How a Proposed Hypothesis during My PhD Training Shaped My Career

Abstract

In this memoir-style essay, I have narrated the evolution of my scientific career, as deeply influenced by my PhD training and the mentorship of Professor Eli Sercarz. Starting in his lab, and continuing to my own laboratory, many of the questions we have pursued link in some way to Eli’s ideas. In this essay, I have summarized the path that I followed after graduating from his lab and highlight findings along the way. I apologize to my colleagues whose work was not discussed here due to the nature of this review and space limitations.

I. PROLOGUE

My love for experimental science began when I first listened to the story of T and B cells told by Farrokh Modabber, my very first professor of immunology at Pahlavi Medical School (currently Shiraz University), who himself had been a student of Professor Eli Sercarz, who at the time was an Assistant Professor of Immunology at Harvard University in Cambridge, Massachusetts. As a young student, I was so captivated by the intricacies of the immune system that I decided to take a different route than medical school and become an immunologist. It took a couple of years for me to complete a biology degree, move to Teheran, and get a job at the Influenza Research Laboratory at Teheran University as a technician, for my first taste of research. And, it was only by sheer luck that Farrokh Modabber was recruited to the very department as the chairman and the director of a master’s program in immunology, jointly taught by professors at Harvard and Teheran University. I could not have asked for a better opportunity than that! I joined the program and graduated in two years. The next step was to continue in a PhD program. UCLA and Eli Sercarz’s laboratory seemed the most appealing option for me. Hence, I feel that I am both Eli’s scientific child and grandchild!

Eli Sercarz’s lab ran like a big family, consisting mainly of intense postdocs and a handful of graduate students. The lab was quite big, occupying the fourth floor and almost the entire west wing of the biology building at UCLA. Dr. Alexander Miller and Eli jointly advised the students and postdocs. While Eli was very imaginative, with tens of ideas at a given time, Alex had his feet on the ground. Thus, their two opposite personalities made a well-balanced force in discussions. As such, the lab was extremely interactive and productive. Whenever a new relevant paper was published, someone would start discussing it with another lab member, which would quickly attract others nearby and before you knew it, there was a big crowd listening and hashing out the paper. These gatherings could pop up spontaneously sometimes several times in a day with all the lab members participating. In such a fertile scientific environment, one could not help but learn and think about scientific problems deeply, and from different angles.

In such a training environment, science was a deep commitment, and the lab members became not only your colleagues, but also your friends—sometimes your only friends. Indeed, to date, two of my closest friends are from that time. We had many outings, including sailing trips and house parties, and most importantly, when Eli would invite a speaker, the entire lab would go along for dinner with the speaker. In short, in Eli’s lab, we embraced our beings as developing scientists; there was no separation between work and fun, because the environment of the lab made learning the central goal and the most enjoyable activity in life.

II. ELI’S CONTRIBUTIONS TO IMMUNOLOGY

Many pioneering concepts have come out from Eli Sercarz’s laboratory; many of which remain subjects of intense studies to date. Some examples are highlighted below.

1. Formulated the X-Y-Z scheme of immunocyte maturation, which indicated that memory induction involves a qualitative change of state of the naïve lymphocyte to a memory cell (Y) that can be driven irreversibly to a Z state (termed exhaustion) (Immunological Tolerance, pp. 73–83, 1961; Strategies of Immune Regulation, pp. l99–2l0, 1980).

2. Was the first to use congeners of a protein to study Ir gene control. Thus, H-2^b mice were able to respond to REL (ring-necked pheasant lysozyme) and TEL (turkey lysozyme) but not to HEL or BEL (bob-white quail lysozyme) (Eur J Immunol 1975;5:317–324).

3. Defined different public idiotypic determinants affecting primary and secondary responses to HEL (Nature 1980;287:541–542; J Immunol 1980;125:1017–1025).

4. Was first to show that immunodominance involved a selective engagement of only a part of the Ag-specific repertoire (Eur J Immunol 1982;12:535–540).

5. Showed that a response to a single antigenic determinant was complex and comprised of T cells with differing fine specificities. Later, this was refined to show that the T cells directed against the distinct specificities of a short 12-mer peptide used distinct Vb genes for each specificity and the response for each specificity was limited (J Immunol 1984;133:2075–2078).

6. Showed that distant or immediate flanking regions of determinants on a well-characterized protein (HEL) could affect processing and presentation (Ann Immunol 1984;135:14 and Discussion, p. 39; J Mol Cell Immunol 1985;1:369–377).

7. Tolerance is induced to dominant determinants and sometimes to subdominant determinants, but the T cell repertoire to most subdominant and cryptic determinants remains intact (Nature 1989;342:183–185; Proc Natl Acad Sci U S A 1992;89:416–420).

8. Used pepscans to define dominant and cryptic T cell determinants in several mouse haplotypes (J Exp Med 1991;173:609–617).

9. Showed that during alloresponses, a major part of the response is indirect, and specific for allo-MHC peptides seen in a self-context (J Exp Med 1992;175:305–308).

10. Showed that there can be holes in the T cell repertoire: because of the absence of half of the T cell repertoire, a response to a determinant can be lost (Proc Natl Acad Sci U S A 1991;88:9503–9507).

