LAG-3 – a very discerning immune checkpoint

For much of its existence the co-receptor CD4 has been shadowed by a related protein called LAG-3. Now the focus of considerable interest as an ‘immune checkpoint’, new work by Maruhashi et al.¹ suggests the surprising finding that LAG-3 inhibits T-cell responses, but only those driven by especially stable MHC class II/peptide complexes.

An interesting effect of the immune checkpoint phenomenon is the way it can elevate a protein. This is true even for PD-1, which achieved a degree of celebrity when Rafi Ahmed and colleagues showed that it marked “exhausted” T cells arising in the course of chronic infections², but whose status was transformed when anti-PD-1 antibodies produced objective clinical responses in melanoma patients right from the get-go³. A newer example is LAG-3 which, along with TIM-3 and TIGIT, forms part of a second wave of checkpoint targets because it is expressed alongside PD-1 on tumour infiltrating lymphocytes and has been associated with T-cell exhaustion⁴.

LAG-3 was discovered almost 30 years ago by Triebel and colleagues⁵, firstly in the form of a transcript expressed by an IL-2-dependent natural killer cell line. This encoded a protein, like CD4, with four immunoglobulin superfamily domains, and whilst LAG-3 shared only very limited similarity with the co-receptor (20%), unusual shared sequences next to the cysteines in the first and third domains confirmed, along with their chromosomal locations, that they also share an ancient ancestor. Like CD4, LAG-3 had a short cytoplasmic domain, but unlike it, LAG-3 was expressed not only on activated CD4⁺ T cells but also on activated CD8s, NK cells, B cells and dendritic cells. Also in contrast to CD4, LAG-3 blockade produced more rather than less T-cell signaling. The CD4 homology immediately suggested that it would bind MHC class II (MHC II), which was quickly confirmed by Triebel et al. using cell binding assays⁶. The reasonable expectation would then have been that LAG-3 binding would correlate strictly with MHC II expression, but this is not at all what Maruhashi and colleagues now find.

They set out to re-evaluate the binding specificity of LAG-3, using a pentameric form of the receptor as a staining reagent in flow cytometric assays. To their surprise they found that although LAG-3 binding correlated to some extent with MHC II expression, there were striking exceptions. Substantial numbers of conventional dendritic cells, most mature bone marrow-derived B cells, and nearly all plasmacytoid dendritic cells bound little or no LAG-3, despite expressing substantial amounts of MHC II. Maruhashi et al. then went onto show that cell-lines presenting exogenous, classical MHC II-binding peptides invoked comparable T-cell responses regardless of whether they bound the LAG-3 pentamer but only responses to those cell lines that bound the pentamer could be inhibited by T cell-expressed LAG-3. This suggested either that there was a non-MHC II ligand for LAG-3, or that LAG-3 binding to MHC II relies on some co-factor.

Maruhashi et al. turned to expression-cloning to search for the alternative ligand for LAG-3 or the binding-permissive co-factor. Tellingly, what they isolated was the Ciita gene. Ciita encodes an IFNγ-inducible transcriptional co-activator of not just MHC II but also, among others, proteins involved in the assembly and peptide editing of MHC II, i.e. Ii and DM. Whilst this did not completely exclude the possibility that there is a non-MHC II ligand for LAG-3 still to be discovered (CRISPR-based searches⁷ are best for this), it was strong circumstantial evidence for MHC II being the dominant player in LAG-3 interactions. And what’s more, Maruhashi et al. were able to explain all their observations in terms of Ciita’s role in creating stable MHC II/peptide complexes. They showed that ectopic Ciita expression dramatically increased the binding activity of LAG-3 non-binding antigen presenting cells, which in turn restored their ability to suppress the activation of LAG-3 expressing T-cells (Fig. 1a). Conversely, deleting Ciita or the Ii and DM genes completely eliminated LAG-3 binding, while only partially reducing MHC II expression. Finally, Maruhashi et al. established a direct link between LAG-3 binding and the stability of MHC II/peptide complexes by showing that MHC II expressed with covalently attached peptides, i.e. super-stable pMHC II, bound strongly to LAG-3 in a Ciita-independent manner, whereas disruption of the anchor residues of these peptides, or complexes formed with other low-affinity peptides, did not. In other experiments Maruhashi et al. showed that LAG-3 did not block TCR/MHC II or CD4/MHC II binding, and that LAG-3 relies on its cytoplasmic domain to suppress T-cell activation, despite it not having conventional signaling motifs, such as ITIMs.

