My interest in immunology began while I was a graduate student at the Massachusetts Institute of Technology (MIT) working on DNA replication in bacteria (1). At the time there were few immunologists and no immunology graduate students at MIT. Norman Klinman came from the University of Pennsylvania to give a lecture. It made such a strong impression on me that I still remember the title, “The Biography of the B Cell.” I had no previous exposure to immunology and I was fascinated by the subject matter. I decided to do my postdoctoral research in immunology.
That led me to the Albert Einstein College of Medicine where I joined the laboratory of Stanley Nathenson, who was working on the structure of MHC molecules, and Matthew Scharff, who was doing somatic cell hybridization with myeloma cells. My project was to select for cells expressing mutant MHC molecules in vitro based on escape from recognition by anti-MHC Abs. I also tried to make hybridomas against MHC molecules so I could use mAbs to select MHC mutants. I didn’t know how to immunize mice very well, so even though I got lots of hybridomas (to my knowledge, some of the first polyethylene glycol–produced hybridomas), they didn't have the specificity I wanted. That was my first foray into immunology and, as you’ll hear in a minute, although it wasn't very productive at the time, what I learned at Einstein about MHC mutants greatly impacted my research in the future.
My second postdoctoral position took me back to Boston, to the Pathology Department at Harvard Medical School, led by Baruj Benacerraf. I chose to work with Steve Burakoff, who was studying the specificity of alloreactive. We collaborated with Matthew Mescher to identify subcellular material that was able to stimulate alloreactive CTL as well as CTL specific for hapten-modified syngeneic cells. This led to several manuscripts demonstrating that MHC molecules are necessary and sufficient for T cell recognition of these Ags (2, 3).
At the time, the mid- to late seventies, the nature of the TCR was a mystery and the subject of much debate. Why was there such a high frequency of alloreactive T cells? Zinkernagel and Doherty, Bevan, and others found that Ag recognition was MHC restricted, and that added to the confusion. Were there two receptors, one for MHC, and one for Ag, or one that saw both? If there were two receptors, was one similar to Ab? I started looking for the T receptor using Abs specific for Ab idiotypes, as many people believed the TCR contained an Ab-like component. I didn't find anything on T cells using such Abs.
I then moved to La Jolla to start my own laboratory at The Scripps Research Institute. I took with me what I had learned about H-2Kb mutants from Stan Nathenson's laboratory and about CTL from Steve Burakoff. Stan was sequencing H-2Kb molecules from strains of mice harboring H-2 mutations as identified by skin graft rejection by Roger Melvold and Henry Kohn. The sequences revealed that changes of just one or two amino acid residues in the Kb molecule were sufficient to stimulate a very vigorous alloresponse (4). I realized I could use these mutants to study the fine specificity of receptors on clones of CTL specific for Kb. So I set up an assay where I cloned H-2Kb–specific T cells by limiting dilution in ninety-six–well plates, and expanded the clones using T cell growth factor, which we didn’t know at the time contained IL-2. After about a week, there were sufficient numbers of cells in each clone so that we could interrogate their specificity against a panel of nine different target cells, seven of which expressed different mutations in Kb. With seven different Kb molecules I could, in theory, discriminate 128 different TCR specificities. Over 50 different types of receptor specificities were identified in these experiments using Kb- specific alloreactive CTL obtained from H-2d (B10.D2) mice (5).
In order to determine how the MHC of the responder affected the composition of the receptor repertoire, we performed the same experiment looking at the repertoire of Kb-specific CTL from B10.BR mice (H-2k), and F1 animals (B10.D2 x B10.BR)f1 (6). We found that several specificities that recurred frequently in the repertoire of B10.D2 animals were no longer evident in the Kb-specific repertoire of B10.BR mice. However, there were other specificities that appeared, suggesting that there was a very strong influence by the MHC on the choice of receptors that were used to recognize the Kb molecule. Furthermore, some parental specificities were eliminated in F1 mice whereas others were retained. These results provided evidence for both tolerance, which eliminated some recognition patterns observed in the F1, and for positive selection. The question of how the repertoire is shaped by tolerance and positive selection remains an area of active investigation.
So what was the basis for this great diversity in specificities against a single alloantigen? A big clue came several years later when Björkman and Wiley published the x-ray crystallographic structure of the HLA-A2.1 molecule. They found what appeared to be an Ag binding groove formed by two α-helices sitting above a platform composed of eight antiparallel β strands (7). The groove was lined with the polymorphic residues of the molecule (8). Most exciting was the discovery that this groove contained electron-dense material, consistent with the presence of a diverse array of peptides. This discovery suggested that the high frequency of alloreactive cells could reflect the vast diversity of processed self-peptides that were presented within the peptide-binding groove of the MHC molecule.
