Investigating The Distinct Role Of Pd-L2 In Modulating Human T Cells Responses

Chapter 1: Introduction

Cancer Immunotherapy, the 2013 “Breakthrough of the Year” leverages the patient’s immune

system to seek out and specifically kill the cancer cells[1]. The origin of the idea to harness

the potential of the immune system in cancer treatment dates back to the 19th century when

William Coley injected his cancer patient with heat inactivated bacteria and observed tumor

regression as a result of an activated immune system[2]. The intellectual momentum that

began with Coley’s observation took several decades of scientific research to finally place

immunotherapy at the centerstage of cancer treatment in the 21st century[3,4]. The concerted

efforts of basic science researchers and clinicians over the past century finally led to The Food

and Drug Administration (FDA) approval of Ipilimumab in 2011, a monoclonal antibody for the

treatment of unresectable or metastatic melanoma and firmly established Immunotherapy as

the fifth pillar of cancer treatment[5,6].

Allison and his colleagues’ curiosity to study the fundamental mechanisms of T cell regulation

led to their discovery of CTLA-4, the molecular target of Ipilimumab on T cells[7,8]. In parallel,

Honjo and his colleagues serendipitously discovered PD-1 which like CTLA-4 inhibited T cell

effector function albeit via a different mechanism of action[9,10]. With the FDA approval of

Pembrolizumab in 2015, a monoclonal antibody against PD-1, a distinct class of

Immunotherapy for cancer called the Immune Checkpoint Inhibitors (ICI) was formed[11,12].

In the past decade the FDA approved seven antibodies targeting the PD-1 co-receptor and

its ligand PD-L1 for the treatment of more than 85 oncology indications[13]. The incremental

understanding provided by studying the underlying fundamental mechanisms constantly

informed novel discoveries and treatments in the clinic. Currently more than 2000 clinical trials

are evaluating the efficacy of anti-PD1/PDL1 mAbs in combination with other treatments for

cancer. While ICI revolutionized cancer treatment and demonstrated that long-term remission

even with discontinued treatment is possible in a minority of patients, those who are refractory

to or relapse after treatment continue to remain in the majority[14,15].

The clinical applicability of ICI and other combination therapies can be expanded only through

a comprehensive understanding of the molecular mechanisms that dictate the dynamic

interaction between the tumor and the immune system. The clinical development of novel

treatment strategies relies on rigorous basic science research and its subsequent clinical

translation.

In Chapter 1, I will present background information necessary to understand and appreciate

the significance of my PhD research work. I will explain the background information in the

following order and present the objective of this dissertation:

1) Hide and seek: Cancer Vs Immune System

2) Immune checkpoint protein: PD-1 co-receptor

3) The role of PD-1 in different contexts of immune response

4) The therapeutic landscape of PD-1/PD-L1/PD-L2 signaling axis

1.1 Hide and seek: Cancer Vs Immune System

The field of oncology until the early 20th century relied on radiotherapy and chemotherapeutic

approaches for advancing cancer treatment[16]. In 1957 Burnet and Thomas postulated that

cancer cells accumulate “non-self” antigens that can stimulate the immune system and result

in regression of the tumor[17]. Their hypothesis received its long overdue validation in 2001

when Schreiber and colleagues showed that immune deficient mice, specifically lacking IFN-

ʏ and lymphocytes were more susceptible to carcinogen induced sarcoma than immune-

competent mice[18]. Schreiber refined the concept of cancer immunosurveillance and

proposed the term ‘cancer immunoediting’ and emphasized the ability of cancer cells to adapt

and outgrow the survival pressures exerted by the immune system and escape immune

detection[19,20]. Cancer immunoediting progresses through three distinct phases of

elimination, equilibrium, and escape[21–23].

The elimination phase retains the concept of immunosurveillance wherein immune cells patrol

the body to restrict the growth of malignant cells through inflammation. The immune system

specifically recognizes tumor-specific antigens (TSA) on transformed cells and eliminates

them through a diverse range of immune effector cells and molecules[24]. Seminal studies

like the discovery of interferons by Isaacs and Lindenmann in 1957, the role of T cells in

adaptive immunity by Miller in 1967 and the discovery of dendritic cells by Steinman and Cohn

in 1973 laid the foundation to infer the inherent anti-tumor potential of the immune system[25–

29]. The elimination phase ensures that developing tumors are destroyed and are clinically

undetectable in the body. The innate and adaptive immune systems work in harmony to carry

out immunosurveillance and protect the host from the development of any malignancies.

Since the end point of the elimination phase is the absence of cancer, experimental models

accurately representing this phase are limited[30].

Tumor cells that survive the elimination phase enter into a dynamic equilibrium with the

immune system during which the immune system is able to only contain the growth of the

tumor and not fully eradicate the tumor. During this phase of tumor dormancy, the transformed

cells continue to accumulate mutations as a result of the cancer cell’s intrinsic genetic

instability and gives rise to new variants that provide increased protection from any anti-tumor

immune response[31].

The Darwinian selection of variants leads to the escape phase of cancer immunoediting

wherein tumor cells grow and metastasize uncontrollably by avoiding recognition and

elimination by the immune system[32]. Adaptations in tumor cell intrinsic mechanisms like

mutational landscape, metabolic fitness, etc. together with tumor cell extrinsic factors like the

tumor microenvironment (TME), host-related factors like microbiota drive the immune escape

of cancer[32–40]. The function of T cells to specifically recognize tumor associated antigens

in the context of MHC molecules to carry out direct cytotoxic reactions against neoplastic cells

and form immunological memory placed them in the spotlight to dismantle the escape phase

and promote anti-tumor immunity[41]. While immune cells like T cells and NK cells are

emblematic of the cancer-killing potential of the immune system, an effective anti-tumor

response requires a series of stepwise events explained in the Cancer Immunity Cycle that

precedes the final blow served by T cells and NK cells[42].

The TME orchestrates a complex network of interactions that creates roadblocks at various

stages of the cancer immunity cycle to prevent an onslaught by the immune system. TME

constitutes molecular players like endothelial cells, immune cells, extracellular matrix, and

vasculature that create a favorable neighborhood for malignant cells to grow and flourish into

aggressive cancers[43–45]. The hallmark features of a pro-tumor TME are a hypoxic niche,

immunosuppressive stromal cells, increased lactate and ROS metabolism, expression of

immune checkpoint proteins and the more recently appreciated a stiff mechanical scaffold and

an innervated niche that supports metastasis[46,47]. Therapeutically targeting the TME is like

a game of Jenga but with the goal of destabilizing the structure. A single block will not weaken

the structure, however a strategic removal of a combination of blocks will let the tower fall. In

the last several decades cancer immunologists are focused on attacking the tumor from

multiple fronts to achieve durable clinical responses[48].

Allison and Honjo’s seminal discovery that blockade of immune checkpoint proteins can

mobilize anti-immune response was the advent of modern immunotherapy against cancer.

For the first time immune checkpoint inhibitors made long term remission a possibility for some

cancer patients and provided a new vision for cancer management. In spite of the hope that

ICI gave rise to, it is not the ultimate answer for cancer treatment. It is important to

acknowledge that innovation and improvement of novel therapeutics against immune

checkpoint proteins can be achieved only through a comprehensive understanding of the

associated immunobiology, which in turn will aid scientists and clinicians to anticipate possible

evasion mechanisms of the tumor or be aware of any potential off-target clinical effects.

1.2 Immune checkpoint protein: PD-1 co-receptor

Immune checkpoint proteins are gatekeepers that maintain self-tolerance and temper the

intensity of immune responses in normal human physiology to prevent any collateral

damage[49,50]. A number of immune checkpoint proteins have been discovered and

investigated for therapeutic benefit against cancer, for example, PD-1, CTLA-4, PD-L1, PD-

L2, TIGIT, LAG3, TIM3, ILT2, B7H3, BTLA, KIR, NKG2A and CD96. FDA-approved inhibitors

against PD-1, CTLA-4 and PD-L1 revolutionized the landscape of cancer immunotherapy and

other proteins are being actively investigated for the next generation of immune checkpoint

blockade[51,52]. A monoclonal antibody against CTLA-4 (Ipilimumab) was the first immune

checkpoint inhibitor that received FDA approval in 2011 for unresectable, late-stage

melanoma. While Ipilimumab and Tremelimumab remain as the only FDA-approved

monoclonal antibody (mAb) against CTLA-4, our arsenal against PD-1 is expanding with 4

FDA-approved mAbs (Nivolumab, Pembrolizumab, Cemiplimab and Dostarlimab)[6,53].

Many attribute the severity of side-effects associated with CTLA-4 blockade and it’s restricted

mode of action in lymphoid tissues as the reason behind the preference for PD-1

blockade[54,55].

Programmed cell death-1 (PD-1 or CD279) is a transmembrane protein and a member of the

CD28/B7 superfamily[56]. It is mainly expressed on activated T cells and binds to two known

ligands, PD-L1 (CD274) or PD-L2 (CD273) to inhibit proliferation and effector functions of T

cells. PD-1 is a type 1 transmembrane glycoprotein with a single extracellular IgV domain, a

hydrophobic transmembrane domain, and a cytoplasmic tail with two pivotal signaling

domains, namely, immunoreceptor tyrosine-based inhibitory motif (ITIM) and immunoreceptor

tyrosine-based inhibitory motif (ITSM)[57]. Initially thought to be expressed only on activated

T cells, PD-1 expression is now appreciated in other cells such as Natural Killer (NK) cells,

Dendritic cells (DCs), Macrophages, Tumor cells, B cells, Myeloid derived suppressor cells

(MDSCs), and Tregs[58,59]. It is generally inferred from previous studies that PD-1 on effector

immune cells inhibit their effector function whereas PD-1 on suppressive immune cells like

Tregs remains controversial and warrants future investigation. This further emphasizes the

complexity of immune therapy and the need to continue investigating the immunobiology of

therapeutic targets for yet to be discovered biological impacts.

The extracellular domain of PD-1 contains a β sheet face that engages the ligands. Murine

PD-L1 and PD-L2 have comparable binding affinities to PD-1 but human PD-L2 exhibits a 2-

6 fold higher affinity to human PD-1 than PD-L1[60]. Structural studies of human PD-1/ PD-

L1 and PD-1/PD-L2 complexes revealed that distinct amino acid residues in the extracellular

domain of PD-1 are engaged by PD-L1 and PD-L2 resulting in significant differences in the

interfaces formed[61]. Whether these differences result in major conformational

rearrangement of the transmembrane protein is yet to be demonstrated. While it is classically

accepted that PD-1 on T cells binds to PD-L1 on antigen presenting cells (APCs), recent

reports support a possible cis interaction between PD-1 and PD-L1 as well[62]. Besides

interacting with PD-1, PD-L1 is also reported to interact with CD80, a co-stimulatory ligand

that engages the CD28 receptor on T cells[63]. PD-L2 does not bind to CD80 and is reported

to bind a second receptor called Repulsive guidance molecule-b (RGMb) in mouse models.

CD80 binds to PD-L1 when co-expressed on the same cell (cis-CD80/PD-L1) and disrupts

the PD-1/PD-L1 complex formation while retaining the CD28/CD80 interaction. The cis-

CD80/PD-L1 complex recruits PD-L1 away from PD-1 and CD80 from CTLA-4 mediated

trans-endocytosis resulting in a net co-stimulatory effect on T cells[64]. The role of PD-L2

remains unresolved with conflicting evidence suggesting both co-stimulatory as well as co-

inhibitory functions. The non-linear interactions between immune checkpoint proteins add

several layers of complexity that need to be considered for designing therapeutic

interventions.

