Molecular prediction of clinical response to anti-PD-1/ anti-PD-L1 immune checkpoint inhibitors: New perspectives for precision medicine and mass spectrometry-based investigations

Monoclonal antibodies (mAbs) acting as immune checkpoint inhibitors (ICIs) are among the most frequently used immunotherapies in oncology. However, precision medicine approaches to adapt the treatment to the patient are still poorly exploited. Given the risk of severe adverse reactions, predicting patient eligibility for ICI therapy represents a great asset for precision medicine. Today, the extended panel of mass spectrometric approaches, accompanied by newly developed sample preparation methods is a strategy of choice for responder and non-responder stratification on a molecular basis, and early detection of resistance. In this perspective article, we review the biodisposition of mAbs, the interest in molecular stratification of patients treated with these mAbs, and the possible analytical strategies to achieve this goal, with a major emphasis on mass spectrometric approaches.


| INTRODUCTION
Immune checkpoints (IC), for example, cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), programmed cell death protein 1 (PD-1) or PD-1 ligand 1 (PD-L1)) are surface proteins of immune cells that regulate the antitumoral immune-response. These ICs can be expressed with or without molecular modifications by the tumor cells (eg, PD-L1 protein) to inhibit the immune response ( Figure 1A). Current IC inhibitors (ICIs) are therapeutic monoclonal antibodies (mAbs) blocking the interactions between the IC on the target cells and their ligands ( Figure 1B). ICIs, such as the ones targeting PD-1 or PD-L1, have revolutionized antineoplastic immunotherapy 1,2 by improving overall survival (OS) and progression-free survival (PFS) for patients treated for various types of malignancies. 3 However, despite these achievements, only a subset of patients benefits from these moderately welltolerated therapies, 3 while others exhibit either primary or acquired resistance. Currently, the most important challenge for many anti-PD-1 or anti-PD-L1 therapies is to identify nonresponder patients in order to give the most adapted therapy as early as possible. This would allow avoiding exposure to drugs causing potentially severe or serious adverse reactions 4 in patients who will not benefit from these potentially harmful compounds. In that regard, predictive biomarkers could help anticipating the occurrence of resistance during the treatment course and thus foster rapid treatment adjustment.
Recently, three predictive biomarkers were approved by the US Food and Drug Administration (FDA) for patient selection for ICI therapies in clinical practice 5 : (i) the percentage of PD-L1+ cells in tissue, expressed as tumor, immune or combined proportion scores, (ii) the presence of high microsatellite instabilities (MSI-H) and (iii) tumor mutational burden, as extensively described in Supplementary information 1 (SI-1) section S1. However, for many patients, these biomarkers do not reliably predict outcome (eg, patients lacking PD-L1 expression on tumor cells may still respond to ICI treatment) 6 and exhibit some limitations given the high diversity and related cut-offs of available tests depending on the tumor type. 3,5 Many clinical studies integrating medical imaging (eg, tumor size evaluation), molecular biology investigations (eg, RNA-sequencing, tumor genome analyses, phenotyping), molecular histology (eg, abundance of PD-L1+ tumor cells or specific immune cells in the tumor microenvironment [TME] 7 ), or host-related characteristics 3,6,8 thus aimed to find further predictive biomarkers of treatment response. This highlights the current necessity to collect and interpret data more comprehensively.
Mass spectrometry (MS) is a powerful and versatile analytical technique enabling comprehensive analysis of complex biological matrices 9 (eg, proteomic analyses), as well as accurate and specific quantification of a wide range of compounds (eg, from small molecule drugs to therapeutic mAbs and related compounds). However, only few studies have been conducted to apply MS techniques to discover predictive biomarkers for response to ICI therapy. This likely results from a limited access to a sufficient amount of meaningful biological samples (eg, tissue or cells) given the invasive nature of sample collection. Furthermore, these mostly focused on either plasma proteomics or targeted detection of ICs in tissues. 10 This perspective article reviews the disposition of the anti-PD-1 and anti-PD-L1 therapeutic antibodies to understand how exposureresponse (E-R) relationship could be better evaluated, as well as the potential predictive markers for precision medicine that could be investigated using MS strategies. The possible methods for the retrieval of ICIs and surrogate markers from the different compartments of interest will be evaluated, as well as the different mass spectrometric strategies and methods for analysis. The clinical and therapeutic context of the different pathologies treated with ICIs will be discussed with respect to feasibility of sample collection and downstream analyses. The present manuscript is accompanied by an important document "Supplementary information 1" (SI-1) that supports statements with more extensive explanations and illustrations.  IMMUNE CHECKPOINT INHIBITOR  THERAPIES   When establishing the E-R relationship, drug exposure is sometimes   equated with the administered dose (first critical pharmacological   parameter, 11 SI-1 Table S1.1), but most often measured as free (second critical pharmacological parameter, 11 SI-1 Table S1.2, Figure 2, 1a) or total drug concentration in plasma or serum (depending on the analytical strategy used 12 ). In this case, exposure is then defined as area under the curve of the concentration-time curve or average concentration (c avg ), or as trough (minimal, c min ) or peak (maximal, c max ) concentration of the first or later treatment cycles (pharmacokinetics [PK]). Different approaches have been used to characterize the E-R relationship in ICI therapy (ie, exposure-efficacy or exposure-safety relationships), depending on the type of ICI (eg, anti-CTLA-4 or anti-PD-1) or even depending on the tumor type or clinical trial. 2,[13][14][15][16] Regarding anti-PD-1 and anti-PD-L1 antibodies (eg, pembrolizumab, nivolumab, atezolizumab), most published trials did not find an obvious association of efficacy (eg, OS, PFS, overall response [OR]) or safety (eg, immune-related adverse events or adverse events leading to treatment discontinuation or death) with either the administered dose or the drug concentration in blood. 2,[13][14][15][16][17][18][19] Only rare studies show contradictory results where an E-R relationship for ICI efficacy was found, but these often neglected several key clinical and biological parameters (eg, tumor burden or mAb clearance) in the multivariate F I G U R E 2 Disposition of monoclonal antibodies (mAb) in the intravascular compartment (eg, blood) and tumor microenvironment (ie, interstitial compartment) during treatment with an anti-PD-1 immune checkpoint inhibitor (ICI), highlighting factors influencing ICI exposure and clearance (1-8) and potential predictive biomarkers of treatment efficacy (gray inserts) analyses. 20  before further research is initiated.

| Distribution of immune checkpoint inhibitors
ICIs are administered intravenously (IV) and further diffuse from the blood compartment into interstitial compartments 21,22 to reach the cellular site of action (SOA) (third critical pharmacological parameter, 11 SI-1 Table S1.3) where the molecular SOA (ie, PD-1 or PD-L1) needs to be present to display the therapeutic action (SI-1 Table S1.6) via noncovalent binding of the mAb (SI-1 Table S1.7). 11 Given their macromolecular structure, distribution of mAbs is limited (volume of distribution, V ss = 3-8 L), and is mainly mediated by extravasation through convective transport from the blood compartment to the interstitial compartment (ie, tissue space or TME in the case of solid tumors). 21,22 For mAbs, these compartments appear of interest for investigations of treatment response prediction and monitoring of either free or target-bound mAb fractions.

| Mechanisms and sites of action of immune checkpoint inhibitors
For an effective immune response, T cells need to be activated through the recognition of the tumor cells via T cell receptor binding to major histocompatibility complex antigen (MHC) antigen on these latter, but also, and perhaps mainly, through T cell co-stimulation via the interaction of the cluster of differentiation 28 (CD28) with CD80 and CD86 (also called B7-1 and B7-2, respectively, expressed by antigen-presenting cells). 23 Activation of the PD-1/PD-L1 pathway results in the inactivation of the T cell, even when the co-stimulatory pathways are active ( Figure 1A), which can be reversed by blocking the PD-1/PD-L1 interaction with mAbs ( Figure 1B).
The intended anatomical, cellular, and molecular SOA 11 vary with the mode of action of the ICI. With ICIs targeting PD-L1 (eg, atezolizumab), the direct cellular SOA are PD-L1+ cells, such as tumor cells or host immune cells. 24 These can be localized more or less distant from the blood compartment, depending on whether they circulate in the blood or are embedded in the tumor tissue. In turn, PD-1-expressing T cells are the indirect or secondary cellular SOA of anti-PD-L1 therapies, since they aim at blocking PD-1/PD-L1 pathway.
With ICIs targeting PD-1 (eg, nivolumab or pembrolizumab), the direct cellular SOA are the T cells that can circulate in the blood or are embedded in the anatomical SOA, whereas the indirect cellular SOA are the PD-L1+ cells (ie, tumor cells or host immune cells).
Drug exposure at the cellular SOA may depend on the compartment in which the direct and indirect SOA are located (ie, lymphatic system, blood or TME). For instance, regarding anti-PD-1 ICIs, T cells can be found in the lymphatic system, the peripheral blood, and the TME, but the anatomical SOA is rather believed to be in the tumor tissue. 23,25 Moreover, recent findings suggest that the actual sites of T cell priming (ie, lymph nodes) can also importantly modulate the efficacy of anti-PD-1 therapy. In fact, the response to PD-1 blockade appears to depend on the expression of CD28 on CD8+ T cells, and anti-PD-1 therapy even appears to trigger the release of CD28+-CD8+ T cells in the peripheral blood. 23,26 Localization and mechanisms of ICI action remain difficult to fully investigate and understand, and this lack of complete understanding of all involved anatomical, cellular, and molecular SOA (direct or indirect) will need to be considered when searching for predictive biomarkers of ICI efficacy. Off-site binding could also occur between the ICI and its molecular SOA solubilized in plasma (soluble target, for example, sPD-1 or sPD-L1) (SI-1 Table S1.8, Figure 2, 4). For sPD-L1 for instance, it was previously shown that some soluble forms retain its immunosuppressive function when binding to PD-1+ T cells. 28,29 In such case, blockage of sPD-L1 could foster treatment efficacy. However, the biological activity of other forms of sPD-L1 and of sPD-1 is less well understood, and, if inactive, these forms could reduce therapeutic efficacy by consuming the available fraction of administered mAb. 28,30 Another important off-site factor is the presence of PD-L1+

| Interfering events during the disposition of immune checkpoint inhibitors
tumor-derived exosomes (TEXs) in the circulation. PD-L1+ TEXs could regulate PD-L1 expression on tumor cells and also on immune cells, and inhibit the immune response, [31][32][33][34] and they seem to contribute to the immunosuppression in patients with various cancers to a greater extent than sPD-L1. 28,29 Anti-PD-1 ICIs can also reverse the immunosuppressive effect of PD-L1+ TEXs by blocking their binding to PD-1+ T cells and a similarly efficient blockade might be achieved when using anti-PD-L1 ICIs. 33 Additionally, T-cell-derived PD-1+ exosomes might also predict ICI efficacy. 35

| Antibody degradation and aggregation
In vivo degradation of mAbs through fragmentation or aggregation in the circulation or tissues (Figure 2, 7) can also limit efficacy and safety (eg, loss of binding structure, increased immunogenicity). 36 During the early drug development phases, degradation and aggregation of mAbs are usually investigated in buffer. However, these modifications should also be monitored in blood (ie, serum or plasma) during the early drug development phases and candidate mAbs exhibiting significant fragmentation or aggregation in the blood should not be selected for further development. 36 In the later stages of clinical development, such phenomena would then be minor and should not have a significant impact on the efficacy of the treatment.
Additionally, mAbs are often subject to posttranslational modifications (PTMs) during mAb production and storage, drug administration, and in vivo (eg, deamidation, isomerization, oxidation or glycation). 36 Such modifications can increase the risk of immunogenicity and considerably reduce treatment efficacy. Therefore, it is important to control these critical quality attributes, also called product quality attributes. 37,38 But even if the manufacturing-and storage-related PTMs are closely monitored and controlled, it remains difficult to anticipate which PTMs will additionally occur in vivo and to which extent this modifies PK properties and efficacy of a mAb (Figure 2, 8  parameters that could be accessed to predict the action of drugs (Table 1) can be grouped as follows: (i) abundance, localization and molecular variants of ICs; (ii) endogenous protein biomarkers in blood or in TME; and (iii) effective exposure to therapeutic mAb in blood or in the TME. To exemplify the retrieval of molecular information related to these areas of interest, an anti-PD-1 ICI will be used as illustrative example to better define the different molecular parameters (Table 1 and Molecular measurements prior to ICI administration would serve as baseline levels for predictive biomarker discovery, while molecular measurements during and after treatment could help detecting significant changes in protein content that could be correlated with clinical outcome or acquired resistance (Figure 3, Q12, Q13, Q14 and Q15).

| Immune checkpoints as predictive biomarkers
Depending on the expression level or the abundance of their cellular and molecular SOA (direct or indirect, Table 1

Peripheral blood mononuclear cells (PBMC)
Immune cells circulating in blood.

Plasma
Plasma proteins including shed extracellular domain of direct and indirect molecular SOAs, i.e., soluble immune checkpoints (sSOAs).

Q14 -Changes in P4?
Potential predictive biomarkers of therapy efficacy. To be investigated by analyzing biological samples before therapy by correlation with clinical data for patient stratification.
Monitoring of exposure to the ICI. To be investigated with analysis of accessible biological samples taken during therapy.

Potential markers of resistance acquisition.
To be investigated by comparison of analytical data from accessible biological samples taken before, during and after therapy.
Q12a -Differential amounts of P2a in TME before and after therapy? Q13a -Differential amounts of P3a in TME before and after therapy? Q15a -Changes in P5 and P6 in TME after therapy?

Q16 -Changes in P4
before and after therapy ?
P2-P3 Interaction level (P4b) F I G U R E 3 Timeline for immune checkpoint inhibitor (ICI) therapy from cancer diagnosis based on the example of nonsmall cell lung cancer patients treated with an anti-PD-1 monoclonal antibody. Related key molecular parameters (PXx, Table 1) and questions (QXx) to be answered during biomarker discovery and pharmacokinetic parameter evaluation molecular state of the intended cellular and molecular SOA and related ligands (Figure 3, Q1d).

| Localization
The localization of PD-1+ and PD-L1+ cells might be of importance in the search for predictive biomarkers. 45 In addition, the spatial organization of PD-1+ T cells within the tumor (tumor-infiltrating lymphocytes) seems to play an important role for the prediction of ICI action. Cold tumors (ie, tumors with low immune infiltrates) seem to not respond to ICIs contrarily to hot tumors (ie, tumor with high immune infiltrates). 46,47 Spatial information through histology coupled to MS-based "spatialomics" 48 may then be highly valuable for patient stratification.

| Forms of immune checkpoints
Not only ICs on cell surface, but also soluble (ie, sPD-1, sPD-Ls) or exosomal ICs (eg, exoPD-L1) could influence the efficacy of ICI treatments and might be used as predictive efficacy or resistance biomarkers. 28,29,49 Their validation as circulating biomarkers would ease the detection of acquired resistance with noninvasive sample collection (ie, blood collection) and could be a precious support for early dose adjustments on the basis of therapeutic drug monitoring (TDM).

| Endogenous proteins as predictive biomarkers
Comprehensive protein assays from blood or tissues (SI-1, section S3.2) can help discriminating protein signatures of potential responders from nonresponders at baseline and during therapy. In general, protein analyses would inform about the TME or the circulating molecular environments of patients before, during and after therapy.

Receptor occupancies
Knowing IC abundances and the amount of mAb bound to its target will give the intratumoral and peripheral receptor occupancies (ITRO and PRO). 2,14,[59][60][61] Because efficacy and related variations are expected to be closely related to the level of mAb bound to its molecular target at its cellular SOA, the ITRO should be one of the first investigated markers (Figure 3, Q10a), along with the PRO (Figure 3, Q10b). In tumor tissues, the chances to detect the drug after the last dose would depend on the half-life of the studied mAb (usually about 20 days) and on the time between last dose and biopsy, but tissue sample retrieval is rarely possible throughout the therapy or even shortly after the end of a treatment course. In such cases, it would be of interest to assess whether the PRO could act as a surrogate for the ITRO, which would enable a more regular exposure monitoring.

| Monitoring off-site interactions and modified forms of the therapeutic monoclonal antibody
Ultimately, detected changes of one of the described modulators (see in opposition to the mAb fraction bound to other molecular or cellular targets in blood and/or in the TME (eg, binding to sPD-1 [ Figure 3, Q11], to exoPD-L1, or to human anti-human antibody [HAHA]). Comprehensive molecular analysis would also help detecting molecular variants of the ICI and support the monitoring of nonfunctional mAb variants.
The panel of monitoring options for ICI exposure and predictive biomarker discovery is large. Ideally, the comparison of accessible molecular data between tissue and blood would inform if the circulation contains good surrogates to the biomarkers found in the TME (eg, in tissue) and would help assessing the usefulness of plasma for TDM. In this context, if putative biomarker levels are found similar or related between blood and TME, plasma, serum or PBMCs could be used instead of tissue for further prediction of resistances to ICI therapy (Figure 3, Q4). This would allow for more frequent monitoring of molecular evolution between samples at regular time points throughout therapy and the clear determination of molecular markers of resistance.

| SAMPLE RETRIEVAL
For sample retrieval, it appears crucial to target the right anatomical SOA. In theory, cellular targets expressing the molecular SOA of anti-PD-1 and anti-PD-L1 can be found in lymph nodes, in the blood, and in the TME. However, even when some biological actions are triggered in the target compartments, it does not necessarily mean that the distinct expression level of the related protein is the decisive indicator for the choice of therapy. For instance, when sampling in the anatomical SOA is not occurring systematically before or during therapy, it will be difficult to base the choice of therapy on biomarkers from this anatomical region. Additionally, focusing on one compartment only (eg, tissue or TME) might neglect other critical pharmacological parameters (SI-1, Table S1). Therefore, the search for putative markers of future clinical outcome or of acquired resistance should take all the compartments into account and study to which extent the selected markers are meaningful and from which compartment. However, while blood is easily accessible, other anatomical targets can be difficult to reach 11 and some promising putative markers could therefore be difficult to apply and validate in terms of technical and/or ethical feasibility.

| Plasma and peripheral blood mononuclear cell collection
In clinical trials, blood is the most accessible compartment when collecting biological samples, explaining why predictive biomarker discovery usually relies on plasma or serum. 10

| Tissue collection
Depending on the targeted cancer and patient parameters (eg, stage of disease), tissue or biopsy collection is mostly performed for diagnostic purposes. In certain trials, further tissue samples may be taken after disease progression to decide on the next lines of therapy, but in general, the collection of tissue or biopsies remains limited, depending on the accessibility of the targeted anatomical site (eg, tumor developing on skin surface as opposed to tumor or tumor metastases developing in an organ). Considering the time points for tissue collection will help to determine the chances to find the molecular parameter to assess (Table 1, Figure 3), and in many cases, it might only be possible to study predictive biomarkers in samples collected at baseline. Moreover, even when tissue can be sampled from the anatomical SOA, the size of the biological sample can be very small. Therefore, attention must be paid to adapt the analytical strategies to sample size in order to get as much molecular information as possible.  [69][70][71][72][73] Thus, the next section will focus on this approach.

| Comprehensive or targeted protein analysis with MS-based proteomics
MS-based approaches can be performed as (i) unbiased proteomic analyses with relative quantification (ie, differential quantification by comparing different subgroups of patients) or (ii) targeted proteomic analyses with the possibility to get absolute quantification of defined protein targets. 74

| Comprehensive proteomics
Proteomics can be applied to different biological matrices including tissue, cells, and plasma, and will help to comprehensively identify proteins from the different patients of the cohort using a bottom-up strategy 75 (ie, sample proteins are first digested and the obtained peptides are monitored using LC-MS/MS to identify and quantify the intact proteins). When comparing proteomic data to clinical data including disease outcome, it becomes possible to identify key proteins or patterns that discriminate responders from nonresponders, during therapy or already at baseline. 10 In this context, relative quantification can be achieved by comparing key proteins or different groups of samples (eg, samples from responders and nonresponders).
It is noteworthy that while it is challenging to obtain absolute quantification using untargeted proteomic assays, the unbiased character of such a comprehensive approach also allows detecting modified forms of the key compounds (eg, variants of therapeutic mAbs or related ICs), which could also bring precious insight for the discovery of biomarkers to predict responses. Therefore, depending on the questions (Figure 3), the choice of analytical strategy is crucial in order to establish the key molecular parameters.

| Targeted proteomics
Besides the comprehensive untargeted analytical approach, it is also possible to develop targeted quantitative proteomic assays to assess the concentration of specified key proteins within a sample by monitoring specific fragments using LC-MS/MS in multiple reaction monitoring (MRM) or parallel reaction monitoring (PRM) modes. 74 For instance, dedicated methods exist for the absolute quantification of mAbs in biological samples (eg, plasma) that rely on affinity purification of mAbs (immunoglobulin G-like compounds) using protein A or protein G resin, and their enzymatic digestion before the quantification of proteotypic peptides using LC-MS/MS in MRM mode. 76 Depending on the analytical platform settings (eg, type of mass analyzer-triple quadrupole or Orbitrap for instance, applied LC gradient and data acquisition settings) and the applied sample preparation workflow (eg, enrichment step, affinity selection of specific class of compounds), it is then possible to quantify one or multiple proteins (ie, absolute or relative quantification). In theory, this would allow monitoring the administrated ICI, the direct and indirect molecular targets and possible additional protein markers in one analysis.

| Tissue proteomics
Relevance of MS assays for the discovery of predictive biomarkers Protein analysis is already extensively applied for diagnostics and biomarker detection in tissue samples from cancer patients using molecular biology or molecular histology (eg, IHC, phenotyping). 5,7,[77][78][79] Mass spectrometry-based biomarker discovery and validation are challenging to reach in a short term, because developments for MS-based proteomics of tumor tissue samples requires time and the availability of numerous relevant tissue samples, which are generally used for already applied diagnostics and predictive tests. Therefore, the clinical relevance compared to existing biomarkers and analyses first needs to be evaluated.
However, MS-based proteomics can potentially give more comprehensive information on the TME and merge the different markers and key proteins (eg, ICI, PD-1, PD-Ls) in one analysis. MS would also enable the comprehensive analysis of very limited sample amounts by using proteomic methods for small sample volumes ("microproteomics" 80 81,82 This could allow to overcome one major challenge in such development, that is, reaching a good enough sensitivity, allowing the quantification of the given ICI or related molecular SOAs. Finally, for extremely low densities of targeted cells in tissues (eg, infiltrating T cells), single-cell proteomics methods may have to be considered. 80 Finally, the localization of ICIs, ICs, and proteomic markers identified by microproteomics could be confirmed by high resolution spatial proteomics. In that regards, desorption/ionization (DI) methods hold good promise for mass spectrometric imaging ICs and proteomic markers, especially with the emergence of photocleavable mass-tag antibodies for the multiplexed immunohistochemical MALDI imaging of tissue biomarkers. 48 In addition to the identification of "cold" or "hot" tumors (SI-1, S3.1.2), cell identity and associated molecular parameters (eg, ICs abundance, ratio ICI/ICs) could be determined.

| Plasma and peripheral blood mononuclear cell proteomics
Relevance of MS assays for the discovery of predictive biomarkers Using blood samples, it is possible (i) to have access to higher sample volumes (in the 1-10 mL range), thus enabling the use of parallel analyses with different strategies from one sample, (ii) to determine which sample type is be the most relevant (eg, serum, plasma, PBMC fraction) and (iii) to monitor putative biomarkers more regularly throughout the therapy. Similar to tissues, MS analyses enable the use of multiplexed assays and a more comprehensive recording of molecular data. For short-term developments, blood samples are then ideal to discover circulating biomarkers of ICI efficacy. On a longer term, efforts should be made to establish whether specific contents of blood (eg, PBMCs, CTCs, TEXs) can be used as surrogates for the remote detection of TME biomarkers. Additionally, it should also be clarified whether biomarkers should be evaluated in plasma or serum. 10,83 Molecular parameters to investigate and related technical developments First, PBMC and plasma should be used to perform targeted mAb quantification ( Figure 3, Q9b and Q9c).
For PBMC proteomic analyses, CD proteins appear as promising candidates to monitor the cells that were determined as relevant for the drug action. For instance, the expression of cell-type specific CDs (eg, CD28+CD8) could be normalized to CD45, expressed on all common immune cells. 23,26,84 The related ratios could subsequently be linked to molecular SOA amounts (Figure 3, Q1c and Q2c), as well as to ICI levels.
Overall, the comprehensive proteomic analyses (Figure 3 Finally, after choosing the appropriate extraction strategy (eg, ultracentrifugation, steric exclusion chromatography), 85 it would also be possible to extract TEXs from plasma samples, and complete the molecular data from plasma and PBMCs.

| CONCLUSION
ICIs have revolutionized the vision of targeted therapies using biologicals, and extensive research is conducted to further improve these immunotherapies by going toward precision medicine. However, although some decisional biomarkers (eg, PD-L1 expression levels) are already applied to support the choice of ICI therapy, these do not necessarily associate with efficacy or resistance. Consequently, innovative predictive biomarkers would help to early identify responders or patients at risk of therapy failure (eg, resistance to ICIs). Moreover, this could support the choice of second-line therapies or of the most beneficial time of administration. In this perspective article, current aspects of ICI pharmacology were reviewed and MS-based analytical approaches were suggested to discover additional biomarkers for the prediction of efficacy and resistance to ICI therapy. This aims to support the elucidation of individual E-R relationships and hopefully ultimately better tailor ICI therapies to the often-evolving changes at their site of action in the course of cancer therapies. The work reported in the paper has been performed by the authors, unless clearly specified in the text.