We thought one way to prove this hypothesis would be to find a target cell line that expressed Kb but not the peptide recognized by a particular Kb-specific CTL clone, and therefore would resist lysis. To increase the probability that the target would not present cognate peptide, we used as the target a human cell line (Jurkat) that was transfected to express Kb. A postdoctoral fellow in the laboratory, William (Bill) Heath, tested many different CTL clones specific for Kb to see if we could find some that were unable to recognize Kb on the transfected Jurkat Kb cells. Most CTL clones recognized both mouse cells expressing Kb (EL-4) and the Jurkat Kb cells; however, he found several that did not. Could we reconstitute recognition of the Jurkat Kb cells if we provided them with peptides produced from the mouse cells? Bill treated cell lysates containing cytoplasmic proteins from EL-4 cells with cyanogen bromide to produce peptides. Amazingly, following incubation of the Jurkat Kb targets with these peptides, they became susceptible to lysis (9). What a eureka moment! We had evidence that peptides were required for recognition of an alloantigen by CTL.
A major advance at that time was the identification of cell lines that were defective in their Ag processing machinery and displayed empty MHC class I molecules on their cell surface. These cells could be incubated with the peptide of choice and, in the presence of β2-microglobulin, the MHC molecules would form around the peptide (10, 11). To obtain the allopeptides we would acid-wash cells expressing Kb to denature the class I molecules, which released most of the peptides into the media. The peptides were subjected to HPLC fractionation and each fraction was used to pulse cells expressing empty MHC molecules. This allowed us to distinguish the receptor specificity of different CTL clones based on their recognition of peptides present in different HPLC fractions (12). We concluded that a diverse array of endogenous peptides was available for binding cell MHC molecules, thereby explaining the high frequency of alloreactive T cells.
Of interest, we observed that as a consequence of peptide-dependent allorecognition, some Kb-specific clones exhibited cell-type–specific recognition among different murine targets (13). Others exhibited tumor specificity such that some CTL clones raised against tumor cells did not kill normal cells. This result suggested that some of the alloreactive CTL we had produced by stimulation of T cells with EL-4 tumor cells were specific for Ag expressed uniquely by those tumor cells. This specificity was of great interest because there are no tolerance barriers to prevent recognition of peptides on alloantigens, whereas many tumor-expressed Ags may cause tolerance in a syngeneic host. One such tumor-specific Ag turned out to be a peptide from an endogenous retrovirus expressing the tumor line (14).
Another project we were working on involved xenorecognition, recognition of MHC molecules of a different species. We were fascinated by the idea that, despite the fact that alloreactivity was very vigorous, mouse T cells were unable to respond to HLA molecules expressed by human cells. To explore the mechanistic basis for such lack of xenorecognition, we produced a transgenic mouse expressing HLA-A2.1. Indeed, cells from the transgenic mouse were still unable to efficiently stimulate an alloresponse, suggesting an inability to recognize the A2.1 molecule even when it was expressed on mouse cells. How could this be explained? One hypothesis was that the TCR repertoire of the mouse had not evolved to recognize MHC of other species. However, another possibility was that species differences did not reside within the highly variable TCR repertoire, but rather in the inability of the mouse CD8 coreceptor to interact with HLA. CD8 recognition of class I MHC was known to be critical for CD8 T cell activation; however, the site of CD8–MHC interaction was unknown. Mike Irwin and Bill Heath decided to test the hypothesis that mouse CD8 was unable to recognize human MHC due to species-specific differences in the invariant α3 domain of class 1.
With much effort, Mike was able to obtain mouse CTL that were specific for the A2.1 molecule expressed by human cells. He then compared the ability of these CTL to recognize human cells that expressed either A2.1, or a chimeric form of the molecule comprised of the polymorphic α1, α2 domains of A2.1, spliced onto the α3 transmembrane and intracellular domains of mouse Kb. We found greatly increased recognition of the targets expressing the chimeric A2/Kb molecule as compared with those expressing intact A2.1 (15). This was consistent with our hypothesis that the barrier to xenorecognition was not the TCR, which interacted with the polymorphic α1/α2 domains, but rather the inability of the mouse CD8 to recognize the α3 domain of HLA. Bill Heath obtained target cells expressing MHC molecules containing mutations in the α3 domain of the mouse class I and performed experiments that further indicated CD8 interacted with a site in the α3 domain of MHC (16, 17). David Margulies’ laboratory eventually proved this at the molecular level when they published the structural interaction of MHC class 1 and CD8αβ (16).