11. Showed that autoimmune response can be initiated by a very focused response to a single determinant, which spreads to include other determinants on the same antigen (intramolecular spread) as well as determinants on other antigens (intermolecular spread) (Nature 1992;358:155–157; Immunol Today 1993;14:203–208).

12. Showed that autoimmune response depends on driver clones, and when they are downregulated the spreading stops. Likewise, diversification can occur to involve regulatory complementary determinants, which then limits the spread.¹

13. Showed that responsiveness to the native form of a autoantigen such as mouse lysozyme (ML) is absent while response to its peptides exists, especially among peptides that bind poorly to the MHC (J Immunol 1999;163:4232–4237).

14. Showed that Th1/Th2 choice can depend on the affinity of a determinant for the MHC. In a system where the affinity for MHC depends on a single residue and can differ over a 10,000-fold range, high affinity peptides induce a Th1 response in vivo and low affinity peptides, a Th2 response (Proc Natl Acad Sci U S A 1995;92:9510–9514).

15. Showed that determinant capture is a mechanism involving MHC-II competition for binding to a site on an unfolding antigen. A determinant on the long peptide binds to the MHC-II molecule with the greatest affinity, thereby preventing a different MHC-II from binding. Trimming then follows to yield the shorter peptides (J Exp Med 1993;178:1675–1680).

16. Showed that competitive determinant capture can also preempt tolerance induction and thereby protect T cells directed against the preempted determinant from negative selection (J Invest Med 1996;44:234A; Proc Natl Acad Sci U S A 2003;100(9):5342–7).

17. Showed that the T cell receptor is polyspecific or degenerate, in that it can combine with many distinct ligands. This furnishes an explanation for thymic positive selection, for molecular mimicry, for the high frequency of alloreactivity, etc. (J Immunol 1993;151:5000–5010; J Exp Med 1995;182:531–39; J Autoimmun 2001;16:201–209).

Because of his creative mind, Eli enjoyed inventing new terminologies and concepts, and introducing them to the immunology community.

The terminologies and concepts introduced by Eli Sercarz include:

1. Immune exhaustion

2. Determinant capture

3. Determinant spreading in autoimmunity

4. T cell suppressor–inducing determinants

5. Processing creates the immunologic self

6. MHC-guided antigen processing

7. Determinant cores and envelopes

8. TCRs exhibit degenerate specificity

9. Only dominant determinants tolerize; thus T cell repertoires versus cryptic and subdominant determinants accounts for autoimmunity

III: MOLECULAR STUDIES ON ANTIGEN PRESENTATION USING MINIMALISTIC APPROACHES

My own laboratory is among those that have devoted their research to understanding some of the concepts Eli introduced. By the end of my PhD thesis, I came up with a hypothesis that differential antigen presentation could prompt different responses in T cells. Hence, T cell activation or suppression/tolerance could be triggered by differences in antigen presentation, an active area of high interest in my lab to date. After trying to address the question by cellular immunology methods and facing much frustration during a short postdoctoral training, it became clear to me that the best way to address the question was a direct approach. Hence, I aimed to better understand the molecular basis of antigen presentation. To do just that, I joined the laboratory of Professor Harden McConnell in the Chemistry Department of Stanford University who had just published a series of papers demonstrating that for activation of T cell hybridomas, purified MHC II inserted in planar membranes and added specific peptides were all that was needed.² This was the first demonstration of the minimal requirements for T cell activation.

Harden McConnell’s lab, also mainly run by postdocs, was very different from Eli’s; it was a multidisciplinary lab. There were several groups of postdocs trained in vastly different areas of science: mathematicians, engineers, NMR spectroscopists, high power laser fluorescence microscopists, and more. Only a small group of two to three fellows worked on immunology problems. We did not have regular lab meetings; each of us directly talked to Harden and discussed our research. The field of antigen presentation was soaringly competitive at that time. So, to survive his lab, I had to learn biochemistry and physical chemistry in a very short time! My project was to study the interactions of fluorescent antigenic peptide with purified MHCII, I-E^k molecules inserted in planar membranes using fluorescent peptides. The results were surprising: we could find that binding of peptides to purified MHC II involved kinetic intermediates and had a fast phase and a slow phase. We interpreted our observations by thinking that some conformational changes in MHC II molecules must be necessary for the slow phase to follow.³

In my next venture, in Dr. Ronald Germain’s lab at NIH, we continued the project and showed that the conformational changes in MHC II could indeed be detected in a simple SDS-PAGE assay where samples were not boiled.⁴ The assay was developed in McConnell’s laboratory where it was shown that MHC II molecules migrated in SDS-PAGE gel differently when dissociated from the bound peptide.^5,6 We were able to show that the conformation of MHC II and its folding depended upon binding of the antigenic peptides. As such, MHC II molecules that were tightly bound to peptides migrated as compact dimers whereas MHC II molecules with loosely bound peptides migrated as larger size complexes and were called floppy dimers, or fell apart as single α and β chains.⁴ Importantly, these findings were verified in vivo. Newly synthesized murine MHC II molecules bound to invariant chain fell apart in SDS gel (SDS-sensitive), whereas, mature peptides bound to MHCII formed compact dimers and were SDS-resistant.^7–9 It was also shown that SDS stability did not always correlate with the stability of pMHC complexes. Altered MHC mutants bound peptides loosely yet formed the characteristic SDS-stable conformation.^10,11 The remarkable characteristic of MHC class II to resist SDS denaturation when in complex antigenic peptides allowed new discoveries that revealed steps in MHC class II synthesis, association with invariant chain (Ii), exposure to antigen-processing enzymes, MHC II trafficking, interaction with accessory molecules, peptide binding and editing, and more as reviewed in detail.^12,13