Figure. LAG-3 binding does not always correlate with MHC II expression.

(a, left panel) LAG-3 fails to bind MHC II expressed by most plasmacytoid DCs but also mature bone-marrow-derived B cells, despite high levels of MHC II expression by these cells. Maruhashi et al. show that this is because LAG-3 does not bind to unstable MHC II/peptide complexes (left panel). (Right panel) IFNγ triggers expression of the transactivator CIITA, which in turn induces DM expression and peptide editing, the formation of stable MHC II/peptide complexes, and LAG-3 binding, whereupon LAG-3 exerts its inhibitory effects. (b) Ribbon diagrams of HLA-DR (PDB: 1DLH) with the CD4 binding-site¹² colored blue and the region influenced by MHC II/peptide complex stability (residues with RMSD >2Å versus the DM-bound DR1 complex⁹) colored yellow. The structure is shown from the point of view of the T-cell receptor (top panel) and from the side (i.e. rotated 90°; bottom panel).

In this way Maruhashi et al. have shown convincingly that LAG-3 preferentially suppresses the activation of T cells and, presumably other LAG-3 expressing cells, by binding MHC II proteins presenting only stably-bound peptides. The questions now are: how does LAG-3 distinguish between stably- and unstably-bound peptides, and why? Peptide-dependent conformational variation in MHC II proteins was first detected as changes in their migration during SDS-polyacrylamide gel electrophoresis⁸. Comparisons of stable HLA-DR/peptide complexes, with DM-stabilised HLA-DR complexed with a low-affinity peptide⁹, have revealed that stable peptide binding triggers conformational changes at one end of the peptide-binding helix of DRα and the floor of the peptide-binding groove next to it. By binding this part of MHC II, LAG-3 might be able to ‘read’ complex stability. This region is opposite the CD4 binding site (Fig. 1b), perhaps accounting for why CD4 and LAG-3 can bind simultaneously to MHC II. It is nevertheless surprising that they can do this given their shared history. This is because it implies either that one of the orthologues lost MHC II binding to the shared site and regained it at the second site, or that it somehow worked its way around to the other side of the molecule, two scenarios that seem equally unappealing. The structure of a LAG-3/MHC II complex, when it comes, will be very interesting.

Understanding why LAG-3 is ligand-selective is also something for the future. The IFNγ- and CIITA-dependent generation of stable MHC II complexes will ensure that LAG-3 will only ever get to exert its inhibitory effects in the context of on-going immune responses. But this requirement is already fulfilled by LAG-3 expression being activation dependent, like other immune checkpoints. It can also be argued from a structural viewpoint that it isn’t so surprising that LAG-3 is restricted to binding the final, stably-folded form of MHC II. Most protein interactions rely on interfaces formed by large arrays of atoms positioned stably in three-dimensional space. Simply by producing folding intermediates capable of reaching the cell surface, MHC II’s singular maturation pathway may allow LAG-3 to exhibit apparent selectivity. But none of this is to say that the fact of selective MHC II ligation by LAG-3 will not have important implications for immune function. Maruhashi et al. propose that autoimmunity could be provoked by T cells that escape negative selection mediated by DM-edited peptides in the thymus and can’t be suppressed by LAG-3 in the periphery because they react with unstable MHC II/peptide complexes. Supporting this idea, the authors note that CD4⁺ T-cells capable of responding to unstable peptides are known to be highly diabetogenic in non-obese diabetic mice¹⁰.

Does any of this alter LAG-3’s status as a target for immunotherapy? Probably not: the twenty-four on-going or completed trials of LAG-3-based biologicals¹¹ will settle that matter. Partly due to the success of anti-PD-1 and -CTLA-4 antibodies, but mostly because there are now so many trials (>1100 in the case of PD-1), many other immunotherapy targets and combinations are unlikely ever be triaged in this way. Careful studies of immune checkpoint biology, in the manner of Maruhashi et al., will need to underpin candidate selection, with the bonus that there will be yet more surprises.