Considering that the A2/Kb molecule was much more efficient than fully human A2.1 at interacting with mouse CD8, we thought that if we produced a transgenic mouse expressing the chimeric molecule, we would have a mouse that could be used to identify A2-restricted peptide epitopes. This project was taken on by Antonella Vitiello in the laboratory (18). She was very interested in using such mice to identify viral and tumor epitopes, but she first needed to determine whether mouse cells actually processed and presented the same peptide epitopes presented by human cells. It was known that following infection with influenza virus, the dominant CTL epitope recognized by human CTL in association with HLA A2.1 was a peptide spanning residues 58–66 of the viral matrix protein. She immunized the A2/Kb mice with influenza, cloned their influenza-specific CTL, and assessed their specificity. Essentially all of the CTL restricted by A2/Kb were found to be specific for matrix peptide 58–66 that was the dominant epitope in humans (18). This specificity validated the model, and numerous laboratories have since used mice expressing A2/Kb and similar chimeric molecules to identify tumor and viral epitopes.
These studies led to another idea. CD8 facilitates T cell activation by bringing Lck to the signaling domains of the TCR complex. We had shown that in the absence of successful interaction between CD8 and MHC, T cells were difficult to obtain as only very high-affinity TCRs could successfully promote T cell activation (19). At the time, Matthias Theobald was trying to identify peptide epitopes of p53 molecule, a molecule overexpressed by many tumors. We thought such peptides might serve as the basis for a tumor vaccine that would work against many different types of tumors. Matthias compared the amount of peptide required for lysis by A2-restricted CTL obtained from transgenic mice expressing either the intact A2 molecule or A2/Kb. The CTL from the A2 mice required 10-fold less peptide to achieve lysis comparable to CTL from A2/Kb mice (20). In fact, the TCRs were of such high affinity that their soluble forms were capable of binding cells expressing cognate peptide-MHC molecules (21). We thought such high-affinity receptors would be exceptional for lysing human A2–expressing tumor cells, but in order for this to be of value, we would first need a method to get the receptors into T cells of tumor-bearing patients. It took 10 years before retroviral vectors expressing TCRs were developed. Richard Morgan was doing this successfully in Steven Rosenberg’s department at the National Cancer Institute, so we gave him the p53-specific TCRs to see if that would work to kill human tumors. Peripheral blood T cells were transduced with the high-affinity p53-specific TCRs and found to be very good at recognition of human tumor cells that expressed both A2 and p53 (22). Unfortunately, p53 was also found to be expressed in activated human lymphoblasts and the transduced CTL killed highly activated T cells, thereby precluding their value as an immunotherapy. But the concept of transducing T cells with tumor-specific TCRs has been used successfully by many groups and may someday become one of the standard weapons in the armamentarium of cancer immunotherapy.
Through these studies on tumor Ags and TCR affinity we became very interested in understanding the relationship between self-tolerance, autoimmunity, and tumor immunity. Autoimmunity is a breakdown in mechanisms of self-tolerance in the immune system, whereas the goal of tumor immunity is to develop an autoimmune response to the tumor. We thought if we could understand the basis for self-tolerance to tissue Ags, we might learn how to manipulate self-tolerance mechanisms to focus the immune system on the destruction of tumor cells. We might also learn how to restore tolerance in autoimmune patients.
In collaboration with David Lo, we developed a transgenic model in which the rat insulin promoter was used to drive expression of the influenza hemagglutinin (HA) in the insulin producing β cells of the pancreatic islets (InsHA mice). Upon immunization with influenza, the avidity of the T cell response in mice expressing the InsHA transgene was found to be 10- to 30-fold lower than that of conventional mice, and this lower avidity was largely due to differences in TCR affinity (23). To learn where tolerance first occurred to the peripherally expressed HA Ag, we also produced a transgenic mouse (Clone 4 TCR) expressing a class 1–restricted TCR specific for an HA peptide restricted by Kd. We found no evidence for thymic deletion of the Clone 4 T cells, but did find evidence for peripheral deletion in the pancreatic lymph nodes where the cells first became activated by cognate Ag (24). Bill Heath was at the Walter and Eliza Hall Institute in Melbourne performing similar experiments using mice expressing OVA in the islets (RIP-OVA mice) to examine the fate of OT-1 CD8 cells specific for OVA. In an elegant set of experiments, he and his colleagues demonstrated that Ag expressed in the islets was picked up by dendritic cells and cross-presented in the pancreatic lymph nodes (25). In our model, upon activation by such cross-presented Ag in the pancreatic lymph nodes, the Clone 4 cells underwent Bim-mediated apoptosis and never left the pancreatic lymph nodes (26). We also learned such tolerance was prevented by the presence of CD4 cells specific for HA, or by the presence of pathogens or inflammatory cytokines that activated the cross-presenting dendritic cells, which explained the difference between T cell tolerance and development of effector function (27).
