial immunological attack by reducing their immunogenicity through immunological editing. This escape mechanism leads to the selection of poorly immunogenic cancer cell variants which down-modulate MHC molecules, or express immunosuppressive molecules ^{[11][12][13][14][15][16][17]}[58,59,60,61,62,63,64].

The CD4 T cell contribution to anti-cancer immunity has been much less studied. Even so, the evidence supporting their anti-tumor capacities is compelling, principally by regulating innate and adaptive immunity ^[18][19][65,66]. Indeed, their importance is highlighted by the study of immunoedited cancer cells, in which mutations in MHC-II-restricted neoantigens are more potently selected during tumorigenesis ^[20][21][22][67,68,69]. This reflects the significant contribution of CD4 T cells in immunosurveillance.

CD4 T cells differentiate into several subsets with different regulatory roles. This diversity of CD4 T cell subsets reflects the variety of immune responses they regulate ^[23][70]. CD4 T cells differentiate during antigen presentation towards different subtypes depending on the cytokine milieu. During antigen presentation, mostly at peripheral lymph nodes, antigen-presenting cells such as DCs present antigen peptides complexed to MHC-II molecules. These MHC-II-peptide complexes are recognised by naïve CD4 T cells. In this process, CD4 T cells receive three signals; the first through their TCR, the second the integration of positive and negative co-stimulation. The third one is called cytokine priming, and consists on stimulation by cytokines produced by antigen presenting cells and those present in the microenvironment. The specific cytokines determine the CD4 T helper (Th) subtype that will be activated. The major Th subtypes are Th1, Th2, Th17 and inducible regulatory T cells (Tregs).

Classically, the CD4 Th1 subtype is associated to anti-tumor immunity. Th1 cells act in concert with antigen presenting cells for CD8 T cell priming and differentiation towards CTLs in a process called T cell licensing ^[24][25][71,72]. During this process, Th1 cells and APCs produce pro-inflammatory cytokines such as IFNγ and IL-12 ^[26][27][28][73,74,75]. Th1 cells are also responsible for DC licensing through engagement of CD40L with CD40 on the DC surface ^[29][30][31][76,77,78]. During this process, DCs mature by up-regulating the co-stimulatory molecules CD80, CD86 and CD40L, together with concomitant IL-12 and IL-15 secretion for cytokine priming ^{[32][33][34][35][36][37]}[79,80,81,82,83,84]. Activated DCs prime naïve CD8 T cells towards CTLs and memory phenotypes. CD8 T cells acquire CTL effector functions through co-stimulation by CD27 with CD70 via CD40-CD40L signaling between APCs and CD8 T cells ^[18][38][39][65,85,86]. Indeed, CD8 T cell priming in the absence of CD4 T fails to fully activate CTLs leading to limited expansion of anergic CD8 T cells with dysfunctional phenotypes, and lack of CD8 T cell memory ^{[18][25][40][41][42]}[65,72,87,88,89]. The activity of CD4 Th1 cells potentiates anti-tumor responses from NK and M1-type macrophages, which further promotes tumor killing causing the release of more TAAs for T cell priming ^[43][44][90,91]. CD4 T helper cells can also differentiate into Th2 and Th17 subtypes, characterized by expression of cytokines such as IL10 and IL4 for the former, and IL-6, IFN-γ and IL-17 for the latter ^{[45][46][47][48][49]}[92,93,94,95,96]. These T helper subtypes are generally associated with tumor progression, although this may be context-specific. For example, CD4 Th2 cells are required for long-term memory responses, while Th17 cells can induce potent inflammation that can amplify anti-tumor immunity ^{[50][51][52][53][54][55]}[97,98,99,100,101,102]. Regulatory CD4 Tregs consist of at least three main subtypes, natural Tregs, inducible Tregs, and Tr1 CD4 T cells that are involved in maintaining central and peripheral tolerance. These CD4 T cells present potent immunosuppressive activities by several mechanisms, ranging from cell-to-cell contacts and expression of immunosuppressive cytokines ^{[56][57][58][59][60][61][62][63]}[103,104,105,106,107,108,109,110]. In certain conditions, CD4 T cells can also acquire direct cytotoxicity through production of IFN-γ and TNF-α, expression of FasL and TRIAL, and cytotoxic granules ^[64][65][66][111,112,113].

2. CD4 T Cell Differentiation Phenotypes according to Effector Functions

Upon antigen presentation, both CD8 and CD4 T cells expand exponentially and differentiate into effector phenotypes, including CTLs and the different subsets of T helper cells. These T cells are short-lived, but a small pool survives as long-lived memory subsets after antigen clearance. These memory T cells can last decades and are critical for recall responses. Memory T cells need less requirements for activation and mobilization following antigen presentation compared to naïve T cells. Following antigen re-encounter, memory T cells undergo fast activation and expansion, leading to stronger effector T cells ^{[67][68][69][70]}[114,115,116,117]. Different T cell differentiation phenotypes can be readily distinguished in humans by evaluating the expression profiles of CD62L and CD45RA. CD62L+ CD45RA+ naïve T cells migrate out of the thymus towards secondary lymphoid organs by the expression of CD62L ^[70][117]. Memory T cells can be differentiated into two main types based on their location and migration patterns: Memory T cells residing in secondary lymphoid organs are represented by central memory subsets, while those migrating into sites of inflammation are termed effector memory T cells. As such, memory T cells lose CD45RA expression, which allows these subsets to move between secondary lymphoid organs. Effector memory T cells further lose CD62L expression, as these cells will remain tissue-resident and upregulate chemokine and cytokine receptors required for chemotaxis towards sites of inflammation. From the effector memory pool, T cells can then re-express CD45RA (effector memory cells that re-express RA, or EMRA), which is also a marker of terminal differentiation. EMRA cells end up accumulating during the lifetime of the individual ^[71][118]. In addition to this classification, human T cell differentiation can also be studied based on CD27/CD28 expression profiles. Thus, poorly differentiated T cells which includes naïve and central memory phenotypes co-express both markers. Then, T cells progressively lose first CD27 expression, and then CD28, leading to highly differentiated T cells which englobe effector memory and EMRA T cells ^{[72][73][74][75][76]}[29,119,120,121,122]. CD27- and CD28- CD4 T cells are considered senescent T cells in humans.

3. Systemic CD4 Immunity as a Biomarker of Response to PD-L1/PD-1 Blockade Immunotherapy

Biomarkers of response to immunotherapies have typically been looked for within the tumor, or by evaluating tumor infiltration by immune cells. This is also the case for PD-L1/PD-1 blockade. For the latter case, the assessment of PD-L1 expression in tumor biopsies is the most well-established. Nevertheless, other markers including neoantigen expression, mutational status, DNA repair proteins and infiltration with TILs and immunosuppressive cells have been used. PD-L1 expression in tumors is the most extensively used, but its reliability could be a function of the tumor type, and also of the specific PD-L1/PD-1 blockade drug used ^{[77][78][79][80][81]}[18,125,167,168,169]. Tumor mutational burden and transcriptomic analyses of TILs can also have predictive value in patients undergoing PD-L1/PD-1 blockade ^{[82][83][84][85]}[170,171,172,173]. It is evident that a major drawback in implementing these biomarkers in clinical practice is, apart from their price, the unavailability of sufficient material from tumor biopsies in many cases. A second issue is whether a sample represents the tumor heterogeneity found in patients. Indeed, PD-L1/PD-1 blockade is carried out systemically and will have a broad impact in the immune system. These global effects will probably contribute to clinical responses, and systemic immune biomarkers could be an inexpensive way forward for their clinical use ^{[10][72][86][87][88][89][90][91][92]}[19,29,45,48,50,57,144,174,175]. Therefore, immune profiling of immune cells in peripheral blood samples could be a promising non-invasive procedure to evaluate biomarkers of response in immunotherapies. Profiling studies in combination with tumor sampling could improve patient stratification for immunotherapies. For example, early expansion of PD-1+ CD8 T cells in peripheral blood by flow cytometry correlated with clinical efficacy, which was confirmed in patients with thymic epithelial tumors ^[10][90][57,144]. In the first study, T cell expansion was evaluated by assessing proliferation through the expression of Ki67 after the first week of treatment. The predictive value of this functional immunological biomarker was validated in two independent cohorts of NSCLC patients. Similarly, genome-wide sequencing of peripheral T cell populations uncovered early clonal expansion that was associated with clinical responses in NSCLC patients undergoing PD-1 blockade ^[93][176]. This phenomenon seems to be common to many tumor types, for example in metastatic melanoma patients treated with PD-1 blockers ^[94][146]. High-dimensional techniques for the analysis of multiple populations can increase the accuracy for the identification of cellular biomarkers of response. For example, the identification of a CCR7-CD27-CD8 T cell subset that expanded systemically after PD-1 blockade in melanoma patients ^[95][147]. All these data clearly indicate that quantification of proliferating CD8 T cell populations following PD-L1/PD-1 blockade may be suitable biomarkers of response, or at least as a biomarker for real-time monitoring of responses in patients using a non-invasive technique. These biomarkers have the advantage of helping the clinician in decision-making during the early onset of immunotherapies. However, they may not be useful for the early identification of hyperprogressors ^[96][31]. Therefore, quantification of CD8 T cell subsets before the start of immunotherapies may not be sufficient for the identification of responders and non-responders. Other high-dimensional techniques with the power of identifying multiple immune cell populations have been applied for the identification of predictive biomarkers. For example, mass cytometry (CYTOF) which has been applied for the analysis of peripheral blood populations in metastatic melanoma patients undergoing PD-1 blockade before treatment initiation. One pioneering study showed that elevated CD14+ CD16- HLA-DR^high monocytes before applying immunotherapy correlated with a significant increase in progression free survival (PFS) ^[97][177].  The application of multi-parametric flow cytometry highlighted the elevation of PD-1, PD-L1 and PD-L1 in T cells as biomarkers of worse survival ^[98][179]. However, this might not apply to myeloid cell populations, in which elevation of PD-L1 correlate with efficacy of PD-L1 blockade with atezolizumab in NSCLC patients with PD-L1-negative tumors ^[86][19]. Nevertheless, none of the above immune cell-based biomarkers have been fully validated in prospective studies. In addition, some of the high-dimensional technologies such as CYTOF are difficult to standardize and apply into routine clinical practice. Work from independent research groups is demonstrating the potential of quantifying CD4 T cell subsets in peripheral blood to predict the efficacy of PD-L1/PD-1 blockade immunotherapies. In a small-scale study in metastatic melanoma patients, elevation of central memory T cells was associated with prolonged survival ^[99][180]. A higher ratio of central memory versus effector T cell subsets correlated with benefit from PD-1 blockade in a small cohort of NSCLC patients ^[100][181]. Two prospective studies, independently showed that evaluating the dynamics of CD4 T cell populations in peripheral blood can predict clinical outcome in patients treated with PD-L1/PD-1 blockade immunotherapy ^{[72][88][89][92]}[29,48,50,175]. Both studies independently found similar quantitative thresholds for evaluation of CD4 memory T cells in peripheral blood before the start of immunotherapies. CD4 T cells exhibited a CD27-CD28^low phenotype, which included both enriched central and effector memory CD4 T cell subsets ^[72][29]. Response rates of about 50% were observed in patients with more than 40% of the T cells with this biomarker phenotype. All patients with percentages below this threshold were non-responders and with a significant increase in risk for developing hyperprogression ^[72][101][29,30]. Kagamu and collaborators independently found a population of CD4 T cells (CD62L^low effector memory subset) with similar threshold values ^[88][89][48,50]. Elevation of this subset and reduction in Tregs were significantly represented in responders to PD-1 blockade immunotherapy. A high ratio between effector memory CD4 T cells and Treg cells could identify responders from progressors. Other immunological biomarkers have also been studied in many cancer types and for a variety of treatments, including immunotherapies and PD-L1/PD-1 blockade. These include absolute and relative neutrophil and lymphocyte numbers, and the classical neutrophil-to-lymphocyte ratio (NLR), amongst others ^{[102][103][104][105]}[182,183,184,185]. These markers have been shown to have prognostic rather than predictive power. This is a subtle difference, but whereas a prognostic marker is treatment-independent, a truly predictive biomarker is one specifically associated with a particular therapy. However, the quantification of these biomarkers is not standardized, and their practical value might be limited ^[87][45]. Modifications to the classical NLR have been introduced to incorporate other immune populations, especially other myeloid cells apart from neutrophils. For example, the derived NLR (dNLR) that together with quantification of lactate dehydrogenase is used for the calculation of the lung immune prognostic index for PD-L1/PD-1 blockade ^[106][186]. However, all these classical prognostic markers are based on standard clinical blood analyses that do not discriminate the variety of immune cell types and their activation status in peripheral blood ^[87][45]. To achieve this, high-dimensional techniques, such as multiparametric flow cytometry, can provide more accurate quantification of different cell types, leading to better correlations with clinical outcomes. Using these techniques, the prognostic value of cell populations such as monocytes, neutrophils and other granulocytes have been confirmed in NSCLC patients undergoing PD-L1/PD-1 blockade immunotherapies ^[87][91][45,174].