A. Conformational Changes in MHC II upon Peptide Binding as the Guiding Principle for Epitope Selection

Later investigations in my independent laboratory demonstrated that for MHC II molecules to bind a new peptide, the molecule had to undergo conformational changes to release the previously bound peptide and convert to a peptide-receptive conformation.^14,15 Once in peptide-receptive conformation, MHC II could bind to new peptides spontaneously. These findings provided the foundation for how MHC II chaperone/editor, HLA-DM (DM), could induce peptide exchange, replacing the class II invariant chain peptide (CLIP) or a poorly bound peptide with a new peptide.^15,16 It became clear that the groove of MHC class II collapses in the absence of a bound peptide and a poor binding peptide, such as CLIP, maintains the groove in proper conformation that can scan peptides or unfolded proteins in the antigen-processing compartments rapidly.^14,15 Thus, in addition to the well-appreciated roles for Ii in preventing unproductive binding of self-peptides present in the ER,¹⁷ and in accompanying the newly synthesized MHC class II to the proper endocytic compartments to intercept with protein antigens,¹⁸ a new and critical role for CLIP as a “place-keeper” became evident.

B. Mechanism of DM in Finding the Immunodominant Determinants during Antigen Processing

As mentioned above, the flexibility of the MHC II groove became the guiding principle for binding to immunogenic peptides. DM (HLA-DM in human, H-2M in mice) is a critical accessory protein in first dissociating CLIP from the newly synthesized MHC II, and then help the molecule to find its suitable peptide. DM does this job by dissociating any peptide that does not fill in the hydrophobic pockets of MHC II groove. But when a sequence of antigenic determinant that would fit the MHC II groove (HLA-DR1) forms a rigid or compact conformation, the complex becomes resistant to DM-mediated dissociation. And, determinants that do not fit the groove and leave the critical P1 pocket of HLA-DR1 empty are susceptible to DM-mediated dissociation.^16,19–27 With the solution of the crystal structure of the DM/DR complex in 2011²⁸ using a cleverly designed DR1/peptide complex that allowed for the DR1 groove to remain open, it became clear that DM would bind the P1 pocket of HLA-DR molecules if empty, and would remain bound until a P1 filling peptide binds the groove, inducing closing of the groove that would displace DM.^28–30 The above findings were complemented by the measured thermodynamics of peptide binding to DR1, indicating that a greater entropic penalty, versus a smaller penalty, was associated with structural rigidity rather than with the flexibility of the pMHC complexes.³¹ These findings suggested that an overall dynamic MHC II conformation, in addition to P1 pocket occupancy, determines susceptibility to HLA-DM-mediated peptide exchange and provides a molecular mechanism for HLA-DM to efficiently target poorly fitting pMHC II complexes, editing them for more stable ones. Hence, in addition to the removal of CLIP, DM helps in shaping epitope selection (more details to follow).

C. Biological Significance of DM

As discussed earlier, DM plays an important role in selecting the right peptides that can stay in the groove of MHC II long enough for T cell recognition.³² This characteristic of DM contributes to T cell immunity in a significant way. Lymphocytes usually respond to a small proportion of the potential determinants on a protein antigen defined as immunodominant.³³ Immunodominant epitopes are the essential targets of the immune response against infectious diseases, cancer, autoimmune diseases, and allergy. Consequently, much attention has been devoted to the understanding of epitope selection and immunodominance. However, in spite of the complexities of antigen processing and presentation, T cell epitope discovery has been a challenging task. Some of the factors contributing to immunodominance are described below.

D. A Reductionist Antigen Processing System Helps in Teasing Out Contributing Factors to Immunodominance

A better understanding of immunodominant epitope selection during antigen processing became feasible by the development of a reductionist antigen processing system with a minimal number of purified proteins.³⁴ This system provided clear evidence that antigen processing and immunodominant epitope selection could be achieved in a tube, which was verified by identification of immunodominant epitopes from several proteins implicated in autoimmunity as well as protein antigens from influenza, malaria, and HIV-1.

The reductionist system helped to tease out several outstanding questions in antigen processing and presentation. We showed that processing of pathogen-derived epitopes was different from autoantigen processing and epitope dominance. First, we learned that immunodominant epitopes of pathogen derived antigens were highly sensitive to proteolysis, unless bound and protected by the groove of MHC II. Hence for a successful processing and presentation of the immunodominant epitopes, antigenic determinants must first be captured by MHC II, and then proteolyzed by endo- and exoproteases.^35,36 These observations contradicted the prevailing dogma that immunodominant peptides are first subjected to proteolysis by antigen processing enzymes, cathepsins, and then bound to the MHC II groove with help of DM.³⁷ As such, these observations also placed focus on structural factors such as the location of the epitopes, and their accessibility to being captured by the groove of MHC II molecules.^38,39 Of great significance, we learned that the dominant epitopes were produced in higher abundance relative to other non-dominant epitopes.^35,36,40,41

E. Epitope Accessibility and Its Relation to Immunodominance

Among many contributing factors to an epitope gaining immunodominance is how accessible the location of sequence is to the groove of the MHC II molecule and/or to the processing enzymes.³⁸ The denaturing environment in the antigen processing compartments (acidic pH and reducing conditions) helps to partially unfold protein antigens to reveal hidden epitopes. Of particular interest is a specialized enzyme, gamma-interferon-inducible lysosomal thiol reductase (GILT), that releases disulfide bonds in proteins,⁴² making denaturation more efficient. Support of the “epitope accessibility” model for immunodominance comes from accumulated evidence that many of the naturally selected epitopes localize on the flexible strands of protein antigens³⁸ or at the C- or N-terminus of protein antigens.^43–45 For a more comprehensive review on the subject of accessibility, the readers are referred to reference.⁴⁶

F. DO (HLA-DO Human; Murine H2-O)

A second Class II accessory molecule, DO (HLA-DO human; murine H2-O), mainly present in B cells and thymic epithelium, also contributes to the selection of immunodominant epitopes. Yet, despite a wealth of mechanistic insights into how DM functions, understanding the contributions of DO to epitope selection has proven to be highly challenging.^47–50 In short, our knowledge about DO can be distilled into two working hypotheses: (1) DO binds to DM to inhibit its activity, mainly removal of the CLIP peptide, and (2) DO differentially affects presentation of structurally diverse peptides and acts as a second chaperone together with DM in fine tuning MHC II repertoire selection. Data in support of the former hypothesis mainly comes from studying overtransfection of DO genes in cell lines, or dendritic cells,^49,51,52 and the recent structural studies of DM/DO interaction. The 3D structure of DM/DO showed that DO binds to DM at the same interface with which DM interacts with DR1.^30,53 Studies supporting the latter hypothesis came from biochemical⁵⁴ and kinetic studies demonstrating that DO only affected association of certain peptides to DR, but had no effect on the dissociation of any tested peptide/DR1 complexes.^50,55 The effect of DO on association directly correlated with peptide sensitivity to DM-mediated dissociation. DO reduced binding of peptides that formed DM-sensitive complexes with DR, and enhanced the binding of peptides that formed DM-resistant complexes. In a nutshell, it was clearly shown that (1) DO works directly on DR1, and not by regulating the effect of DM; (2) DO can only bind the peptide-receptive MHC Class II; and (3) that this peptide-receptive conformation is generated by DM. Hence, we proposed that DM and DO cooperate for a more effective epitope selection. Thus, in one model, DO would reduce presentation of immunodominant epitopes, whereas in the other, DO would increase the abundance of immunodominant epitopes.

Using a combination of various in vivo approaches, including peptide elution, mixed lymphocyte reaction, T-cell receptor (TCR) deep sequencing, tetramer-guided naïve CD4 T-cell precursor enumeration, and whole-body imaging, we reported that DO affects the repertoire of presented self-peptides by B cells and thymic epithelium. DO induces differential effects on epitope presentation and thymic selection, thereby altering CD4 T-cell precursor frequencies. Importantly, by comparing pMHC presentation to specific T cells by APCs from EAE diseased brain from mice that either lacked DO, or were DO sufficient,⁴⁷ we demonstrated that the DO sufficient APC presented higher numbers of MOG peptide/I-A^b compared with DO-knockout APCs. Consistent with our quantitative model, we learned that presence of DO in B cells promotes selection of self-peptides of higher kinetic stability. In addition, the study by Welsh et al. offered substantive in vivo evidence that the selected expression of DO in thymic medulla reduces self-reactivity.

Because of the significance of determinant selection and immunodominance,^34,35 it makes sense that two accessory molecules be at play to ensure optimal selection of immunodominant epitopes from autoantigens in the thymic medulla and in B cells where DO is mainly expressed. The question remains as to what advantages are conferred by the regulated expression of DO in B cells? The above understandings of DM and DO and the quantitative and qualitative differences in display of pMHC during different stages of immune development and differentiation took us back to my graduate school days, and the hypothesis that differences in antigen presentation by different APCs prompts different responses in T cells. I will discuss this more in the upcoming section.

G. The Cell Free Ag Processing System Reveals Epitope Hierarchy

The phenomenon of epitope hierarchy, originally described by Sercarz proposed that while many epitopes within a protein antigen can bind to the groove of MHC II, some gain dominance over the others, hence the terms subdominant or cryptic epitopes were introduced to the immunology world.³³ Cryptic epitopes were defined as epitopes that would only elicit a T cell response if the immunodominant epitopes had been removed or altered. One molecular mechanism envisioned in explaining the phenomenon was that MHC II binding to antigenic determinants could occur prior to peptide generation in the endosomal compartments,^56,57 a concept that was directly supported by the use of our reductionist system. We were able to clearly demonstrate that immunodominant epitopes bind to MHC II as either (1) part of the full antigen or (2) in a large fragment of the parent antigens, and DM helps with the selection of the right determinants.^35,36

H. Artificial Spacer Sequences Designed to Facilitate Recombinant Protein Purification Can Emerge as Immunodominant Epitopes

The competition to gain dominance among potential epitopes of an antigen as described by Sercarz and colleagues considered only the natural antigen sequences. However, when we tested a recombinant malaria vaccine candidate, LSA-NRC, to determine its immunodominant epitope(s), LSA(434–453) located at the C-terminus of the protein emerged as the dominant epitope.^34,58 Interestingly, this epitope contained part of a spacer sequence artificially added to the protein for purification purposes (residues 444–453).⁵⁹ While the LSA sequence ends at Leucine 443, the immunodominant epitope extended aa 443 to include the spacer sequence GGSGSP and four histidine residues. Notably, the artificial epitope was the selected determinant in vivo, as verified by its ability to induce T cell responses in DR1 transgenic mice immunized with LSA-NRC, and in human volunteers immunized with LSA-NRC vaccine preparations.^34,58 Moreover, the peptide containing the spacer sequence, LSA(436–449) was the only peptide to recall strong dose-dependent T cell responses. Consistent with this observations, LSA-NRC protein, which was originally designed as a vaccine candidate in humans, did not protect vaccinated individuals against malaria challenge,⁶⁰ likely because immunized individuals developed immunity to the spacer containing epitope and not to the physiological epitope of the protein.^34,58 Similarly, the immunodominant determinant of a recombinant Peptidylarginine Deiminase 4 (PAD4) protein that contained a spacer shifted the immunodominant epitope to the spacer sequence.⁵⁸ Notably, the spacer added to this PAD4 protein was different in sequence and placement from the spacer added to LSA-NRC. In short, three recombinant proteins all having different spacer sequences preferred part of the spacers as their immunodominant determinants.

I. An Immune Hierarchy is Established When Two Different Epitopes Compete for the Same HLA-DR1 Molecules

The above observations indicated that immune hierarchy exists between epitopes of one antigen. Next, we found that the immunodominant epitope of one antigen might become subdominant or cryptic when other more robust epitopes are simultaneously present. It appears that when proteins are being simultaneously processed, a race for binding to the groove of MHC II is ongoing. A combination of factors contributes to the emergence of one epitope becoming dominant. We found that the location of an epitope at the C- or N-termini of proteins might increase the likelihood of its capture by the MHC II, as well as to proteolysis by the processing enzymes.^35,58 The concept of epitope accessibility also sets the rules for the emergence of epitopes that include the spacer sequences as dominant epitopes. Accessibility of an epitope combined with the right amino acid sequences to form DM resistant complexes with MHC class II, can establish dominance over the other epitopes by being presented at higher quantities to T cells.³⁵ The above findings highlight the underlying principles for immunodominance for MHC class II by providing direct evidence for the contribution of factors intrinsic to antigenic structure or epitope accessibility. Notwithstanding, these observations suggest that determinant hierarchy is established in the antigen processing compartments rather than at T cell level. Moreover, the above findings go against the accepted dogma that once an epitope gains the title immunodominant, it is always immunodominant because the slightest change in the Ag sequence/structure can disturb the determinant selection, and simultaneous processing of multiple antigens creates another level of hierarchy, as was predicted by Eli Sercarz.

Thus, these findings tell a cautionary tale to vaccine designers,^36,61 warning against the assumption that placement of dominant determinants in any context would not affect their selection as dominant epitopes, or mixing several vaccines together might result in effective immunity against all the immunogens in the mix.

IV. T CELL RECOGNITION OF PMHC II

Alongside our work on teasing out mechanisms of antigen processing and immunodominant epitope selection, we have been working on T cell recognition of the presented epitopes on the membranes of APCs. The question was whether peptides that bind to MHC II weakly would send qualitatively different signals to T cells that might lead to anergy, as proposed for altered peptide ligands (APLs).⁶² At the time multiple theories were considered as the guiding principles in driving T cell anergy by APLs. The qualitative model for T cell activation or anergy was among the first, proposing that APL caused alterations in the topology of APL-MHC-TCR interface creating a qualitatively different ligand for TCR recognition.^63–67 An alternative model, kinetic proofreading, emerged using kinetic measurement of interactions between soluble TCR with and soluble MHC proteins in complex with APL.^68–71 The model proposed that APL/MHC complexes dissociated faster than the agonist ligands/MHC from their cognate TCR. Because of extreme low affinities integral to the ternary complexes of TCR/pMHC (K_d values of ~ 10⁻⁴–10⁻⁶M, and lifetimes of ~ 1–100 seconds),⁷² evidence for correlations between the dissociation rate of TCR/peptide/MHC ternary complexes and the different effector functions required highly sensitive technologies using surface plasmon resonance. It was demonstrated that faster dissociation rate for the ternary complex led to the expression of fewer T cell activation markers and/or lower T cell function.^73–75 However, a challenge to the kinetic proofreading theory was that it did not offer a certain threshold common to all agonists, partial agonists, or antagonists with similar effects, and experimental data did not fully support a unifying concept.^72,76–79 Moreover, because of the extremely fast dissociation rates for antagonist pMHC/TCR, and lack of biochemical data hinting at some measurable cellular effects by the antagonist peptides, it was hard to place them together with other partial agonists with measurable lifetime and/or biochemical effect at the cellular levels.