What causes a breakdown of tolerance in autoimmunity? To examine this issue we used NOD mouse that develop spontaneous type 1 diabetes. Fortunately, NOD mice express H-2Kd, which allowed us to produce NOD mice that expressed the Clone 4 TCR. NOD Clone 4 cells were transferred into NOD-InsHA mice that expressed HA in their pancreatic islets. In contrast to what we found in B10.D2 mice, we saw no evidence of tolerance in the NOD mice. The Clone 4 cells proliferated vigorously in the pancreatic lymph nodes and then migrated to the pancreatic islets (28). So what went wrong? In a series of experiments performed in collaboration with Linda Wicker, we found tolerance deficiencies at two different checkpoints. First, deletion could be restored in the pancreatic lymph nodes if the NOD mice expressed protective genetic regions from C57BL/6 mice found on chromosome 3 (Idd3) and chromosome 1 (Idd5). Second, we identified another opportunity to prevent disease by tolerizing T cells within the pancreas. NOD mice carrying a protective region on chromosome 4 (Idd9) did not delete islet Ag–specific T cells in the pancreatic islets, yet they restored T cell tolerance by successfully eliminating the cells in the pancreas (29). However, we were frustrated in attempts to localize specific genetic loci responsible for tolerance. Many genes were known to be present in the Idd loci under investigation, but due to their close proximity, these could not be separated by genetic recombination. Furthermore, due to the lack of availability of embryonic stem cells of NOD origin, we were unable to genetically modify mice to test the function of individual susceptibility loci. Fortunately, with the advent of CRISPR/CAS9 as a method to introduce specific genetic alterations in the germline, we have begun studies to determine how specific genes may alter disease progression. Our first candidate is a phosphatase encoded by PTPN22. A minor allele of this gene has been found to be highly associated with increased incidence of many different autoimmune diseases, including type 1 diabetes (30). We have introduced the mouse ortholog of the proautoimmune allele of this gene into NOD mice and found that it significantly accelerates both the appearance of autoantibodies to islet Ag and the onset of disease. This validates the use of this model to study the basis for accelerated disease in patients harboring the proautoimmune allele.
Returning to tumor immunology, we wanted to apply the information we learned about normal peripheral tolerance to study T cell recognition of tumor cells. Was the recognition of cross-presented tumor Ags stimulatory or tolerogenic? To address this issue we produced InsHA mice that also expressed the large SV40 T Ag under the insulin promoter. These mice spontaneously develop insulinomas expressing HA Ags. Upon transfer of HA-specific CD8 T cells into tumor-bearing mice, the T cells were found to be tolerized by cross-presented tumor Ag in the pancreatic lymph nodes (31). However, just as we had found in our studies of CD8 T cell tolerance in InsHA mice, such tolerance could be overcome and tumor cells could be eradicated if we also supplied tumor-specific CD4 cells. We found the CD4 helper cells needed to be present within the tumor microenvironment to be effective in promoting tumor killing by the CD8 cells (32). Successful tumor eradication required production by the CD4 cells of both IFN-γ and IL-2. IFN-γ was necessary to induce chemokines required to recruit tumor-specific CTL to the site of tumor, and IL-2 was necessary to promote both the survival and cytolytic activity of the CTL within the tumor milieu (33).
We believe the study of autoimmunity will continue to provide important clues on how to mobilize the immune system to promote tumor eradication.
Another important problem we need to solve right now is how to get Congress to provide more funding for the National Institutes of Health. My son, Matt Klinman, who happens to be a comedy writer, has written and produced a video that proposes how we may use CRISPR to achieve this goal. This video was produced by Matt, and any views or opinions presented are solely those of the producers and do not necessarily represent those of The American Association of Immunologists. It is called “Funding Fever” and may be viewed at https://www.youtube.com/watch?v=f-Y73kscmgc.
Abbreviation used in this article: