The primary function of the immune system is to protect against infection by identifying and removing pathogenic disease causing organisms. Additionally it is primed to detect and destroy transformed cells, thereby preventing potential tumor formation. A complex system with many contributing cell populations and biological pathways, the immune system employs multiple control mechanisms to limit the extent of responses to avoid damage to the host when removing pathogens or tumors. One such mechanism utilizes inhibitory markers specific for certain immune cells that are activated to start an immune response. These markers, known as checkpoints, control the activity of these cells, controlling and ending immune activation following successful pathogen clearance, as well as assisting in the maintenance of tolerance to self-antigens. This is an important control mechanism; defects in which can lead to hyperstimulation of the immune system and the development of autoimmune disease. Importantly, this self-regulatory system is exploited by tumors in an effort to avoid surveillance by the immune system; it is now known that tumors seek to prevent immune attack by several mechanisms, including the triggering of certain checkpoint pathways to actively inhibit immune responses directed against them. 

Recently, immune checkpoint pathways have become an area of interest for oncology research and drug development. By preventing this system of negative regulation, the immune system is reactivated against a tumor and mechanisms for its killing and removal reinstated. The multifaceted nature of the immune system means that any interactions utilizing these pathways can have an impact on multiple other pathways, and thorough research into the pharmacology of new therapeutic products is essential in order to assess any downstream effects associated with their function.

Given the function of these products and the complexity of the immune system, assessing the safety of novel checkpoint inhibitor products is challenging and complex, relying heavily on a thorough understanding of the individual pathways being targeted as well as any potential effects resulting from the use of such products in combination. This review will describe the structure and function of the immune system, why checkpoint markers exist and why they are attractive therapeutic targets.

Approaches to assess the safety of new therapeutic products targeting these mechanisms will then be discussed.

The role of the immune system
Two arms of the immune system effectively carry out protection from infection. These are known as the innate and adaptive immune systems. Simplistically, the front-line innate system is non-specific and non-memory forming, providing a rapid but limited response. Adaptive immunity, by contrast, is specific for individual pathogens and this mechanism allows it to form memory.¹

Innate immunity
Every day we are exposed to millions of pathogens, through contact, ingestion and the air we breathe¹ and the first layer of defense against this assault is the innate immune system. Initially, this defense comprises physical barriers such as epithelial surfaces, cilia and secreted substances including tears and saliva, but these are breached by some pathogens. The innate immune system therefore also comprises multiple cell populations, including macrophages, dendritic cells, and granulocytes (eosinophils, basophils, and neutrophils), which mediate killing and removal of pathogens through direct mechanisms or by the production of soluble factors.

The innate immune system employs the recognition of particular molecules that are common to many pathogens, and absent in the host, in order to identify pathogens. Patterns within molecules such as the double-stranded RNA of some viruses, or molecules on the cell surface of bacteria enable the identification of these organisms as foreign and therefore requiring removal. Several types of dedicated receptors, collectively called pattern recognition receptors, recognize these pathogen-associated markers. These may appear as soluble components within the blood (the complement system), or membrane-bound on the surface of host cells (toll-like receptor family).¹ Following recognition by these proteins the pathogen is targeted and destroyed. Additionally, an inflammatory response is produced, which attracts and activates components of the adaptive immune system.

Although the innate immune system is an important initial line of defense, it is limited by the lack of memory; even though a particular pathogen has been successfully removed, it can subsequently cause reinfection if encountered by the host in the future. Additionally, there is no mechanism for amplification of an innate anti-pathogen response and rapid multiplication of the infectious agent may overcome the response mediated by the innate system and overwhelm the host. 

Adaptive immunity
The more sophisticated adaptive immune system present in higher organisms relies predominantly on CD4 and CD8 positive T cells, which help the immune response and drive cell-mediated immunity respectively, and antibody-producing B cells, which drive the humoral response, to specifically identify and target the pathogen for removal, leading to a highly efficient and specific response.

The cell-mediated component of adaptive immunity requires antigens from the pathogen to be presented to receptors expressed on the surface of T cells, in combination with signals to activate the cell. Upon recognition of these antigens as foreign, the T cells become activated and can proliferate and differentiate into effector cells. Two populations of T cells can be activated; CD4+ T helper cells which ‘help’ the effectiveness of the anti-pathogen response by enhancing the function of other cell populations, and CD8+ cytotoxic T cells which act directly on pathogens or tumor cells to eliminate them. CD8+ T cells respond to endogenous antigens presented by MHC class I on the surface of many cell populations, while CD4+ T cells are activated by exogenous antigens which have been processed and presented on MHC class II by specialized antigen presenting cells (APCs). APCs may be cells from the innate immune system such as dendritic cells and macrophages, or components of the adaptive response such as B cells. 

In the humoral immune response, antibodies, attached either to the outside of B cells or found in extracellular fluids, are the primary modulator of the immune response, through their direct binding with an antigen and activation of a response. Binding of antibody may inactivate viruses or microbial toxins (such as tetanus or diphtheria toxins) by blocking their ability to bind to receptors on host cells. More importantly, antibody binding marks pathogens for destruction, recruiting the phagocytic cells of the innate immune system to ingest them. Importantly, each T or B cell activated during a given infection, is specific for that pathogen only, providing a highly efficient and directed immune response against the infectious agent. Both types of cellular responses have methods of amplifying the response once an antigen is recognized and, importantly, once an adaptive immune response has been mediated against a given pathogen, specific cell populations recognizing the pathogen are retained in the circulation and remain primed to recognize this pathogen on reinfection. This provides the basis of immunological memory, which ensures that any subsequent re-encountering of the same pathogen leads to rapid clearance and the absence of symptomatic infection.

Tolerance
The immune system is an extremely complex system that swiftly activates in response to a threat. However, the mechanisms employed to kill pathogens can also mediate damage to the host and it is therefore vital that control mechanisms are incorporated to maintain a controlled response, thereby limiting tissue damage. While the innate immune system relies on the recognition of particular types of molecules that are common to many pathogens but absent in the host, other safety mechanisms exist in the adaptive immune system to provide ‘tolerance,’ enabling the body to control the extent and duration of an immune response, as well as preventing unwanted responses mediated against self-antigens. 

As T and B cells mature in the thymus and bone marrow respectively, central tolerance mechanisms ensure that only cells that can distinguish between self and non-self antigens are released into the circulation; those cells that respond inappropriately to self-antigens are removed from the pool of developing cells and do not enter the periphery. Additionally, the selection process ensures that only those cells that are able to respond to signals delivered via the T or B cell receptor are released into the periphery, this ensures that cells released into the blood are able to identify foreign antigens and respond appropriately, while ignoring self-antigens and therefore avoiding the development of autoimmune responses.

While central tolerance is extremely effective, there is still a need to control cells following activation in the periphery. This next layer of control termed peripheral tolerance is mediated by multiple mechanisms and serves to limit self-damage and further refine the response, ensuring that the immune system reverts to the resting state once the pathogen is removed. One mechanism by which this control is provided is through regulatory T cells, which actively suppress the function of other T cells, and inhibitory cytokines such as interleukins and transforming growth factor-beta produced after pathogen removal.¹ T cells themselves appear to control their response following activation and pathogen removal by becoming anergic or upregulating ligands for inhibitory receptors which then mediate a negative signal to the cell to prevent further activation. These inhibitory pathways utilize specific markers expressed on APCs or the T cells themselves, often termed checkpoints, which deliver an inhibitory signal to the cell, preventing further differentiation or proliferation, or by inducing cell death. One of these pathways is mediated by cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4), the first immune-checkpoint receptor to be clinically targeted.² This receptor binds to the ligands of a costimulatory marker critical for T cell activation at high affinity, out-competing their binding and thereby preventing delivery of the positive co-stimulation signal while preventing activation through delivery of an inhibitory signal through CTLA-4 itself. Another checkpoint marker, programmed death 1 (PD-1), the expression of which is induced following activation of various cells, including T cells, and non-T lymphocytes, is also showing promise as a clinical target. The mechanism used by tumor cells to exploit these pathways is discussed below.²

Checkpoint inhibitors
Tumors have developed multiple ways of evading the immune system, so that tumor killing becomes less efficient. A major focus in cancer immunotherapy is increasing tumor surveillance by preventing this evasion.

As shown in Figure 1, a large number of checkpoint markers on T cells bind to their respective ligands and receptors on APCs such as dendritic cells and macrophages to effectively switch off the T cells after activation. Tumors utilize these pathways to evade the immune system and avoid killing mechanisms; following upregulation of checkpoint ligands on their surfaces these ligands bind to receptors on the activated T cells specific for the tumor, thereby limiting the anti-tumor response and promoting survival of the tumor.

Because many of these inhibitory markers are upregulated by T cells following activation or exhaustion, blocking the signal between the inhibitory ligand on the tumor surface and its receptor on the T cell (for example, between CTLA-4 and CD80/86, or PD-1 and PD-L1) will reactivate cells. Suppressing these negative signals with blocking monoclonal antibodies can therefore drive T cells to more effectively kill the tumor. This is a novel and very exciting way to stimulate the immune system and is the basis underlying a number of exciting drugs on the market or in development, which target a range of different checkpoint markers.

PD-1 is a checkpoint marker expressed on T cells which binds to programmed death ligand 1 (PD-L1), a marker on some normal (and cancer) cells. When PD-1 binds to PD-L1, it inhibits the T cell response and thereby stops the cell from destroying the PD-L1-positive tumor. Some tumors express large amounts of PD-L1, and it is thought that this is one mechanism by which they evade immune attack. Monoclonal antibodies that target either PD-1 or PD-L1 block this binding and prevent the inhibition of the immune response against the cancer cells. These drugs have shown considerable promise in treating certain cancers.

Three licensed checkpoint inhibitor drugs are currently available: the anti-CTLA-4 mAb ipilimumab (Yervoy; Bristol-Myers Squibb), discussed in more detail later in this article, and the anti-PD-1 monoclonal antibodies nivolumab (Opdivo, Bristol-Myers Squibb) and pembrolizumab (Keytruda, Merck), while many more are currently in development.

Safety concerns
Of particular note, the inhibitory checkpoint receptors and the signaling mechanisms they mediate are an essential part of the physiological immune response, with expression induced following normal adaptive T cell activation to limit the response after pathogen clearance. Upregulation of these inhibitory checkpoint receptors will decrease responsiveness to antigenic stimulation. Likewise, ‘hijacking’ of this system by cancer cells, will make the immune system less responsive to the tumor. As observed clinically using checkpoint inhibitor products, inhibition of these targets can restore T cell function and enhance T cell-mediated tumor killing; however, taking the brakes off the immune system by blocking these receptors may alter the balance of the immune system toward hypersensitivity. By removing one or more of the normal control mechanisms, checkpoint inhibitor-targeting products could potentially promote inflammation and autoimmunity in addition to enhanced tumor cell killing.

This concern is in fact consistent with the phenotypes of mouse models lacking checkpoint markers. CTLA-4 knockout mice demonstrate an early fatal autoimmune syndrome characterized by lymphoproliferation,³ while some strains of PD-1 knockout animals develop autoimmunity later in life.⁴ Similarly, patients with genetic anomalies in certain checkpoint inhibitor genes are at increased risk of autoimmune disorders such as systemic lupus erythematosus, multiple sclerosis, or type 1 diabetes, which further demonstrates the potential side effects of modifying the way that these pathways function. 

Safety assessment considerations for checkpoint inhibitors
When considering the approach required for safety assessment of checkpoint inhibitor products, the intended immune activation should in itself be an initial cause for concern. In addition to the more general effects associated with immune activation, the potential risk of autoimmunity should be considered throughout non-clinical safety assessment of checkpoint inhibitor products.

It is therefore important to understand the magnitude and downstream effects of any increases in immune activation when designing non-clinical safety assessment packages for checkpoint inhibitor monoclonal antibodies. Additionally, the relative immune status of non-clinical animal models and the subsequent intended clinical population is also an essential consideration. As checkpoint markers are upregulated following activation of T cell populations, there may be less of these targets present in the relatively inexperienced immune systems of the animals used for non-clinical safety assessment programs. An understanding of the relative density of target expression and therefore potential pharmacology following treatment will help to elucidate how closely the expected clinical response will be replicated. If necessary, consideration should be given to whether it would be more useful to increase the expression of these markers by activating the immune system prior to administration of the novel checkpoint inhibitor product in order to induce a response more representative of that expected in the clinic.

Regardless of the marker targeted by a monoclonal antibody, there are a number of standard evaluations required as part of the safety program prior to its first clinical use. As for all monoclonal antibody products, tissue cross-reactivity studies are required prior to Phase I trials to characterize binding in humans and to identify whether this is as expected based on the known expression of the target. Pharmacokinetic (PK) analyses are important to assess the half-life of the product and therefore the duration that biological effects may be observed for following dose administration. Furthermore, repeat-dose toxicity studies must be performed in pharmacologically relevant species, considering which population of cells may be affected by inhibiting the intended pathway and assessing their function following dose administration to understand the safety profile expected in the clinic following dosing.

For monoclonal antibodies targeting checkpoint markers in particular, a robust assessment of the immune system and its function is required. A number of assessments of general immune status are already included within most standard pivotal toxicology packages and therefore provide an initial insight. Hematology assessments provide information on total and absolute differential leukocyte counts, while standard clinical chemistry panels include globulin levels and albumin/globulin ratios. Additionally, gross pathology of lymphoid organs and tissues, organ weights and histopathology of the lymphoid organs (bone marrow, spleen, lymph node, and thymus) are also performed, providing additional information on immune status in treated animals.

Assessment of the major immune cell populations such as T, B, and NK cells is now commonly included in relevant safety assessment programs and can be expanded to include specific sub-populations based on the expression pattern of the specific checkpoint marker being targeted, or its ligand. However, it is most important that a functional assessment of immune system is included to assess the safety profile of checkpoint-targeting products and a number of methods are available to address this. Assessment of T cell-dependent antibody responses (TDAR) has long been a key method for the assessment of immunotoxicity. Animals are vaccinated with a model antigen such as keyhole limpet hemocyanin (KLH) and the resulting antigen-specific antibody production is measured. The TDAR method is particularly useful, as although the final end point is the assessment of antibody responses, and therefore serves as a measurement of the humoral response, effective T cell function is also required to support this, and the method therefore provides a mechanism for assessing the function of the adaptive immune response as a whole. Traditionally the method has been used to assess immune function in the presence of products expected to ablate the immune system where assessment of more subtle changes in immune function is not required. However, by optimizing the KLH immunization regimen it is possible to avoid induction of maximal responses and therefore potentially also assess whether enhanced immune function is observed, as may be expected for checkpoint inhibitor products, and indeed observed in the case of Ipilimumab. Envigo has recently re-validated our TDAR immunization regimen in the non-human primate to induce sub-maximal responses, enabling the assessment activation of the immune response as well as its inhibition (see Figure 2). Such functional assessments are critical to understanding the effects mediated by checkpoint inhibitor products, as simple snapshots of circulating immune populations are unlikely to be sufficiently informative on the effects of treatment and the potential for over stimulation of the immune system.



Challenges for safety assessment of checkpoint inhibitor products
For checkpoint inhibitor products, the classical ‘no adverse event level’ approach to safety assessment may be less relevant than for molecules targeting other markers. First-in-human dosing calculations are instead likely to be based on the pharmacologically active dose derived from extensive non-clinical investigations, including the relative potency between human and non-clinical species, receptor occupancy studies, and extensive PK modeling to inform clinical dosing and dose escalation. The balance between the intended therapeutic pharmacology and the dose-limiting ‘exaggerated pharmacology’ is also critical, and dose level setting may be limited by the intended activity of the product rather than by toxicity per se.

Priming of the immune response prior to treatment with a checkpoint inhibitor is also an important consideration. One challenge is that animal models may be ‘clean’; while there may have been some level of previous antigen exposure, it is likely to be far less than in vaccinated and immune-experienced human patients. Less prior immune activation will result in lower expression of checkpoint markers and therefore a reduced amount of available target relative to the clinical population. The immune system may, thus require some amount of stimulation before the effect of a checkpoint inhibitor is determined to more closely mimic the patient population. However, it is also important to consider whether stimulation of these models by vaccination before assessing the checkpoint inhibitor response; could in fact lead to over prediction of inflammatory or autoimmune effects.

Considerations for the selection of animal models
Animal models used for safety assessment programs are often relatively immune-inexperienced and therefore may not have large resident populations of activated T cells. Furthermore, chronic autoimmune effects may be difficult to model in healthy, standard, non-clinical animal models. As discussed above, such models may also lack or have limited expression of checkpoint receptors. This can create a challenge in finding suitable models to represent the intended clinical population. Various differences in the population need to be considered too; for example, chemotherapy patients may have a very different immune status to that of a healthy animal.

As stated in ICH S6 (R1), for the safety assessment of any biopharmaceutical such as a monoclonal antibody, the non-clinical model(s) must be selected such that the intended clinical biology can be observed and all safety assessment studies must therefore be performed in pharmacologically relevant species. It must be demonstrated that the product can bind to its target to induce the desired pharmacological response and ideally this response should be measurable, in terms of function. For checkpoint-targeting products, the immune system must therefore be experienced enough to express the target, and sufficient activated T cells must be present to mediate the potential clinical biology.

Example of a non-clinical safety program for a checkpoint inhibitor monoclonal antibody: ipilimumab
The anti-CTLA-4 monoclonal antibody ipilimumab was the first checkpoint inhibitor product licensed and was approved in 2011 for the treatment of melanoma. Its safety assessment program, reported by the European Medicines Agency, provides a useful base for studies required as part of the non-clinical safety evaluation for a checkpoint inhibitor product.⁵

In the case of ipilimumab, a standard monoclonal antibody safety assessment program was performed. This included in vitro binding studies, showing high-affinity binding to CTLA-4 and inhibition of binding of its ligands CD80 and CD86 to human CTLA-4. The assessment also included in vitro and in vivo complement-dependent cytotoxicity and antibody-dependent T cell-mediated cytotoxicity tests, and efficacy studies using a colon carcinoma model in a human CTLA-4 transgenic mouse. Repeat dose toxicity studies assessing effects of treatment with ipilimumab alone were performed in the non-human primate and were well tolerated, with few signs observed. Interestingly, however, adverse effects were observed when animals were treated with ipilimumab in conjunction with an additional checkpoint inhibitor product.

In addition to classical safety studies, extensive non-GLP T cell-dependent antibody response (TDAR) studies including immune activation were also performed. Animals were vaccinated with a range of antigens to stimulate the immune system before investigating the effect of inhibiting CTLA-4 signaling by administering ipilimumab. This increased the proportion of cells expressing CTLA-4 and therefore increased the potential for any effects due to inhibition of this signaling pathway to be observed. The appearance of adverse events again correlated with this increased activation of cells, despite the fact that immunophenotypic analysis of T cells and their subsets indicated no global changes in proportions of circulating cells. This reinforces the hypothesis that the mode of action of ipilimumab was to increase immune function rather than alter its composition.

Clinical experience of checkpoint inhibitors: effect on safety evaluation
In addition to the pre-marketing safety data generated during the development of ipilimumab, clinical experience with checkpoint inhibitors can help to inform the design of non-clinical safety assessment programs. Indeed, signs and symptoms expected for immune system activation have been observed in the clinic, and adverse events including pneumonitis, colitis, skin effects (rash, pruritus, and vitiligo), conjunctivitis, uveitis, hepatotoxicity, thyroid toxicity, and nephritis have been reported.

Clinical efficacy has been shown to be patient- and tumor-specific; for example, in Phase 3 trials of both CTLA-4 and PD-1 inhibitor products, increased survival has been observed in patients with melanoma. By contrast, in a Phase 3 study of the anti-CTLA-4 mAb ipilimumab in patients with non-small cell lung cancer, there was no increase in overall survival. The response to treatment is likely to be affected by additional patient- and/or tumor-related factors, and stratification has, therefore, become a key focus. In addition to differential effects in various tumor types, specific tumor features, individual patient characteristics (for example, pre- or post-chemotherapy, state of the immune system, co-medications) and inhibitor combinations are currently being investigated. 

Combinations of checkpoint inhibitor products are now being utilized, with the potential for additive effects by targeting multiple pathways. The effects of combination therapy relative to monotherapy will need to be assessed for any product that is destined for combination use in the clinic and may demonstrate an altered safety profile. Indeed, combination studies assessing ipilimumab in combination with anti-CD137 or anti-PD-1 antibodies conducted in cynomolgus monkeys showed increased stimulation of the immune system and correlated with the reporting of adverse effects in these animals. Not all findings in cynomolgus monkeys were directly predictive of all clinical findings however. For example, the incidence of immune-mediated toxicities in monkeys was lower than that subsequently seen in patients, therefore, confirming the importance of understanding the relevance of the immune status of the animal model relative to the intended patient population so that all results can be suitably considered with respect to the clinical population.

These findings all indicate the importance of aiming to increase the potential to translate understanding from non-clinical studies to clinical practice, while understanding the limitations of these studies.

Summary
Checkpoint inhibitors are a novel and exciting class of drugs that are already making a difference in patients’ lives. However, manipulating the immune system to promote enhanced activity also presents the potential to introduce immunotoxic effects.

A translational approach to safety assessment is required to model the potential risks in non-clinical safety assessment in order to predict the clinical situation. Focusing on understanding the biology of these products and how they alter the function of the immune system is crucial in order to design appropriate safety programs. Functional immune assessments and a thorough understanding of how the selected non-clinical model represents the intended clinical population should therefore be an essential component of each non-clinical safety program for these products. 
 
References

1. Alberts B, Johnson A, Lewis J et al. Molecular Biology of the Cell, 4th edn. New York: Garland Science, 2002.
2. Pardoll DM. The blockade of immune checkpoints in cancer immunotherapy. Nat Rev Cancer 2016; 12: 252–264.
3. European Medicines Agency. Assessment Report For Yervoy (ipilimumab). EMA, 2011. www.ema.europa.eu/docs/en_GB/document_library/EPAR_-_Public_assessment_report/human/002213/WC500109302.pdf
4. Nishimura H, Okazaki T, Tanakaya Y et al. Autoimmune dilated cardiomyopathy in PD-1 receptor-deficient mice. Science 2001; 291: 319–22.
5. Tivol EA, Borriello F, Schweitzer AN et al. Loss of CTLA-4 leads to massive lymphoproliferation and fatal multiorgan tissue destruction, revealing a critical negative regulatory role of CTLA-4. Immunity 1995; 3: 541–547.

---

Kirsty Harper is head of biologics at Envigo. She has led the UK biologics team since October 2017, having been a member of the group since June 2013. Kirsty and her team support customers by designing safety studies and non-clinical development programs for biologics in response to specific requests as well as providing scientific support and advice. Prior joining Envigo, she was employed as principal scientist at Oxford Immunotec where she was responsible for pipeline product development projects and the provision of immunological advice and expertise.