Mutational studies show that PD-1/PD-L1 inhibitory function depends on the Lck mediated

ITSM phosphotyrosine that preferentially recruits SHP-2 phosphatase to downregulate T cell

activation[65–67]. Single cell imaging experiments elucidated a molecular mechanism of PD-

1 inhibition wherein PD-1/PD-L clustered with TCR to form a microcluster that induces the

dephosphorylation of the membrane proximal TCR signaling molecules[68]. The role of ITIM

in PD-1 mediated signaling is explained by contradicting models of protein interaction. In one

model it is postulated that ITSM strongly binds to the C-SH2 domain of SHP-2 and recruits

the phosphatase proximal to the membrane, while the ITIM binds to the N-SH2 domain and

activates the phosphatase. In a different model it is shown that the N-SH2 and C-SH2 domains

of SHP-2 could bridge two phosphorylated ITSM regions on two PD-1 tails to form a PD-1:

PD-1 dimer at the plasma membrane[69]. While it is mostly accepted that SHP-2 is the only

immediate partner of PD-1, but a recent report showed mice with SHP-2 deleted T cells

capable of PD-1 mediated inhibition indicative of a possible SHP-2 independent downstream

signaling of PD-1[70]. PD-1 ligation ultimately results in the dephosphorylation of the TCR-

CD3 and CD28 signaling axis leading to the inhibition of PI3K/Akt, MAPK and mTOR

pathways in T cells[71]. PD-1 is also able to reprogram the metabolic state of T cells by

favoring fatty acids in β-oxidation over glycolysis and by decreasing the threshold of TGF-β

mediated signaling[72].

1.3 Diverse roles of the PD-1 signaling axis

Immune checkpoint proteins regulate T cell responses through ligand-receptor interactions

during different contexts of inflammation and T cell differentiation. In addition to regulating the

effector functions of T cells, PD-1 can also regulate T cell memory responses as well. At the

time of first encounter with an antigen, the strength of the TCR signal will dictate the balance

between effector responses vs memory responses[73,74]. The precise role of PD-1 in

influencing immunological memory is dependent on variables such as type of immune

pressure, tissue context, stage of T cell response and TCR signal strength.

PD-1 was first implicated in the regulation of T-cell tolerance and autoimmunity when the

phenotype of PD-1-/- mice was characterized by late onset of lupus-like disease and mild

glomerulonephritis in C57BL/6 mice[75]. The inhibitory signal delivered by the PD-1/PD-

L1/PD-L2 axis can regulate both central and peripheral T-cell tolerance in normal human

physiology[76].For example, the PD-1/PD-L1 complex restricts positive selection in the CD4-

CD8- (DN) to CD4+CD8+ (DP) stage of T cell development in the thymus. In the periphery,

PD-1 inhibits the initial phase of activation and expansion of self-reactive T cells and prevents

any organ injury by self-reactive T cells. PD-1 is also known to enforce a cell intrinsic

mechanism of restraint on Tregs, such that PD-1 deficient Tregs exhibit enhances

immunosuppressive function[77]. PD-1 and PD-L1 were implicated in the pathogenesis of

Type 1 diabetes (T1D) in mouse models. Similarly, the PD-1 signaling axis was studied in the

context of Multiple Sclerosis, Inflammatory bowel disease and rheumatoid arthritis[76]. These

studies focused only on the PD-1/PD-L1 interaction and extrapolated it to be true for PD-1/PD-

L2 binding as well. Only recently the specific role of the PD-1/PD-L2 axis in airway immune

tolerance was studied to show that PD-L2 promoted the maintenance of Foxp3+ pTregs by

supporting their metabolic activity and Foxp3 stability[78]. This observation is aligned with a

different study that showed a positive correlation between the severity of asthma and absence

of PD-L2[79].

During acute infections like malaria and chronic infections like HIV or Hepatitis B virus immune

checkpoint molecules are upregulated on T cells and restrict immune-mediated clearance of

the infection[80]. The severity of PD-1 mediated exhaustion in T cells is proportional to the

viral load that ultimately results in the elimination of antigen-specific T cells[81]. Recent studies

of malaria contributed to the contention around PD-L2 and showed that higher expression of

PD-L2 on blood DCs correlated with lower parasitemia[82]. The study postulated that PD-

1/PD-L2 complex outcompetes PD-1/PD-L2 interaction to protect T cells from PD-1/PD-L1

mediated exhaustion during malarial infection. During chronic HIV or HBV infection PD-1 is

overexpressed on CD4+ and CD8+ T cells and correlated with increased disease progression

and higher viral load. Immune checkpoint blockade in viral infections like HIV could have

beneficial outcomes like increased vaccine responsiveness or increased HIV-specific T cell

function to eliminate HIV-infected cells[83].

Maternal-fetal immunotolerance is a unique challenge faced by the immune system where the

mother’s immune system has to restraint immune responses against the semiallogenic fetus

while protecting the mother from other infections[84]. At the maternal-fetal interface high

expression of PD-1 is found on decidual T cells that reduce Th1 type of immune response at

the maternal-fetal interface during a healthy human pregnancy[85]. In mice, blockade of the

PD-1 pathway during pregnancy increased T cell infiltration and Th1 cytokine production in

the placenta resulting in decreased pup survival[86]. During gestation PD-ligands showed

different expression patterns, but the physiological relevance of the difference is still known.

PD-L1 is expressed on the villous synctiotrophoblast and extravillous cytotrophoblast while

PD-L2 is limited to villous cytotrophoblast[87,88]. An interesting study demonstrated that the

function of PD-L2 was modified during the evolutionary transition from marsupial to placental

mammals, emphasizing a possible unique role of PD-L2 in the placenta that is not shared with

PD-L1[89].

The PD-1 pathway is efficiently hijacked by the tumor to evade the immune system and

bolsters the clinical progression of the disease. PD-1+ tumor infiltrating lymphocytes (TILs)

observed in several human cancers are characterized by impaired effector cytokine

production, increased expression of multiple inhibitory receptors, altered metabolism and loss

of proliferating capacity[90]. PD-L2 is widely expressed across many types of cancers, in

tumor cells, endothelial cells as well as on immunosuppressive stromal cells. The expression

of PD-L2 in human cancers is only recently appreciated and was shown to be a better

predictor of progression-free survival with pembrolizumab in head and neck squamous cell

carcinoma than PD-L1 [91]. However, preclinical studies in murine models suggest that PD-

L2 promotes tumor immunity through a PD-1 independent mechanism[92]. In the last decade

several studies investigated the role of PD-1 on immune cells other than T cell as well as on

nonimmune cells[59,93–95]. The discovery of PD-1 on tumor cells and subsequent research

on it has led to the “tumor cell-intrinsic PD-1/L1 paradox”[96,97]. While PD-1 is a potential

oncogene in pancreatic, melanoma, and bladder cancer cells, but is reported to be a tumor

suppressor in non-small cell lung cancer and colon cancer cells. A previous study reported

the rapid disease progression of stage IV NSCLC in a 61-year old woman after treatment with

pembrolizumab[98–100]. The precise molecular mechanism underlying this paradox is

important to decipher since it will inform current and future clinical potential of PD-1 blockade.

This also emphasizes the non-canonical cis-PD-1/PD-L1 interaction that may be at play on

the surface of tumor cells expressing PD-1 and PD-L1[101]. Recently, it was reported that

PD-1 on dendritic cells (DCs) neutralized PD-L1 in cis to inhibit PD-1 signaling in T cells[102].

Additionally, it is also important to understand the role of PD-L2 during non-canonical

interactions of PD-1 since human cancers are now known to upregulate both the PD-ligands.

The field has come a long way from describing the PD-1 signaling axis as a simple on/off

switch that regulates T cell function to an intricate signaling repertoire characterized by a

gradient in the intensity of responses. Since PD-1 and its ligands are expressed on a wide

variety of cells, therefore current and future therapies should be cognizant of the functional

consequence of disrupting the PD-1/PD-L1/PD-L2 signaling axis on different cellular

compartments that may yield new potential targets for combinatorial therapies.

1.4 Therapeutic landscape of the PD-1/PD-L1/PD-L2 signaling axis

As of 2022, 4,897 active clinical trials are investigating anti-PD1/PDL1 mAbs as a single agent

or in combination with other therapies[103]. The majority of these clinical trials are evaluating

combination therapies affirming that the future of immunotherapy for cancer relies on

mechanism of action based synergies of drug therapies. FDA-approved mAbs against PD-

1/PD-L1 include Pembrolizumab, Nivolumab, Durvalumab, Atezolizumab, Avelumab,

Cemiplimab and Dostarlimb[6]. Pembrolizumab, is a human IgG4κ monoclonal antibody

against PD-1 that was first approved in 2014 in metastatic melanoma patients who were

refractory to CTLA-4 therapy and is investigated in the highest number of clinical trials in

comparison to other mAbs in the PD-1/PD-L1 arsenal. Since 2014, anti PD-1/PD-L1 mAbs

have been FDA-approved for several cancers with the latest one on February 9, 2023, for

dostarlimab in adult patients with mismatch repair deficient (dMMR) recurrent or advanced

endometrial cancer. Through these clinical trials the field of Immunotherapy aims to achieve

the holy grail of cancer treatment, that is, maximal tumor clearance with minimal treatment

associated toxicities. 10% of patients receiving anti-PD-1 antibodies have grade 3 immune

related adverse events (irAEs) that mostly occur within 6 months of treatment[104]. The most

common irAEs associated with PD-1 blockade include pneumonitis, hepatitis, colitis,

endocrinopathies and diarrhea[105–107]. A meta-analysis of 19 clinical trials studying anti-

PD-1/PD-L1 antibodies showed a higher incidence of pneumonitis in patients receiving anti-

PD-L1 antibodies. The comparatively lower incidence of anti-PD-L1 blockade may be due to

the preservation of PD-1/PD-L2 interaction that maintained immune homeostasis in

patients[108]. A recent study showed that PD-L1 antibodies like atezolizumab, avelumab and

durvalumab were superior to PD-1 antibodies like pembrolizumab and nivolumab to disrupt

the PD-1 signaling axis[109]. This was the first time therapeutic antibodies with similar

mechanisms of action were compared directly against each other. Even though both PD-1

and PD-L1 antibodies target the PD-1 signaling axis, the multi-directional interactions of

proteins in the B7/CD28 family results in unique clinical consequences of these

therapies[110].

While the current clinical landscape of PD-1/PD-L1 blockade reflects innovation and a

promising future for immunotherapy and cancer treatment, it also necessitates the need to

avoid duplicative development programs. Through the inception of Project Orbis, the FDA

Oncology Center of Excellence seeks to harmonize the various drug development programs

and achieve greater efficiency in patient care and treatment[111]. The unprecedented durable

responses of PD-1/PD-L1 blockade are a reality for only a small population of patients, while

the majority either do not respond to the therapy at all (primary resistance) or relapse after a

short period of response to immunotherapy (acquired resistance)[112]. Resistance to PD-

1/PD-L1 blockade can be driven by tumor cell intrinsic or tumor cell extrinsic factors[113].

Tumor cell intrinsic factors include mechanisms like altered antigen-presenting machinery,

expression of immunosuppressive ligands, mutations in interferon gamma signaling,

oncogenic signaling through the MAPK pathway, and upregulation of genes associated with

mesenchymal transformation[113–117]. Tumor cell extrinsic factors that support resistance to

immunotherapy include cells like Tregs, MDSCs, M2 macrophages and an

immunosuppressive metabolic microenvironment[118–120]. Alternate inhibitory immune

checkpoints like TIM-3, LAG-3, TIGIT and VISTA can compensate for the PD-1/PD-L1

blockade and result in unresponsive cancers to therapy[121]. Several cancers employ

immune suppression, immune exclusion, and immune ignorance to prevent anti-tumor

immune response[122]. The majority of modern immunotherapy is focused on combinatorial

approaches to strategically tear down the above mentioned resistance mechanisms of the

tumor. A clear understanding of the immunobiology underlying the therapeutic targets is

fundamental to designing combinatorial therapies with optimal synergies in the clinic.

1.5 Objective of the dissertation

The need for durable therapy against cancer encouraged the clinical translation of immune

checkpoint blockade and the subsequent accelerated clinical expansion to different cancer

types. The therapeutic efficacy of immune checkpoint proteins is built upon an incomplete

understanding of the mechanistic biology that dictates the crosstalk between these proteins.

The newly discovered cis interaction between PD-L1 and CD80 expands on the current model

of PD-1/PD-L1 and CTLA-4/CD80 signaling axis and presents a new perspective for

therapeutic intervention[64]. The paradoxical role of tumor-intrinsic PD-1 as an oncogene in

some cancers vs as a tumor suppressor in others was a recent discovery that may help refine

PD-1 blockade as a therapy in the clinic. Studies like this make it clear to the field that there

is still a lot unknown about the complex multi-directional interactions of PD-1 and its ligands.

Despite the concerted efforts between immunologists and clinicians to delineate the PD-1

signaling axis, our understanding will continue to remain incomplete until we appreciate the

role of PD-L2 in PD-1 mediated T cell inhibition. So far, PD-L2 is assumed to be a redundant

ligand of the PD-1 co-receptor on T cells based on preclinical mouse models that may not

accurately represent the role of PD-L2 in humans. Through this dissertation work we seek to

provide a deeper insight into the molecular relationship between PD-1, PD-L1 and PD-L2 and

to highlight the functional outcome of the differential interaction between PD-1, PD-L1 and

PD-L2. We hypothesize that PD-L2 is solely a co-inhibitory ligand in humans that

transduces an inhibitory signal via PD-1 that is distinct from PD-L1. We propose to (1)

define the function of PD-L2 in humans, (2) delineate the inhibitory signal mediated by PD-L2

through PD-1 and (3) elucidate the functional and physiological consequence of PD-L2

mediated inhibition of T cell function.

Chapter 2: Methodology

2.2. Cell lines

Chinese Hamster Ovarian (CHO) cells expressing anti-CD3 protein on the surface and PD-1

expressing Jurkat T cells with a luciferase gene under the control of the NFAT response

element (NFAT-RE) were purchased from the Promega Corporation as part of the PD-1/PD-

L1 Blockade Bioassay. The CHO cells were retrovirally transduced to overexpress human

RGMb, PD-L1 alone, human PD-L2 alone and both human PD-L1 and PD-L2. The cells were

bulk sorted by flow cytometry for surface level PD-ligand expression.

The mesothelin specific CAR-T cells and 3T3 cells overexpressing mesothelin were a gift from

Dr. Pawel K. Mazur lab (UT MD Anderson Cancer Center, Houston, TX). The murine glioma

cell line GL261 was a gift from Amy Heimberger lab (UT MD Anderson Cancer Center,

Houston, TX). The murine cancer cell lines and human cancer cell lines from Figure1 were

purchased from American Type Culture Collection (ATCC; Manassas, VA).

The lentiviral plasmid vector carrying Cas9 and gRNA targeting human PD-1 were purchased

from GeneCopoeia Inc (Rockville, Maryland) and transiently transfected into Jurkat T cells.

The transfected cells were sorted by flow cytometry for lack of PD-1 expression on the surface.

We tested the PD-1 knock-down Jurkat T cells for surface expression of PD-1 after activation

with CD3/CD28 to ensure knock down of endogenous PD-1. The plasmid vectors for the

mutant forms of PD-1 were purchased from Genscript Biotech Corp (Piscataway, NJ) and

retrovirally transduced in PD-1 knock-down Jurkat cells.

2.2. Quantification of PD-ligands expression

We assessed the surface expression of PD-L1 and PD-L2 in murine and human cancer cell

lines by flow cytometry. The cytokines used in the experiment were purchased from

PeproTech and reconstituted according to manufacturer’s instructions. We quantified the

number of PD-ligand molecules on the surface of retrovirally transduced CHOs by using PE

quantibrite beads purchased from BD Biosciences.

2.3. NFAT-RE Reporter Cell Assay

All bioluminescence assays were conducted according to manufacturer’s protocols (Promega

PD-1/PD-L1 Blockade Bioassay). In short, CHO cells expressing the indicated PD-ligands

were grown, counted and plated at a concentration of 4X104cells/well in a 96-well flat bottom

(white) plates. The cells were allowed to adhere to the surface of the plate overnight and the

following day Jurkat T cells were harvested, washed and counted and added to the wells

containing CHO cells at a concentration of 5X104cells/ well. The indicated antibodies were

titrated and added to the well accordingly. The cells were incubated for 6 hours at 37ºC. At

the end of the incubation time Bio-Glo reagent was added to the wells and the luminescence

was measured using a luminometer (Schematic 1). Dr. Chantale Bernatchez was kind

enough to allow us to use her laboratory’s luminometer. The same experimental set up was

used to study the impact of PD-ligands on Jurkat T cells expressing mutant forms of PD-1 on

the surface.

2.4. Cytometric Bead Array

The cytometric bead array was performed according to manufacturer’s protocols (BD

Cytometric Bead Array Human Th1/Th2/Th17 Cytokine Kit) to quantify the levels of cytokines

in the supernatant of our co-culture experiments. In short, the human TH1/Th2/Th17 cytokine

standards were reconstituted and titrated in assay diluent and mixed with PE detection

reagent. The volume of capture beads corresponding to each of 7 cytokines required for the

assay was calculated and mixed. The supernatants from the co-culture experiments were

harvested to be cell-free and mixed with the capture beads in a 1:1 ratio. The supernatants

were then incubated at room temperature for 3 hours on a dry surface before acquiring the

samples by flow cytometry.

2.5. PDL2-RGMb Interaction Assay

We purchased recombinant human RGMb protein from Sino Biological Inc. (USA) and

recombinant human PD-L2 Fc chimera protein from R&D Systems. 293T and CHO cells were

retrovirally transduced to overexpress RGMb(His Tag) protein on the surface. We tested the

ability of recombinant RGMb and PDL2 to interact with the surface of CHO-PDL2 and 293T-

RGMb cells, respectively. The recombinant protein was detected by secondary antibodies

against the His tag and Fc region. We also assessed the interaction between RGMb and PDL2

using NTA-Atto 488 conjugates that specifically bind to His-tag. The NTA-Atto-488 kit was

purchased from Millipore Sigma.

2.6. Microarray analysis

Microarray analysis was performed on total RNA isolated from Jurkat T cells after co-culture

with CHO cells overexpressing PD-ligands (Schematic 1) using the RNeasy Mini Kit (Qiagen)

according to the manufacturer’s instructions. RNA quality was assessed using an Agilent 2100

Bioanalyzer and Affymetrix (ThermoFisher) performed the gene expression profiling. The

robust multi-array average (RMA) was used to normalize raw data from batches using

R/Bioconductor. Differentially enriched genes were identified using the DESeq R package and

filtered using a q value<0.001. We performed pathway analysis using GSEA preranked

platform and filtered differentially regulated pathways using a false discovery rate <0.25.

2.7. RPPA Analysis

PD-1+ Jurkat T cells were co-cultured with CHO cells overexpressing the indicated ligands

for 6 hours. After the co-culture the Jurkat T cells were harvested and spun down to remove

contaminating CHO cells and washed twice with cold PBS and dry pellets were stored at -

80ºC. The experiment was performed twice independently, and the cell pellets were sent to

the Proteomic Facility at the Baylor College of Medicine and UT MD Anderson Cancer Center.

We performed a RPPA experiment with pre-activated PBMC isolated human CD4+ and CD8+

T cells co-cultured with CHO cells overexpressing PD-Ligands for 24 hours and the T cells

were harvested and spun down before being flash frozen. The dried cell pellets were sent to

the Proteomics Facility at Baylor College of Medicine. For each pathway considered in the

analysis the members were decided based on literature. The pathway score is then calculated

as the sum of the relative protein level of all positive regulatory members minus that of

negative regulatory members in a particular pathway[123].

2.8. Statistical analysis

Graphical representation and statistical analysis of the data was performed using GraphPad

Prism (Version 8, GraphPad Software, San Diego, CA). Data shown in this dissertation show

mean +/- SD. Sample groups in the experiments were assessed using ANOVA analysis

followed by Tukey’s multiple comparison test. All parametric statistical analysis were

performed based on the assumption of normality and homoscedasticity.

Chapter 3: Results

3.1. Mouse PD-L2 does not reflect the immunobiology of human PD-L2

Despite being a higher-affinity binding ligand of the PD-1 co-receptor, PD-L2 is largely

excluded from the arsenal of ICI therapy due to previous failed reports of PDL2 blockade in

transplantable mouse tumor models[89]. The failure of PDL2 blockade to control tumor growth

may be because of its reported T cell co-stimulatory function in mice[124]. Murine PD-L2

(mPD-L2) can co-stimulate CD4+ T cells through the engagement of a second receptor called

Repulsive Guidance Molecule B (RGM-b)[125]. However, the co-stimulatory function of

human PD-L2 (hPD-L2) is often assumed to be like its murine counterpart despite the lack of

any substantial evidence.

Recent reports demonstrated PD-L2 expression across many human tumor tissues and

emphasized the clinical relevance of the ligand as a predictive biomarker of response to α PD-

1 therapy[126]. The widespread expression of PD-L2 on human tumors and its association

with poor prognosis is not accurately reflected by the inefficacy of PD-L2 blockade observed

in mouse models. Hence, we hypothesized that murine PD-L2 does not recapitulate the

underlying biology of human PD-L2 and differs in its function.

In order to investigate the precise function of PD-L2 in humans, we first assessed the surface

expression of PD-L2 on murine and human cancer cell lines through flow cytometry. Murine

tumor cell lines expressed PD-L1 on the surface and upregulated the expression when treated

with cytokines like IFNʏ and IL-2 (Figure 1a; Supplemental Fig S1b). We did not observe

any surface expression of PD-L2 on murine tumor cell lines even after induction with

cytokines. Contrary to mouse biology several human tumor cell lines expressed comparable

basal levels of PD-L1 and PD-L2 (Figure 1b; Supplemental Fig S1a). Moreover, some

human cancer cell lines like SNU-387 and HDLM-2 showed higher surface expression of PD-

L2 than PD-L1 demonstrated by the 2-4 fold increase in mean fluorescence intensity. The

preferential lack of expression of PD-L2 on murine tumor cell lines and widespread expression

of PD-L2 on human tumor cell lines provided preliminary evidence for the possibility of

divergent regulation of PD-L2 in humans and mice.

In order to further emphasize the differences in the function of PD-L2 in mice and humans,

we investigated whether like mPD-L2, hPD-L2 could interact with RGM-b. We tested the direct

binding of hPD-L2 to human RGM-b (hRGM-b) in an in vitro protein-protein interaction assay

and found that recombinant hPD-L2-Fc fusion protein did not bind to 293T cells

overexpressing hRGM-b (Supplemental Fig S1c) on the surface (Supplemental Fig S1d).

However, at the highest concentration tested (10µg) we observed weak binding of His-tagged

hRGM-b recombinant protein to the surface of CHO cells overexpressing hPD-L2 (CHO-

hPDL2). In order to exclude any non-specific binding of anti-His tag secondary antibody, we

validated the interaction between hPD-L2 and hRGM-b in an assay using Ni-NTA-Atto

conjugates that specifically bind to His-tag with minimal cross-reactivity. The Ni-NTA-Atto

conjugates recognized the His-tagged hRGM-b overexpressed on CHO cells and confirmed

the validity of the experimental setup. However, soluble His-tagged hRGM-b recombinant

protein did not specifically bind to the surface of CHO parental and CHO-hPDL2 cells and

hence was not detected by the Ni-NTA-Atto conjugates (Figure 1c).

Our findings showed that PDL2 expression in the context of mouse tumors did not replicate

the expression pattern of human PDL2 and emphasized the possible divergence in the

function of PD-L2 between humans and mouse models. We directly interrogated the

interaction between PD-L2 and RGM-b in humans and found that unlike their mouse analogs,

human PD-L2 does not bind to human RGM-b.

Figure 1: Characterization of the biological differences between mouse and human PD-

L2. (a) Surface expression of PD-L1 and PD-L2 across murine cancer cell lines before (light

blue: PD-L1; light red: PD-L2) and after (deep blue: PD-L1; deep red: PD-L2; black:

Unstained) induction with IFNʏ and IL-4 (100µg/ml). (b) Surface expression of PD-L1 and PD-

L2 across human cancer cell lines at the basal level without cytokine inductions (Blue: PD-L1;

Red: PD-L2; Black: Unstained). (c) Mean fluorescence intensity denoting the detection of His-

tagged RGMb protein by Ni-NTA-Atto conjugates.

Supplementary Figure S1: Extended data on biological differences between mouse and

human PD-L2. Mean fluorescence intensity denoting the surface expression of PD-L1 and

PD-L2 across human cancer cell lines (a) and murine cancer cell lines (b). (c) Mean

fluorescence intensity denoting surface expression of hRGM-b in 293T cells retrovirally

transduced to overexpress hRGM-b.

3.2. Human PD-L2 is solely a co-inhibitory ligand that does not bind RGM-b

We wanted to determine the precise role of hPD-L2 in T cell activation by quantifying the

degree of dephosphorylation in the TCR-CD3 signaling axis mediated via the PD-1 co-

receptor. Since PD-L2 is expressed on several human cancers as observed in our results as

well as in previous scientific literature, we postulated that human PD-L2 is a co-inhibitory

ligand of T cell function and does not retain the co-stimulatory role observed in mice.

We retrovirally transduced CHO cells to overexpress hPD-L1 alone (CHO-hPDL1), hPD-L2

alone (CHO-hPDL2) and both hPD-L1 and hPD-L2 (CHO-cisPDL1/L2) on the surface

(Supplemental Fig S2a, S2b, S2c). Transduced CHO cells were sorted based on surface

expression of comparable number of PD-1 ligands. We quantified the number of ligands using

PE-quantibrite beads as 2000 ligands on the surface of CHO-hPDL1, CHO-hPDL2 cells and

around 5000 ligands on CHO-cisPDL1/L2 cells (Supplemental Fig S2d). We also quantified

the number of PD-ligands on the surface of human cancer cell lines and found that BFTC-909

expressed ~15,000 ligands of PD-L1 and PD-L2 on the surface, whereas KARPAS expressed

higher number of PD-L2 than PD-L1 ligands on the surface (Supplemental Fig S2e). The

CHO cells also expressed the anti-CD3 scFv protein fragment on the surface that activated

the TCR-CD3 axis in Jurkat T cells, resulting in bioluminescence due to the luciferase reporter

gene under the regulation of the NFAT-response element (RE) (Schematic 1). As a negative

control, we used CHO-S cells that did not express the anti-CD3 protein fragment (No

activation). Using this reporter system, we found that CHO-hPDL2 like CHO-PDL1

suppressed T cell activation and did not co-stimulate T cells (Figure 2a). PD-1 blocking

antibody restored T cell activation demonstrated by the increase in bioluminescence and

confirmed that PD-1 on Jurkat T cells mediated the inhibition of the TCR-CD3 axis. Similarly,

anti-PDL1 and anti-PDL2 blocking antibodies showed that it was the specific PD-1/PD-L1 and

PD-1/PD-L2 interaction that inhibited the activation of Jurkat T cells. While PD-L2 blocking

antibody completely relieved PD-1/PD-L2 inhibition, it only partially restored inhibition

mediated by a mix of CHO-hPDL1 and CHO-hPDL2 (CHO-transPD-L1/L2: 4X104 cells of

CHO-hPDL1 + 4X104 cells of CHO-hPDL2) since PD-1 continued to interact with PD-L1.

Surprisingly, only PD-L1 blocking antibody and not PD-L2 blocking antibody interfered with

CHO cells co-expressing PD-ligands (CHO-cisPD-L1/L2) inhibition and restored

bioluminescence. The co-expression of PD-L1 and PD-L2 on the same cell may have resulted

in cis-interaction between the ligands such that inhibition via PD-1 was primarily driven by PD-

L1. As expected, PD-L1 blocking antibody efficiently relieved inhibition by CHO-PDL1 as well

as CHO-cisPD-L1/L2 and partially CHO-transPD-L1/L2. The combinatorial blockade of PD-

L1 and PD-L2 replicated the data observed in the mono-blockade of PD-1.

In addition to the influence of PD-ligands via PD-1 on membrane proximal signaling events,

we also tested the effect of PD-L2 on human T cell proliferation and cytokine production. We

co-cultured CFSE stained Jurkat T cells with CHO-hPDL1, CHO-hPDL2 and CHO-transPD-

L1/L2 for 72 hours and assessed the dilution of CFSE by flow cytometry (Figure 2b). CHO-

hPDL1 inhibited proliferation of Jurkat T cells efficiently but CHO-hPDL2 appeared to have a

less potent effect in the experiment. CHO-transPD-L1/L2 exhibited a null effect as the higher

affinity of PD-L2 out-competed PD-L1 to bind to PD-1. The increased proliferation induced by

CHO-transPD-L1/L2 compared to CHO-hPDL2 may be due to the increased number of CHO-

transPD-L1/L2 cells (8X104) used in the experiment, resulting in a 2X engagement of the TCR-

CD3 complex by the αCD3 protein expressed on the CHO cells. We ensured to maintain the

CHO cell number at 4X104 across all the samples in our following experiment. We also

validated the inhibitory function of PD-L2 during antigen-specific T cell activation. We showed

that CAR-T cells in the presence of PD-ligands produced significantly lower amounts of pro-

inflammatory cytokines like IL-2, IFN-g and TNFa when activated by its corresponding antigen

mesothelin for 72 hours (Figure 2c).

Taken together, our data demonstrated that human PD-L2 is exclusively a T cell co-inhibitory

ligand that binds to the PD-1 co-receptor to inhibit T cell cytokine production and NFAT-

mediated T cell activation. However, PD-L2/PD-1 binding inhibited the proliferation of T cells

to a lesser extent than PD-L1 and appeared to be a weaker inhibitory ligand of PD-1 at least

at the level of proliferation.

Schematic 1: Representation of the bioluminescent assay to characterize the nature of

regulation mediated by PD-L1 and PD-L2 through PD-1 when Jurkat T cells are activated via

TCR-CD3 complex.

w”)

‘€

250000

200000;

150000

100000

50000

T T T T T

T T T T

4-3-2

-1

-3

-2

-1

-3 -2

-1

Log,o[Isotype]yg/ml

Logiof[aPD1]yg/ml

Logio[aPD-L1]ug/ml

Log,

[aPD-L2]

g/ml

300000

250000

200000

150000

100000

activation

50000

aCD3

protein

@PD-L1

4-3-2

-1

4-3

-2

-1

Sie

@cis

Log4o[aPD-L1]ug/ml

Log

4o9[aPD-L2]ug/ml

trans

PD-L1/L2

KKK

100;

KKK KKK

801

oOo

]

activation

aCD3

protein

46)

PD-L1

PD-L2

@trans

PD-L1/L2

KK KK

200,

*#**

“STE

80)

__*

1000,

“ee

180}

70:

pee

goo]

~~"

****

160;

604

800;

@|e@

1404

—

700;

1201

30)

600]

[ee

1004

404

«900;

@®|

80;

®@®

30:

@®

400;

3004

Mesothelin

-—-

PD-L1

PD-L2

- -

60-

40}

200]

204

100]

v T T

T T

+ +

I++

1+,

C++

Figure 2: PD-L2 is solely a T cell co-inhibitory ligand in humans. (a) 4X104 CHO cells

overexpressing indicated proteins and 5X104 PD-1+ Jurkat T cells (NFAT-RE-Luciferase

reporter) were co-cultured for 6 hours with a titration of indicated immune checkpoint blockade

antibodies. TCR-CD3 activation was measured by luciferase assay according to

manufacturer’s instructions and represented as bioluminescence units. (b) PD-1+ Jurkat T

cells were co-cultured with indicated CHO cells overexpressing PD-Ligands for 72 hours and

proliferation was quantified by dilution of CFSE stain. (c) Mesothelin specific CAR-T cells were

co-cultured with 3T3 cells overexpressing mesothelin, PD-L1 and PD-L2 for 72 hours and pro-

inflammatory cytokines levels (IL-2, IFNʏ and TNFα) in the supernatant were measured by

Cytokine Bead Array. Statistical significance was calculated using a one-way ANOVA. *p<

0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Supplementary Figure S2: Extended data on the T-cell co-inhibitory function of human

PD-L2. (a)(b) CHO cells expressing the αCD3 protein were retrovirally transduced to

exclusively express PD-L1 (CHO-hPDL1), PD-L2 (CHO-hPDL2) and co-express both the

ligands (CHO-cis-PD-L1/L2) and assessed for surface expression by flow cytometry. (c) Mean

fluorescence intensity of PD-1 expression on the surface of Jurkat T cells. (d) The number of

PD-L1 and PD-L2 molecules on the surface of CHO cells after retroviral transduction and (e)

human cancer cell lines at the basal level are shown. (f) Expression of human PD-L1 and (g)

human PD-L2 on the surface of 3T3 cells by flow cytometry. (h) Western blot of mesothelin in

whole cell lysates of 3T3 cells. Statistical significance was calculated using a one-way

ANOVA. *p< 0.05, **p<0.01, ***p<0.001, ****p<0.0001.

3.3. Human PD-L2 binds PD-1 to induce a unique transcriptional program in Jurkat T

cells

The unique biophysical properties of the PD-1/PD-L2 interaction and its restricted expression

pattern on antigen presenting cells (APCs) and the female reproductive tract as opposed to

the widespread expression of PD-L1 on inflamed tissues strongly indicate towards a possible

nonredundant role of PD-L2 in humans[61,89]. Furthermore, NMR simulation studies of the

extracellular domain of PD-1 elucidated that the PD-ligands bind to the same region on human

PD-1 in seemingly different ways such that PD-L2 formed a smaller interface than PD-L1 to

make the PD-1/PD-L2 complex more stable than the PD-1/PD-L1 complex. We wanted to

investigate whether the differential engagement of PD-1 by its ligands can result in distinct

signaling mechanisms downstream of PD-1. Hence, we hypothesized that human PD-L2

binds to PD-1 to induce an inhibitory mechanism via PD-1 that is different from PD-L1.

While it is known that PD-1 on binding to PD-L1 recruits SHP-2 to attenuate T cell activation,

it was only recently appreciated that PD-1 can also reprogram the transcriptome of an

activated T cell by inhibiting the upregulation of distinct sets of T cell effector genes

[57,67,69,127]. Even though this study was seminal in elucidating the nuanced mechanism of

PD-1 inhibition, much like the body of research before it, the study ignored the role of PD-L2

in PD-1 mediated T cell inhibition.

With an aim to comprehensively study the PD-1 signaling axis we evaluated the genetic

program induced by PD-1 when bound to PD-L1 or PD-L2 using the Jurkat-CHO system

(Schematic 1). We co-cultured Jurkat T cells with each type of adherent CHO cell transduced

to express PD-ligands (CHO-hPDL1, CHO-hPDL2, CHO- cis-CHOPDL1/L2 and trans-

CHOPDL1/L2) for 6 hours and harvested bulk RNA to generate Affymetrix microarray data.

A venn diagram depicting the overlap of differentially enriched genes (adj.p.value<0.05)

identified a transcriptional program of 160 genes and 330 genes that were unique to PD-1/PD-

L2 and PD-1/PD-L1 meditated inhibition, respectively (Figure 3a). To gain more insight into

the physiological relevance of this unique signature we performed Gene Set Enrichment

Analysis (GSEA) (Figure 3c). We observed significant differences in the molecular pathways

regulated by PD-L1 and PD-L2. Hallmark pathways associated with T cell activation like TNFa

Signaling via NFKB, MTORC1 signaling, and Glycolysis were selectively enriched by PD-L2

when compared to PD-L1. We also found cell proliferation related gene sets like G2M

checkpoint, E2F Targets and Mitotic Spindle downregulated by PD-L2 versus PD-L1 (Figure

3b). Our findings suggest that PD-L1 inhibits T cells through the downregulation of pathways

like MTORC1 (downstream to PI3K/Akt pathway) resulting in a transcriptomic signature

aligned with previous studies in the field. However, PD-L2 did not follow the same mechanism

and instead preferentially impacted cell cycle associated pathways. Some of the DEGs that

contributed to the core enrichment of the differential pathways were the upregulation of

BCL2L1L, MXD3, PIFI, KIF2A, KIF3B and SDC2 and the downregulation of FBXO5, IRF1,

IFI35, TAP1, PEA15, SQLE, DHCR24, EBP, MXD3 and FADS1. While dissecting the

differences between PD-L1 and PD-L2 we found that PD-L2 binding to PD-1 significantly

enriched MYC targets and Apoptosis pathways. This was the first time induction of apoptosis

was observed with PD-1/PD-L2 mediated inhibition. T cell activation, inhibition and activation

induced cell death (AICD) is regulated by a fine balance of molecular signaling. Further studies

to confirm the direct correlation between PD-1/PD-L2 signaling and apoptosis are warranted.

We also observed differentially enriched genes in Jurkat T cells exposed to PD-ligands

expressed in cis or trans to each other, however they did not translate into dramatic

differences in the regulation of molecular pathways in the GSEA results (Figure 3d). This

supported our earlier speculation that PD-L1 and PD-L2 when co-expressed on the same cell

may form heterodimers that retain binding affinity to PD-1. Members of the B7 family of

proteins mediate several non-linear protein-protein interactions and confirmation of a cis

interaction between the PD-ligands will be pivotal to improve therapeutics targeting the PD-1

signaling axis. In conclusion the data suggested that the PD-L2 induced a transcriptional

program through the PD-1 co-receptor that is different from PD-L1 to inhibit TCR-CD3

mediated activation. The transcriptional program induced by PD-L2 was less inhibitory in its

phenotype when compared to PD-L1 regulated transcriptional program.

Figure 3: Characterization of the transcriptional program induced in Jurkat T cells by

PD-1/PD-L1 and PD-1/PD-L2 interactions. (a) Venn diagram showing the overlap of

differentially enriched genes (DEGs; adjusted p value <0.05) between Jurkat T cells inhibited

by hPD-L1, hPD-L2, cis-PD-L1/L2 and trans-PD-L1/L2 through PD-1 in the presence of CD3

activation. (b) Gene set enrichment analysis (GSEA) enrichment plots for the indicated

pathways. The y-axis shows the enrichment score (ES) and the x-axis details the genes

(vertical lines) characteristic of the indicated pathways. The red colored band in the plot

correlates to a high gene expression whereas the blue correlates for a low gene expression.

comparing PD-1/PD-L1 inhibited T cells against PD-1/PD-L2 (c) and PD-1/CHO-cis-PD-L1/L2

against (d) PD-1/CHO-trans-PD-L1/L2 inhibition.

Pathway

Apoptosis

PI3K/Akt

PISK/Akt

PI3K/Akt

PISK/Akt

PI3K/Akt

PISK/Akt

PI3K/Akt

PISK/Akt

PI3K/Akt

PISK/Akt

PI3K/Akt

Ras/MAPK

mTOR

Cell

cycle

Cell

cycle

Cell

cycle

Cell

cycle

Cell

cycle

Cell

cycle

Cell

cycle

Cell

cycle

Cell

cycle

Cell

cycle

Cell

cycle

Cell

cycle

Cell

cycle

Cell

cycle

Antibody

Bad_pS112

Bak

Bax

Bdl2A1

Bcl2

Bcl-xL

Bid

Bim

Caspase-3

Caspase-7-cleaved

c-IAP2

Akt1_pS473

Akt_p$473

Akt_pT308

GSK-3a-b_pS21_S9

INPP4b

p27_pT198

PRAS40_pT246

PTEN

Tuberin_pT1462

PI3Kp85

PI3KP110A

PI3Kp11b

mTORpS2448

mTOR

pPDK1($241)

PDK1

B-Raf_pS445

c-Jun_p$73

C-Raf_pS338

JNK_pT183_Y185

MAPK_pT202_Y204

MEK1_pS217_$221

p38_pT180_Y182

p9ORSK_pT573

Shc_pY317

YB1_pS102

A-Raf

c-Jun(60A8)_R_V

p38/MAPK_R_V

p44/42MAPK(Erk1/2)_R_V

4E-BP1_pS65

mTOR_pS2448

p70-S6K_pT389

Rictor_pT1135

S$6_p$235_$236

S6_p$240_$244

PRAS40

PRAS40_pT246

Raptor

mTOR

p70-S6K1

Cdc2_pY15

CDK1

Cydin-B1

Cydin-D1

Cycdlin-D3

Cydin-E1

FoxM1

p27_pT198

p27-Kip-1

PCNA

p-AuroraA(T288)/B(T232)/C(T198)_R_V

AuroraB

p21

p53

Weight

-1

PRP PRP

RP RP RP

PRP

RP BP RP RP BP RP RP BP BP RP BP

BPP

BP BP BP

PRP

PP ee

PPR

RP RP

Supplemental Figure S3: Extended data on RPPA analysis. (a) Differentially enriched

proteins when PD-L2 inhibited T cell is compared to PD-L1 inhibited T cells is depicted. (b)

Pathway score membership and the weight assigned to each member protein.

3.4. Human PD-L1 and PD-L2 elicit different protein repertoires through PD-1

In addition to the microarray experiment we also assessed the functional differences in the

signaling pathways induced by PD-L1 and PD-L2 through Reverse Phase Protein Array

(RPPA). We used the same co-culture experiment as before and flash froze whole cell pellets

for RPPA. In each sample Jurkats were activated only through the TCR-CD3 axis and

provided a simultaneous signal through PD-1. Hierarchical cluster analysis of normalized

RPPA data from MD Anderson revealed dramatic global differences in the proteomic

signatures of Jurkat T cells exposed to each of the PD-ligands (Figure 4a). Despite binding

to the same co-receptor, Jurkats inhibited by PD-L2 clustered closer to uninhibited Jurkats

and did not resemble the PD-L1 induced signaling repertoire. These observations were also

reproduced in an independent RPPA experiment performed at Baylor College of Medicine that

further highlighted the robustness of our findings (Figure 4c).

At a more granular level we studied the differential protein expression between PD-1/PD-L1

and PD-1/PD-L2 interactions. We observed that phosphor-Akt(S473), PRAS40, p110b-PI3K

were preferentially downregulated while PTEN was upregulated by PD-L2 (Supplemental

Figure S3a), suggesting that PD-1 engagement of PD-L2 and not PD-L1 suppressed the

TCR-CD3 mediated activation of the PI3K/Akt pathway in Jurkat T cells. Similar results were

also observed in the Baylor data (Figure 4a, 4c). As we analyzed the RPPA data further we

found that PD-L2 induced the expression of pro-apoptotic proteins like cleaved Caspase-3,

cleaved Caspase-7 and Annexin that aligned with our earlier observation of enriched

apoptosis in the transcriptomic data.

We also characterized the functional association between the differentially regulated proteins

through literature-driven pathway analysis. We focused on five cellular pathways, namely,

PI3K/Akt, mTOR, Apoptosis, Cell Cycle and Ras/MAPK and calculated pathway activity score

based on protein predictors curated from the literature (Supplemental Figure S3b). In the

heatmap, blue color represented a suppressed pathway and red indicated an activated

signature (Figure 4b). Jurkat T cells that received only the TCR-CD3 stimulus showed active

PI3K/Akt, Ras/MAPK, Cell cycle pathways along with an inactive Apoptosis pathway. In

addition to the differential clustering evident in the dendrogram, we also observed an inverse

relationship in the regulation of the assessed cellular pathways between PD-L1 and PD-L2.

We found that PD-L2 inactivated mTOR, PI3K/Akt, Cell cycle and Ras/MAPK pathways at the

proteomic level more potently than PD-L1. This was in contrast to our transcriptomic data

where we concluded that PD-L2 was the weaker inhibitory ligand than PD-L1. Similar to our

transcriptomic data we found the apoptosis pathway to be enriched by the PD-1/PD-L2

interaction. The only point of commonality between PD-L1 and PD-L2 inhibited T cells was

the suppression of the Ras/MAPK pathway by both the PD-ligands. It was surprising to find

that PD-1/PD-L1 did not strongly suppress the PI3K/Akt/mTOR pathway. It is important to

remember that in our experimental set up Jurkat T cells received only a TCR-CD3 activation

in the absence of any co-stimulation. While it is known that CD28 is the primary target for PD-

1/PD-L1 inhibition, our RPPA results suggest that PD-1/PD-L2 inhibition may be more

effective in attenuating the signaling partners downstream to the TCR-CD3 axis. We were

also the first to observe a direct association of PD-L2 and activation of apoptosis in Jurkat T

cells, an impact that was not evident in the PD-L1 sample.

We began our investigation without assuming a redundant role of PD-L2 in T cell function and

reported that PD-L2 binds to PD-1 to induce a distinct inhibitory signaling repertoire that is

supported by transcriptomic and proteomic evidence.

Figure 4: Characterization of the signaling repertoire induced in Jurkat T cells by PD-

1/PD-L1 and PD-1/PD-L2 at the proteomic level. (a) Hierarchical cluster analysis of RPPA

data of Jurkat T cells inhibited by the indicated PD-ligands expressed on CHO cells. This

experiment was performed at the University of Texas MD Anderson Cancer Center

Proteomics Facility (n= 3 technical replicates). Some proteins that were differentially regulated

by PD-L1 and PD-L2 are indicated (red: upregulated; green: downregulated). (b) Heatmaps

demonstrating mean pathway scores after unsupervised hierarchical clustering. (c)

Hierarchical cluster analysis of differentially enriched proteins from the RPPA data performed

at the Baylor College of Medicine (p value cutoff <=0.05).

3.5. PD-L2 engages PD-1 to weakly suppress TCR-CD3 activation and induce apoptosis

in Jurkat T cells.

While both our microarray and RPPA showed enrichment of apoptosis by PD-L2, the

regulation of other pathways associated with T cell activation were incongruent between the

two experiments. The microarray results showed that PD-L1 was a more potent inhibitory

ligand and suppressed pivotal pathways like mTORC1 and TNFα signaling via NFKB.

However, from our RPPA experiment we observed that PD-L2 was better at suppressing TCR-

CD3 activation. Nevertheless, both the platforms helped prove our hypothesis that the

inhibitory signaling induced through PD-1 by PD-L1 and PD-L2 is distinct.

It was imperative that we validated our results in functional assays. We continued with the

same experimental set-up of CHO cells overexpressing PD-ligands and Jurkat T cells

expressing PD-1. We also decided to analyze the temporal kinetics of PD-L1 and PD-L2

mediated inhibition and performed co-culture assays over the course of 24, 48 and 72 hours.

We first tested our earlier observation of upregulation of apoptosis by PD-1/PD-L2 interaction.

Apoptosis of Jurkat T cells was determined by staining with AnnexinV and 7-AAD (Figure 5a).

We also quantified active caspases using Z-VAD-FMK, a caspase inhibitor conjugated to FITC

for detection by flow cytometry (Figure 5b). The caspase inhibitor is a cell permeable

molecule that irreversibly binds to active caspases in Jurkat T cells. We gated PD-1+ Jurkat

T cells to exclude any possible CHO cell contamination and analyzed AnnexinV/7-AAD

staining by flow cytometry. Early apoptosis was characterized as AnnexinV+/7-AAD-, late

apoptosis as AnnexinV+/7-AAD+ and live cells as AnnexinV-/7-AAD- in the FACS dot plots.

The Jurkat T cells that were only activated through the TCR-CD3 axis and not exposed to any

ligands remained consistent with 20% of PD-1+ cells in early apoptosis. Jurkat T cells

stimulated via TCR-CD3 are susceptible to induction of apoptosis. After 24 hours of co-culture,

we did not observe any significant differences in the viability of Jurkats across the different

groups and noticed that CHO-cis-PD-L1/L2 induced a 2-fold increase in apoptosis over the

mono expressing CHO cells, CHO-PDL1 and CHO-PDL2. The differential regulation of

apoptosis between PD-L1 and PD-L2 became evident after 48 hours when we saw that PD-

L2 inhibited Jurkats had a significantly higher percentage of cells in early apoptosis than PD-

L1 inhibited Jurkats despite comparable viability across the two samples. At 72 hours, PD-L2

experienced Jurkats still retained a higher percentage of early apoptotic cells however the

degree of difference between PD-L1 and PD-L2 was reduced. The initial induction of

apoptosis by CHO-cisPD-L1/L2 cells contributed to the dramatic drop in viability to 30% at the

end of the experiment. In addition to AnnexinV staining we also evaluated the activation of

caspases in Jurkat T cells inhibited by PD-ligands. Even though we recorded statistically

significant differences between PD-L1 and PD-L2 in AnnexinV staining, we found that the

levels of activated caspases in Jurkats exposed to PD-L1 or PD-L2 were comparable

suggesting that caspase-independent mechanism of apoptosis may be at play in our

experiment. It was also interesting to find viable Jurkat cells cocultured with CHO-cis-PD-

L1/L2 that were positive for active caspases indicating towards possible non-apoptotic role of

caspases.

Taken together we observed that apoptosis in Jurkat T cells was a consequence of TCR-CD3

activation that was sub-optimally mitigated by PD-L2 due to a weaker inhibitory signal via PD-

1. The preferential induction of apoptosis in Jurkat T cells by CHO-cis-PDL1/L2 may have

relevance in the context of co-expression of the PD-1 ligands in human diseases like cancer.

Figure 5: PD-1/PD-L2 complex mediates a weaker inhibitory signal than PD-1/PD-L1 that

sub-optimally suppresses activation induced apoptosis in Jurkat T cells. (a) Jurkat T

cells were co-cultured with CHO cells overexpressing the indicated ligands for 72 hours and

the percentage of apoptosis was measured by flow cytometry through AnnexinV and 7-AAD

staining. Bar plots depict the percentage of early apoptosis cells (7AAD-AnnexinV+) and live

cells (7AAD-) in each sample. (b) We also measured the levels of active caspase by using the

CaspGLOW Fluorescein Active Caspase Staining kit. The mean fluorescence intensity of

activate caspases in early apoptotic cells across the tested samples is depicted. Statistical

significance was calculated using a one-way ANOVA. *p< 0.05, **p<0.01, ***p<0.001,

****p<0.0001.

3.6. PD-L2 engages PD-1 to inhibit TCR-CD3 activation of primary human T cells

through a mechanism that is distinct from PD-L1

Our work so far leveraged Jurkat T cells as the model system to elucidate signaling

mechanisms associated with the PD-1/PD-L1/PD-L2 signaling axis. While the system

facilitated our high-throughput molecular studies we also acknowledged the short-comings of

a transformed T-cell line and validated our results in primary human T cells. We utilized

PBMC-isolated human T cells to confirm the differential inhibitory signal mediated by PD-L1

and PD-L2 via PD-1. Since naïve T cells do not express PD-1 on the surface we activated

PBMC-isolated human T cells with aCD3/aCD28 coated beads for 40 hours and cocultured

the activated T cells with CHO cells expressing PD-ligands for 24 hours. After coculture we

flash froze whole cell pellets for RPPA analysis. Additionally, we also utilized the same

experimental setup and stained the T cells with Propidium Iodide (PI) and AnnexinV/7-AAD to

evaluate the impact of PD-ligands on cell cycle and apoptosis.

The differential protein repertoire elicited by PD-L1 and PD-L2 in Jurkat T cells was replicated

in CD4+ and CD8+ human T cells as well (Figure 6a, 6b). Hierarchical clustering of the RPPA

results emphasized the distinct protein signaling patterns and confirmed that PD-L2 inhibited

human T cells clustered away from PD-L1 inhibited T cells and resembled more closely to

TCR-CD3 activated human T cells.

When we tested the influence of PD-L1 and PD-L2 on the cell cycle of Jurkat T cells we found

that PD-L2 preferentially accumulated Jurkat T cells in the S phase depicted by Propidium

iodide (PI) staining (Supplemental Figure S6). We chose to corroborate our previous findings

in functional assays using PBMC-isolated human T cells. Since we observed a 2-fold increase

in the % of Jurkat T cells in the S-phase (Supplemental Fig S4), we speculated that PD-L2

unlike PD-L1 did not arrest T cells in the G1/G0 phase rather accumulated T cells in the S

phase. This was also in line with our GSEA findings where we found PDL2 downregulated

pathways like Mitotic Spindle Pathway that contribute to the execution of the S phase in the

cell cycle. When we investigated the impact of PD-ligands on the cell cycle of primary human

T cells we found that PD-L1 restricted T cells in the G1/G0 but PD-L2 did not impede the

progression of the cell cycle and was comparable to activated T cells (Figure 6c, 6d). We

also studied the correlation between PD-L2 and apoptosis in primary human T cells and found

that only after 96 hours of coculture with PD-ligands did we observe a significant difference in

the percentage of apoptotic human T cells (Figure 6e, 6f). Once again, the effect of PD-L2

on human T cells was comparable to that of activated T cells. We conclude that PD-L2 sends

a weaker inhibitory signal through PD-1 which in turn contributes to a downstream signaling

mechanism that was distinct from PD-L1.

Figure 6: PD-L2 binds to PD-1 to induce a signaling repertoire that is distinct from PD-

L1 in CD4+ and CD8+ PBMC-isolated T cells. (a) (b) Hierarchical cluster analysis of RPPA

data of human PBMC-isolated CD4+ and CD8+ T cells co-cultured with CHO cells expressing

the indicated PD-Ligands. (c)(d) Percentage of CD4+ and CD8+ T cells in different stages of

the cell cycle based on propidium iodide staining after 72 hours of co-culture with PD-ligands.

(e)(f) Percentage of early apoptotic CD3+ T cells that are AnnexinV+7-AAD- after co-culture

for indicated time durations with CHO cells overexpressing PD-L1 or PD-L2. Statistical

significance was calculated using a one-way ANOVA. *p< 0.05, **p<0.01, ***p<0.001,

****p<0.0001.

Supplemental Figure S4: PD-L2 engagement of PD-1 arrests the Jurkat T cells in the S

phase of the cell cycle. (a)Jurkat T cells were co-cultured with CHO cells expressing PD-L1,

PD-L2 or co-express PD-L1 and PD-L2 for the indicated time durations and stained with

propidium iodide to quantify the percentage of live cells in different stages of the cell cycle.

Statistical significance was calculated using a one-way ANOVA. *p< 0.05, **p<0.01,

***p<0.001, ****p<0.0001.

3.7. PD-L1 and PD-L2 rely on the ITSM region in PD-1 cytoplasmic tail to repress TCR-

CD3 activation of Jurkat T cells

To further elucidate the differential capacity of PD-L1 and PD-L2 to inhibit T cell activation, we

evaluated the inhibitory effect on membrane proximal TCR signaling partners. We lysed Jurkat

T cells after exposure to PD-ligands (5 mins or 10 mins) and blotted with antibodies against

phospho-epitopes of proteins in the TCR-CD3 signaling axis. We observed that PD-L2 was a

weaker inhibitor than PD-L2 since at 10 mins PD-L1 more potently dephosphorylated proteins

like ZAP70, LAT, ERK and PLCʏ (Figure 7a). However, PD-L2 was still inhibitory relative to

no PD-ligand available for PD-1 engagement.

Several years of research focused on PD-1 contributed towards our current understanding

that the ITSM region in the cytoplasmic tail of PD-1 recruits SHP-2, a phosphatase that is

responsible to dephosphorylate elements of the CD28 and TCR-CD3 signaling axis[67].

However, these studies were largely based on PD-1 engagement of PD-L1 and did not look

into the relevance of PD-L2 in this model. Based on our findings we hypothesized that PD-L2

utilizes the ITIM region in the cytoplasmic tail of PD-1 instead of the ITSM to mediate the

weaker inhibitory signal via PD-1. We performed CRISPR-Cas9 guided knock-down of the

surface expression of endogenous PD-1 in the NFAT-luc reporter Jurkat T cells and confirmed

the levels of PD-1 expression by flow cytometry (Figure 7b) and quantitative-PCR (qPCR)

(Figure 7d). We confirmed the absence of PD-1 expression on Jurkat T cells even after

activation with aCD3/aCD28 dynabeads (Figure 7c). We then transduced the PD-1KD Jurkat

T cells with a phosphor-dead ITIM mutant (Y223F), or a phosho-dead ITSM mutant (Y248F)

or a double mutant with inactive ITIM and ITSM regions (Figure 7e). A phospho-dead mutant

will not be able to create a docking site for any relevant phosphatases even if the extracellular

domain of PD-1 engages a cognate ligand. Similar to our previous experiment in Fig 1, TCR-

CD3 activation is quantified as a measure of bioluminescence. WT Jurkats express PD-1 on

the surface that retain the endogenous cytoplasmic domains and exhibit reduced

bioluminescence due to inhibition by PD-L1 or PD-L2 (Figure 7f). The bioluminescence was

restored by Jurkat T cells wherein PD-1 surface expression was knocked down. Y223F(ITIM)

PD-1 mutant had an intact ITSM region that continued to inhibit T cell activation but the

Y248F(ITSM) PD-1 mutant showed increased bioluminescence suggesting that the ITSM

region is indispensable for PD-1 mediated inhibition of NFAT-activation in T cells. Therefore,

our hypothesis was disproved since both the ligands relied on the ITSM region to transmit the

inhibitory signal downstream of the transmembrane PD-1 co-receptor.

Figure 7: The ITSM region on the cytoplasmic tail of PD-1 is indispensable for PD-1/PD-

L2 mediated inhibition TCR-CD3 activation in Jurkat T cells. (a) Jurkat T cells expressing

PD-1 on the surface were co-cultured with PD-L1 and PD-L2 for 5 and 10 minutes and blotted

for phosphor-proteins associated with the TCR-CD3 signaling axis. (b) Quantification of PD-1

expression in Jurkat T cells transiently transfected with Cas9-gRNA vector specific to PD-1

by flow cytometry, after activation with aCD3/CD28 dynabeads (c) and qPCR (d). (e) KD-1

Jurkats were retrovirally transduced with indicated mutant forms of PD-1 and sorted based on

restored PD-1 expression. (f) 4X104 CHO cells overexpressing indicated proteins and 5X104

mutant forms of PD-1+ Jurkat T cells (NFAT-RE-Luciferase reporter) were co-cultured for 6

hours and TCR-CD3 activation was measured by luciferase assay according to

manufacturer’s instructions and represented as bioluminescence units. Statistical significance

was calculated using a one-way ANOVA. *p< 0.05, **p<0.01, ***p<0.001, ****p<0.0001.

3.8. PD-1/PD-L2 inhibited T cells are poised to functionally recover more readily than

PD-1/PD-L1 inhibited T cells.

Our findings so far discovered and detailed the distinct inhibitory signal mediated by PD-L2

through PD-1 in humans. PD-L2 binds to PD-1 to transduce a weaker inhibitory signal than

PD-L1 resulting in a distinct signaling repertoire downstream to PD-1. We hypothesized that

PD-L2 mediated inhibition of T cells through PD-1 leaves T cells in a less inhibited state than

PD-L1 and allows the inhibited T cells to recover more readily. We tested this hypothesis by

inhibiting pre activated human T cells (to induce expression of PD-1) from 2 different donors

with CHO-hPDL1 and CHO-hPDL2 for 72 hours and then rested them overnight before re-

activating them with CD3/CD28 T cell activator beads for another 72 hours and analyzed the

levels of cytokines in the supernatant by cytokine bead array (Figure 8a). Post inhibition we

detected low levels of cytokines in the supernatant and no difference was observed across

the samples. The absence of co-stimulatory signal for 72 hours during inhibition must have

resulted in the low levels of cytokine produced by the uninhibited T cells (Figure 8a). Post

recovery only PD-L2 inhibited T cells produced IL2 and TNFα significantly more than PD-L1

inhibited T cell and was comparable to the uninhibited T cells (grey) (Figure 8c). This data

supports our hypothesis that PD-L2 inhibited T cells recovered more effectively than PD-L1

inhibited T cells and readily produced effector cytokines like IL-2 and TNFα. In future

investigations it will be important to expand our hypothesis to other effector functions like

cellular cytotoxicity and granzymeB production and titrate the reactivation stimuli to find out

the upper limit of re-activation intensity that propels the PD-L1 inhibited T cells to recover to a

level that is comparable to uninhibited and PD-L2 inhibited T cells.

Figure 8: PD-L2 inhibition through PD-1 allows T cells to functionally recover more

readily than PD-L1. (a) PD-ligand mediated inhibition and recovery protocol as described in

detail in methods. (b) Evaluation of cytokine levels in the supernatant immediately after the

72 hours co-culture with PD-ligands by cytokine bead array. (c) Evaluation of cytokine levels

by using a cytokine bead array after re-activation of T cells using CD3/CD28 T cell activator

beads. Statistical significance was calculated using a one-way ANOVA. *p< 0.05, **p<0.01,

***p<0.001, ****p<0.0001. N= 2 donors n=3 technical replicates.

Chapter 4: Discussion

This dissertation investigated the unique role of PD-L2 in modulating human T cell responses

and discovered the following:

1) Human PD-L2 does not exhibit the T cell co-stimulatory function observed in mice.

2) PD-L2 engages PD-1 to induce a distinct signaling repertoire in human T cells.

3) PD-L2 mediates a weaker inhibitory signal through PD-1 than PD-L1 in human T cells.

4) PD-L2 inhibited T cells are able to functionally recover more readily than PD-L1

inhibited T cells.

As of February 2023, the therapeutic landscape of PD-1/PD-L1 inhibitors consisted of 180+

pharmaceutical companies with 200+ candidates in various stages of drug development[128].

The current arsenal of anti-PD-1/PD-L1 monoclonal antibodies comprises of 7 different drugs

that are FDA-approved for treatment across 24 cancer types[6]. While immune checkpoint

inhibitors revolutionized the field of immuno-oncology it is also important to acknowledge that

innovation and improvement of novel therapeutics against the PD-1 signaling circuit can be

achieved only through a comprehensive understanding of the underlying biology.

Since its discovery in 2001, PD-L2 was assumed to have overlapping immune modulatory

functions with PD-L1 until a second binding partner of PD-L2, repulsive guidance molecule b

(RGMb) was reported in 2014[129]. Thereafter the role of PD-L2 in T cell responses was

controversial because both coinhibitory and costimulatory functions were attributed to the

ligand. Chen et al. showed that a recombinant mutant PD-L2 protein (K113S) that retained its

binding affinity to RGMb but not to PD-1 co-stimulated CD4+ T cell responses to promote Th1

polarization in a murine model for asthma[125]. In parallel, other research groups published

data that denied any dualism in the function of PD-L2 and emphasized PD-L2 as a coinhibitory

ligand to T cell responses[130–132]. Researchers overwhelmingly relied only on PD-L1 as

the inducer of PD-1 in their investigations and as a result the therapeutic potential of PD-L2

remained in the shadows of PD-1/PD-L1 blocking agents. The contentious function of PD-L2

may have contributed to the singular focus of the field on PD-L1 as the cognate ligand of PD1.

I began my PhD work with the general hypothesis that the restricted expression pattern of PD-

L2 in normal human physiology and cancer combined with its unique biophysical properties

underscore PD-L2 as an independently important ligand of the PD-1 signaling axis. My

dissertation demonstrated that PD-L2 is a non-redundant ligand of human PD-1 that inhibits

T cell activation through a mechanism that is distinct from PD-L1. We further characterized

the distinct inhibitory signal and found that PD-L2 inhibited T cells showed higher levels of

apoptosis when compared to PD-L1 and demonstrated slower temporal kinetics in the

dephosphorylation of the TCR-CD3 signaling axis. Additionally, we also showed that the PD-

1/PD-L2 interaction unlike PD-1/PD-L1, allowed T cells to progress through the cell cycle and

avoid being arrested in the G1/G0 phase. Our work challenged prior scientific literature that

assumed a redundant function for PD-L2. We were the first to discover that human PD-L2,

despite its higher binding affinity transmitted a weaker inhibitory signal through PD-1, which

in turn allowed the inhibited T cells to functionally recover more readily when re-exposed to

activating signals. We were also the first to provide experimental evidence that human PD-L2

similar to PD-L1 engaged the ITSM region on the cytoplasmic tail of PD-1 to inhibit TCR-CD3

activation. Through our work we also emphasized the incongruence between mice and human

in the context of PD-L2 immunobiology and cancer. We demonstrated that human PD-L2 did

not retain interaction with RGMb as observed in mice. In the course of our investigation, we

also found that co-expression of PD-L1 and PD-L2 on the same cell may lead to the formation

of a heterodimer that can induce a unique PD-1 inhibitory repertoire different from the

monomer ligands.

In the following pages I will discuss the impact of this dissertation and how our findings laid

the foundation for future investigations to harness the therapeutic potential of PD-L2 in

immunotherapy.

4.1. The paradoxical role of PD-L2: Mouse PD-L2 Vs Human PD-L2

Even though in recent years PD-L2 gained importance as a potential prognostic biomarker in

human cancers, immunologists hesitate to target the ligand to promote anti-tumor

immunity[126,133–139]. Previous studies of PD-L2 blockade failed to demonstrate

therapeutic benefit in orthotopic as well as transplantable murine tumor models[124]. The

clinical prospect of PD-L2 blockade was further dampened by reports that showed possible T

cell co-stimulatory functions of PD-L2 in mice[125,129]. Murine PD-L2 binds to RGMb to form

a signaling supercomplex with other proteins such as BMP and neogenin to facilitate the initial

T cell expansion in draining lymph nodes during respiratory tolerance induction. In a different

study, RGMb independent of PD-L2 was implicated in the development of airway inflammation

and airway hyperreactivity (AHR)[79]. A decade ago, a report demonstrated increased

expression of PD-L2 in the airways of asthmatic individuals and was correlated with severity

of asthma. However, the impact of PD-L2 as a causative or consequential factor of airway

hypersensitivity was not clarified. IL-4 is known to drive AHR and has long been associated

with pathogenesis of allergic disorders[140]. IL-4 is also known to selectively induce the

expression of PD-L2 but not PD-L1[141]. Preclinical investigations of PD-L2 and its role in

modulating immune responses in mice have been contradictory and assumed to be reflective

of its role in humans.

Our study was the first to demonstrate that mouse and human PD-L2 have functionally

different roles. We clearly showed that human PD-L2 binds to only PD-1 on T cells to inhibit

TCR-CD3 activation in human T cells. In our studies we primarily focused on sub-optimal

levels of T cell activation in the absence of CD28 co-stimulation in an attempt to replicate the

tumor microenvironment where co-stimulatory ligands are preferentially downregulated[142].

It will be interesting to evaluate the impact of PD-L2 on T cell responses in the presence of

CD28 co-stimulation since PD-1/PD-L1 complex is known to dephosphorylate the CD28

signaling axis over the TCR-CD3 complex[143]. We also showed that human PD-L2, unlike

mouse PD-L2 did not bind to RGMb.

While our results have conclusively shown that human PD-L2 does not bind to RGMb, future

studies exploring the possibility of new partners for PD-L2 on naïve T cells is warranted. Since

we tested the physical interaction between PD-L2 and RGMb exclusive of other proteins, one

can argue that PD-L2 and RGMb may need to form a complex with BMP/Neogenin to form

stable physiologically relevant interactions[129,144]. Even if it was true that we

underestimated the human PD-L2/RGMb interaction or the presence of alternate co-

stimulating binding partners of PD-L2, the widespread expression of PD-L2 in human cancers

reported in several research articles as well as in our analysis of cancer cell lines support our

result that PD-L2 is a solely inhibitory ligand in humans. This is in contrast to murine cancer

cell lines that do not express PD-L2 even when induced by cytokines. PD-L1 and PD-L2 as

members of the B7/CD28 superfamily of proteins are part of the paradigm that tumor cells

preferentially downregulate the expression of T cell co-stimulatory ligands and upregulate the

expression of co-inhibitory ligands to evade the immune system. Many human malignancies

associated with poor prognosis co-opt the expression of PD-L2 over PD-L1 that further denies

any potential costimulatory function in humans[126]. PD-L2 expression in human cancers was

proved to play a role in tumorigenesis and immune escape as well. In head and neck single

cell carcinoma (HNSCC) and oesophageal cancer patients PD-L2 expression correlated to

poor relapse free survival (RFS) and progression-free survival (PFS)[38,133,135]. PD-L2 is

also reported to be associated with neuroinvasion and low CD8+ TIL density in colon

cancer[145]. In the last 5 years, the concept of PD-L2 as a prognostic biomarker in human

cancers is gaining momentum, and our results clarifying the role of human PD-L2 further

bolsters this momentum.

Published studies delineating the biophysical properties of PD-L2 supported our conclusion

that mouse and human PD-L2 differ in their nature of regulation of T cell

responses[61,89,146]. Even though mouse PD-L2 shares ~70% amino acid sequence identity

with human PD-L2 it still differs in its mechanism of interaction with the PD-1 protein[147].

Human PD-L2 binds to PD-1 through a subset of amino acid residues that create a prominent

pocket absent in mouse PD-1/PD-L2 complex and provides for an attractive small-molecule

drug target[146]. Future investigations delineating the molecular and structural basis for PD-

1/PD-L2, PD-L2/RGMb, PD-1/PD-L1 interactions will help define potential hot-spots for

therapeutic interventions and curate the disruption of protein-protein interactions. Our findings

highlighted the limitations of traditional transgenic and transplantable mouse models of cancer

in the context of PD-1/PD-L1/PD-L2 immunobiology. The co-stimulatory function of murine

PD-L2 lacks relevance in human biology and as a result the translational opportunity of PD-

L2 blockade is misrepresented in traditional mouse models. Our work provides evidence in

support of the use of humanized mice to develop physiologically relevant preclinical models

in the future to optimally target the PD-1 signaling axis for clinical benefit.

4.2. The non-redundant role of PD-L2: Human PD-L1 Vs Human PD-L2

In normal human physiology PD-L1 is widely expressed in hematopoietic and non-

hematopoietic cells but PD-L2 is largely restricted to antigen presenting cells and the female

reproductive tract. Despite the 60% homology of amino acid residues between human PD-L1

and PD-L2, the ligands interact with different binding partners[60,89]. While human PD-L1

binds to CD80 and PD-1, PD-L2 interacts with only PD-1 on the surface of activated human

T cells[64]. Preliminary data in the lab demonstrated that PD-L1 and PD-L2 can interact with

each other in vitro.Previous surface plasmon resonance studies have demonstrated

significant differences in the association and dissociation characteristics of PD-1/PD-L1 and

PD-1/PD-L2 complex formation[89]. The distinct expression pattern of the ligands along with

differences in the structural mechanisms of interactions with cognate receptors argued

strongly in favor of non-overlapping roles of PD-L1 and PD-L2 in immune regulation.

Specifically in the context of anti-tumor immunity, the differential function of mouse and human

PD-L2 reported earlier in our study enabled us to speculate whether selective pressure during

evolution may have led to non-redundant PD-L1 and PD-L2 function in humans. A recent

study supported our theory and showed that PD-L2 evolutionarily originated from a gene

duplication event that diverged from PD-L1 at the level of placental mammals[89,148]. In the

past decade several research groups independently discussed the structural differences

between PD-1/PD-L1 and PD-1/PD-L2 complexes, but none could determine the molecular

and physiological consequence of the differences. We were the first to uncover that PD-L1

and PD-L2 engage PD-1 on human T cells to elicit distinct molecular mechanisms of inhibition

downstream to PD-1 in human T cells. Our discovery of differential signaling patterns induced

by PD-ligands was demonstrated through a robust multi-platform investigation of in vitro and

ex vivo assays of human T cells.

Key to understanding the PD-1 mediated inhibition is uncovering the mechanism underlying

the differential signaling mediated by PD-1. Is the differential signaling the result of unique

structural conformations of the PD-1/PD-L2 complex that induced unique signaling partners

downstream of PD-1? Or is it the functional consequence of a difference in the temporal

kinetics of PD-L1 and PD-L2 mediated PD-1 inhibition? One way we addressed this issue is

by performing site-directed mutagenesis of the signaling domains, ITIM and ITSM in the

cytoplasmic tail of PD-1. PD-L1 is known to engage SHP-2 phosphatase through the ITSM

domain of PD-1 to inhibit T cell activation, and the same was assumed to be true for PD-

L2[57,67,69]. However, our theory that PD-L2 preferred ITIM over ITSM was disproved when

we found that mutagenesis of Y248 in ITSM domain and not Y223 in ITIM domain reversed

PD-1 mediated inhibitory function. Even though in vitro the ITIM is known to activate SHP-2

like the ITSM, the failure to readily inhibit T cell indicates towards a possible conformational

change of PD-1 when bound to a ligand that limits the inherent potential of ITIM to recruit

SHP-2. Alternatively, it may be possible that the PD-1/PD-L1 complex induces ITSM to recruit

SHP-2 whereas PD-1/PD-L2 induces ITSM to recruit another family member such as SHP-1

or SHIP. This theory is supported by studies that evidenced SHP-2 independent mechanism

of PD-1 inhibition in T cells[70]. PD-1 is a transmembrane protein that engages its two known

ligands in unique conformations and a signaling competent cytoplasmic tail is the spark that

ignites the distinct signaling repertoires downstream. Our collaborator Dr. Wan-Lin Lo at the

University of Utah will perform co-immunoprecipitation assays with Jurkat T cells expressing

mutant forms of PD-1 to investigate the molecular partners bound to the cytoplasmic tail of

PD-1 when engaged by PD-L1 or PD-L2.

After validating our “omics” findings in functional assays we concluded that human PD-L2

binds to PD-1 to mediated weaker inhibition than PD-L1 resulting in a distinct signaling

repertoire downstream to PD-1. It is counterintuitive that a higher affinity ligand induces a

weaker signal through the same receptor. A possible explanation can be attributed to the 2-6

fold higher affinity of PD-L2 to PD-1 and the ability of PD-L2 to recruit PD-1 to the TCR

microclusters more potently than PD-L1[149]. Even though previous studies have opined that

the 2-6 fold higher affinity is insufficient to drive differential signaling, we speculate that the

significantly slower dissociation rate of PD-L2 off PD-1 may support the feedback loop of SHP-

2 mediated dephosphorylation of PD-1 cytoplasmic domains resulting in a weaker inhibitory

signal by the PD-1/PD-L2 complex[150]. Since our hypothesis on the role played by the

cytoplasmic domains in the differential signaling did not hold true, the mechanism underlying

the distinct PD-1/PD-L2 signaling remains an open question for further investigation.

4.2. The physiological relevance and therapeutic potential of PD-L2

Ultimately, it was pivotal for our study to elucidate the physiological relevance of the weaker

and distinct inhibitory signal induced by PD-L2 to inform future therapeutic interventions. We

discovered that the PD-1/PD-L2 complex inhibits T cells to a lesser extent and allows them to

be recovered more readily than the PD-1/PD-L1 mediated deeper inhibition of T cells. Our

findings showed that in vitro human T cells inhibited by PD-L2 recovered and produced higher

levels of IL-2 and TNFα that were comparable to uninhibited T cells. Future investigations

focused on evaluating the impact of PD-L2 on the ability of tumor infiltrating T cells to regain

anti-tumor potential will be seminal for immunotherapy against cancer. The more reversible

anergy mediated by human PD-L2 may also have a role to play in physiological contexts other

than cancer like chronic viral infection, mucosal immunity, and autoimmunity[83,151]. Our

findings may also help delineate the functional consequence of PD-L2 expression in maternal

fetal tolerance (MFT) since PD-L2 like PD-L1 is expressed in the placenta and is finely tuned

as the pregnancy progressed through the trimesters[84,85,88].

Our study primarily focused on the impact of PD-1/PD-L2 signaling on T cell responses and

provided the rationale to investigate the presence and consequence of the differential

signaling mediated by PD-L2 through PD-1 in other immune and non-immune cells as well.

More recently the expression of PD-1 has been appreciated in Tregs, tumor cells, NK cells,

dendritic cells and myeloid cells like macrophages[59]. The precise function of PD-1 signaling

axis in each of these cells is yet to be delineated. We provide the premise to perform these

future investigations using PD-L1 as well as PD-L2 as the inducer of PD-1 in their

experiments. This will ensure that we are not underestimating the differential role PD-1

engagement by PD-L1 and PD-L2 in cells other than T cells.

Even though it’s been more than 20 years since its discovery, there remains no disclosed

effort to directly target human PD-L2 therapeutically. This may be due to preclinical murine

models that inadequately represent the function of human PD-L2 or the assumed functional

redundancy with PD-L1 that our study conclusively debunks. Through our work we

established the rationale for PD-L2 to share the spotlight with PD-L1 to achieve optimal anti-

tumor immune responses in the clinic. Can higher expression of PD-L2 in human cancers be

correlated with increased potential of tumor infiltrating T cells to recover its cytotoxic potential

on account of the more reversible anergy mediated by PD-1/PD-L2 interaction? Does this

proposed mechanism explain several recent scientific studies that demonstrate the ability of

PD-L2 expression in human cancers to be more predictive of response to PD-1 blockade

therapy than PD-L1? In the last decade novel antibody-engineering techniques enabled

scientists to engage the innate arm of the immune system (NK cells) to deplete the

immunosuppressive stromal cells in the tumor microenvironment and support cytotoxic T cell

tumor infiltration and anti-tumor function[152]. The restricted expression pattern of PD-L2 on

tumor cells and immunosuppressive stromal cells gives us an opportunity to arm our blockade

antibodies with antibody dependent cellular cytotoxicity potential (ADCC) and strategically

reprogram the tumor microenvironment to support anti-tumor immunity.

In conclusion, the work demonstrated in this dissertation is to our knowledge the first study to

discover the unique mechanisms of PD-1/PD-L1 and PD-1/PD-L2 mediated inhibition of T cell

function. Our findings broadened the field’s understanding of the PD-1 signaling axis and may

have implications in other PD-1 expressing cells as well. We lay the foundational block for

future investigations to study the PD-1 signaling axis in the context of both ligands, PD-L1 and

PD-L2 to gain a comprehensive perspective of the underlying biology. Our study did not

consider the recently appreciated putative signaling domains in the cytoplasmic tails of PD-

L1[153,154]. While we studied the influence of the differential engagement of PD-1/PD-L1 and

PD-1/PD-L2 downstream to PD-1 in T cells, future studies should investigate the impact on

the downstream signaling of PD-L1 and PD-L2. We demonstrate that the PD-1/PD-L1/PD-L2

signaling axis is not binary in its function, instead PD-1 can suppress T cell function to varying

degrees of inhibition depending on the ligands available for ligation. We are the first to show

that PD-L1 induces T cells into a deeper state of exhaustion whereas PD-L2 weakly inhibits

the T cells to a more reversible state of anergy. In the future it will be interesting to see the

relevance of PD-L2’s reversible state of anergy in the context of tumor immunology. Can we

reverse the inhibition mediated by PD-L2+ tumors more readily on account of the reversible

state of anergy? As we gain deeper insights into the structural and mechanistic biology of the

PD-1/PD-L1/PD-L2/CD80 signaling axis, we will be able to rationally curate future

therapeutics against cancer such that we disrupt specific arms of the signaling axis while

retaining those that will protect the patient from immune related adverse events.

Investigating The Distinct Role Of Pd-L2 In Modulating Human T Cells Responses

Students also viewed