Another challenge to the kinetic proofreading theory comes from our work with peptide ligands that dissociated rapidly from the MHC but did induce T cell anergy, similar to the APL. Work by Mirshahidi et al.⁸⁰ postulated that the determining factor for the induction of anergy or activation was the lifetime of the TCR/pMHC ternary complexes. Thus, peptides that dissociate rapidly from the MHC should have similar effects as the APL and should induce anergy. Notably, however, one of the short-lived peptides, HA(Y308A), triggered both activation and anergy but at 10- to 100-fold concentration range. Interestingly, when similar peptide titrations were applied to the agonist HA(306–318) peptide, the same trend was observed; higher doses of the agonist peptide induced activation and lower doses induced anergy in a T cell clone specific for HA(306–318)/DR1.^80–83

A. Agonist Peptides at Low Densities Induce Anergy

The finding that in the presence of signal 2, agonist ligands induced T cell anergy when displayed at low densities on APCs strongly suggested that neither the lifetime of TCR/pMHC, nor the quality or topology of the pMHC presented to T cells are the best determinants of T cell activation or anergy. The observation highlighted a quantitative model for T cell activation. It showed that an overall avidity of peptide-MHC-TCR ternary complexes rather than the affinity of the individual molecular complex is the critical factor for T cell responsiveness. Other laboratories also have made the observation that presentation of low densities of agonist peptide-MHC complexes induced T cell anergy.^84–87

B. The Magnitude of TCR Engagement is a Critical Predictor of T Cell Anergy or Activation

Ag recognition by TCR engagement as detected by internalization and degradation has been established as a correlate of T cell activation.^88,89 Korb et al. demonstrated that at low doses of antigen, fewer than 10 complexes of peptide-MHC were presented per APC and engaged ~ 1000 TCR in T cell clones, which caused T cell anergy as defined by lack of T cell proliferation and IL-2 production.⁸³ In an attempt to compare the number of TCR downregulated upon interaction with the inhibitory doses of agonists, short-lived peptides, or APLs, it became clear that regardless of differences in structural properties, all inhibitory stimulations resulted in downregulation of ~ 1000 TCR molecules, whereas stimulatory doses of agonists or short-lived ligands downregulated > 4000 TCR.⁸⁰

C. Low Density TCR Ligand and APL and Intracellular Signaling

A comparison of phosphorylation patterns of TCR signaling components initiated by low doses of agonist peptide or inhibitory doses of short-lived peptides indicated similar degrees and patterns of phosphorylation as the APL. T cells pretreated with low doses of agonist peptides or inhibitory doses of short-lived peptides exhibited partial phosphorylation of CD3z and below detection levels of pZAP-70, pLAT, and pSLP-76. However, Fyn was fully phosphorylated consistent with the phosphorylation pathway seen in anergy induced by traditional APL.^90–92 Anergy induced by all three forms of ligands, APLs, low doses of agonist peptides, or peptides that dissociate fast was associated by the lack of IL-2 synthesis. Another similarity between the three forms of ligands was the transient phosphorylation of Vav, a known regulator of rearrangement of the actin cytoskeleton and capping of TCR.^92–96 Low densities of agonist peptides or short-lived peptides induced a transient phosphorylation of Vav in T cells treated with inhibitory doses of peptides, and were accompanied with no detectable actin polymerization. Altogether, a clear trend was emerging that short-lived peptide-MHC complexes and low densities of long-lived agonist peptides both induced T cell anergy through engagement of fewer TCRs. Thus, a unifying model explaining how APLs, agonist ligands, and short-lived peptides induce anergy could be the engagement of less than optimal number of T cell receptors by a variety of ligands.

D. Anergy Induced by Low-Density p-MHC was Physiological and Occurred In Vivo

To study T cell anergy induced by low density agonist presentation, Mirshahidi et al. used several transgenic mouse models, including an HLA-DR1 transgenic strain,⁹⁷ naturally populated with heterogeneous CD4 T cells.⁸² Mirshahidi et al. provided evidence that low-avidity engagement of T cells by low densities of agonist pMHC led to the induction of anergy in antigen experienced CD4 T cells past their peak of responsiveness or memory, rather than naïve CD4 T cells in vivo.^80,82,98 In vivo results using short-lived peptides confirmed 10- to 100-fold higher doses of short-lived peptides were necessary to induce anergy in memory T cells in HLA-DR1 transgenic mice.⁸⁰ Needless to say, use of short-lived peptides for induction of anergy in vivo has many advantages over the traditional APL because unlike APL that is strictly specific for a single clonal TCR, short-lived peptides can interact with several T cell clones specific for a given peptide-MHC complex in vivo and tolerize HLA-DR1 transgenic mice. Because APL tolerize single T cell clones their clinical relevance is limited when used for the treatment of pathological self-reactivity in vivo. Indeed, clinical trials with APL have raised important considerations for their use for immunotherapy.^99,100

E. Which Antigen Presenting Cells Induce Anergy in Memory T Cell?

Going back to my original hypothesis, we next addressed the specific interactions of different types of APCs with memory CD4 T cells. A novel ex vivo anergy assay first suggested that B cells induce anergy in memory T cells and an in vivo cell transfer assay further confirmed those observations.⁹⁸ We demonstrated that B cells pulsed with defined doses of Ag anergized memory CD4 cells in vivo. We established that CD11c⁺ DCs did not contribute to anergy induction to CD4 memory T cells, because diphtheria toxin receptor (DTR) transgenic mice that were conditionally depleted of DCs optimally induced anergy in memory CD4 T cells. In agreement with the above findings, B cell deficient μMT mice did not induce anergy in memory T cells. Moreover, we showed that B2 follicular B cells were the specific subpopulation of B cells that rendered memory T cells anergic. Notably, we showed that anergy in this system was mediated by CTLA-4 upregulation on T cells, in agreement with our earlier findings in T cell clones.⁹⁸

In the next set of experiments we tested the potency of B cells that had been immunized by either Ova protein in CpG, or Vaccinia-OVA in vivo to induce anergy in OVA-specific (DO11.10) memory CD4 T cells. Normal BALB/c mice were injected with an immunogenic dose of OVA mixed with CpG, and their B cells were isolated at 4-day intervals and transferred to groups of recipient BALB/c mice harboring DO11.10 memory T cells. No additional immunization was performed. Therefore, the only source of antigen was the pMHC displayed on B cell membranes at quantities proportional with the duration of time post immunization; earlier time points displayed more pMHC, and later time points fewer pMHC. We found that B cells from mice harvested on day 16 or day 20 postinjected with OVA/CpG rendered memory CD4 T cells hyporesponsive. B cells harvested before day 16 (days 4, 8, and 12), and after day 20 (days 24 and 28) were stimulatory and induced positive responses. These experiments suggested that the level of antigen presented by B cells during 16–20 days post-CpG/OVA injection might be just the right amount for rendering CD4 memory T cells unresponsive.

To further test the role of BCR-mediated antigen uptake in this setting, groups of HEL-Ig BCR transgenic IgHelMD4 mice were injected with OVA-HEL in CpG. At different time intervals, B cells from those mice were isolated and injected into B6 mice harboring memory OT-II T cells. Eight to ten days later, cells from those mice were tested for responsiveness in vitro. We found that B cells harvested from IgHelMD4 mice on days 41, 44, and 48 after injecting HEL-OVA caused reduced proliferation and IL-2 production in DO11.10 memory CD4 T cells. Compared with B cells transferred on days 16 and 20 after immunization, B cells with HEL specificity were more efficient in antigen capture and presentation. The response levels on day 4 posttransfer were particularly low, likely because of antigen induced cell death since it is expected that at day 4 postimmunization B cells would carry large amounts of antigen. The above complex experiments revealed that different levels of processed OVA peptide/MHCII displayed on B cells render CD4 memory T cells hyporesponsive/anergic, and that this process is expectedly more efficient when HEL-specific B cells IgHelMD4 mice¹⁰¹ are utilized.

Using CpG or Vaccinia viral infection as models to study the immune response, we observed that purified B cells from the infected or immunized mice could induce hyporesponsive memory T cells. Those experiments showed that antigen capture was indeed 1,000–10,000-fold more efficient when OVA antigen was targeted to HEL-specific BCR transgenic B cells, providing a direct evidence for the role of BCR in capturing of antigen at very levels (10⁻⁵–10⁻⁸ pmol), making it highly unlikely for any other APC to match this level of sensitivity.

Since transferring B cells from infected mice generated hyporesponsive memory CD4 T cells in the recipient mice, it was possible that this state of rest could develop spontaneously in memory CD4 T cells in mice that had recovered from infection. Indeed, we observed that memory T cells began to become hyporesponsive after the contraction phase in the absence of any external interference. When either Vaccinia-OVA, or OVA/CpG were used as mimics of infection, OVA specific CD4 T cells contracted in numbers and became unresponsive to antigen once the effector phase ended. Several studies have already demonstrated that CD4 T cells increase in number during the effector phase and decline over time after gaining memory characteristics following the resolution of infections, consistent with our observation.^102–104 Our findings are also in agreement with the reports that memory T cells adopt a resting state¹⁰³ because of programmed metabolic switches that control glycolysis and/or fatty acid oxidation.^105–107 While those studies indicated that memory T cells are in resting state, we have demonstrated here that CD4 memory T cells undergo a resting state initiated by B cells and triggered by certain low levels of antigen during the resolution of infection. It is noteworthy that low level of antigen presented by B cells is concurrent with low-level expression of danger signals as well. This resting state is a transient condition and may be reversed by antigen and IL-2, a condition that is met during the re-emergence of an infection due to inflammatory conditions. We demonstrate that a second viral infection even after nearly 14 months postinfection is stimulatory to the memory populations that are otherwise fully unresponsive to a peptide-alone challenge in vitro. Similarly, IL-2 plus peptide, or CpG plus antigen administered in vivo recalled vigorous responsiveness in quiescent memory CD4 T cells.

In Vaccinia-OVA injected mice, over half of CD4 DO11.10 cells showed high expression of memory markers, such as CD44, CD127, and CD62L, from which more than half did not undergo homeostatic proliferation. In contrast, only about 10% of CD4 DO11.10 T cells from OVA emulsified in CFA injected mice showed high expression of CD44, CD62L, and CD127. Our observations suggested that continuous presence of Ag as expected in OVA emulsified in CFA mimics a chronic infection, and disrupts induction of high quality memory T cells that are quiescent. Consistent with this notion, B cells that expressed high levels of signal 2, B7-1 did not induce anergy.⁹⁸ On the contrary, Vaccinia-OVA infection clears in few days, hence clearance of danger signals, hence B cell presentation for induction of anergy/dormancy. All these findings document that perhaps dormant memory T cells survive longer and might be less harmful to self-tissues due to cross-reactivity.¹⁰⁸

In summary, our findings highlight an important physiological process that takes place at the end of an infection, when (1) the antigenic load is reduced and danger signals are no longer present, and (2) antigen experienced CD4 T cells need to receive signals that infection has waned. Under such conditions, B cells bearing specific receptors for antigens are the natural choice of the immune system for sending the message to T cells. By capturing antigen at its lowest level and presenting it to the T cells, they promote a resting state. When the need arises to fight infection during a challenge, the quiescent memory T cells get activated and exert their effector function.

Therefore, our studies put forward a novel regulatory mechanism for CD4 memory T cells. Despite the general view that memory T cells are readily activated, our data reveal strict regulation on memory CD4 T cells. First, lack of cell proliferation and reversibility by inflammatory cytokines fulfill the original definition of T cell clonal anergy.¹⁰⁹ Anergized memory T cells maintain low metabolic activity and cell cycle progression, criteria that would preserve cellular energy and might be the key mechanism in long-term survival. However, the critical requirement of anergized memory T cells for inflammatory cytokines for reactivation is another control mechanism of responding to danger.¹¹⁰ Thus, memory T cells, while equipped to respond to antigens rapidly, also require second signals for the initiation of response, similar to naïve T cells. The need for an inflammation induced danger signal for the activation of memory T cells prevents them from autoreactivity.

In recent work currently under preparation, we have examined the dynamics of gene expression in CD4 T cells from naïve, to activation, resolution of infection, and differentiation into long-lived memory.^{106,109–111} Alongside, we discovered a new set of markers for CD4 memory T cells that appear late during memory differentiation. We even extended our new memory markers in humans and found that they represented the most robust responding CD4 T cells to influenza virus in individuals vaccinated against influenza. Thus, through continuous and reproducible research exploring development and persistence of memory CD4 T cells, we define a CD4 memory population as exemplary for its high abundance, lack of response to peptide stimulation in the absence of danger signals, for its excellent polyfunctional recall responses upon secondary challenge, and importantly, its longevity.

V. EPILOGUE

Here, I have touched upon my research career from the start until now. A clear message that hopefully has come through is that a question/hypothesis that I found so appealing during my PhD training could draw a path that has lasted for decades. While at times this journey might have appeared as disjointed, one path being gaining a mechanistic understanding of antigen processing and immunodominant epitope selection, and the other being investigating quantitative aspects of T cell recognition of pMHC, both paths have merged together because B cells, rather than DCs, take control of signaling CD4 T cells to become quiescent and long-lived, and they do it through a quantitative mechanism; also, the MHC II chaperone DO functions by increasing the abundance of immunodominant determinants on B cells. Multiple reports indicate that CD4 T cells in the GC Light Zone (LZ) are Ag-experienced,^112,113 and that B cells in the GC present pMHC to those T cells. Based on our model in which DO enhances the abundance of the immunodominant pMHC,^47,55 as B cells presenting the highest levels of pMHC preferentially populate the GC,¹¹² high densities of the immunodominant epitopes presented on B cells could warrant successful entry into the GC. Once B cells are in the GC, DO level decreases, and so does the abundance of pMHC on B cells.^114–116 The presentation of lower numbers of pMHC would preferentially help with selection of high-affinity Ag-specific CD4 T cells in the LZ for differentiation to become memory precursors, in keeping with the model suggested from our prior findings.

I am thankful….

I could not have been happier for the privilege of having Eli Sercarz as my PhD mentor. Because of his love for science and critical mind, what I learned in his lab made a lifelong impression. I also wish to thank all my students and postdocs who by their hard work, intellectual contributions, and diligence have made all the work described above possible. Importantly, my sincere appreciations go to all my other mentors and colleagues who have contributed with their incredible insights and encouragements for which I remain in their debt forever.