A new study by researchers from Cardiff University School of Medicine-UK has found that the SARS-CoV-2 spike mutation P272L emerged to basically evade killer T-Cells generated by infection and vaccination.
The study findings were published in the peer reviewed journal: Cell.
Emerging evidence indicates that HLA-I-restricted CD8 “killer” T cells contribute to the control of SARS-CoV-2 and the immunity to disease offered by the currently approved vaccines (Rydyznski Moderbacher et al., 2020; Sette and Crotty, 2021). Individual HLA-I alleles can be associated with increased likelihood of control or progression of disease with viruses such as HIV (Collins et al., 2020).
The immune pressure of prevalent host CD8 T cell responses against HIV results in selection of altered viral sequences and viral immune escape (Leslie et al., 2005; Price et al., 1997, Price et al., 1998). Influenza virus is also known to gradually escape from CD8 T cells with one study estimating that following the 1968 H3N2 Hong Kong flu pandemic, the virus has lost a prominent CD8 T cell epitope once every three years (Woolthuis et al., 2016).
We were interested in whether there was evidence that SARS-CoV-2 might also escape from CD8 T cells. Genome sequencing has shown extensive non-synonymous mutations occurred within the SARS-CoV-2 genome during 2020. Some of these mutations alter viral fitness (Plante et al., 2021), while others have been shown to diminish the binding of antibodies (Harvey et al., 2021).
To date, there has been limited study of T cell escape and a broad brush-approach of examining T cell recognition of key transmission variants failed to reveal any evidence that escape was occurring (Tarke et al., 2021). A recent report described a patient on rituximab treatment for non-Hodgkin’s lymphoma who was infected with SARS-CoV-2 for over 300 days without detectable neutralizing antibodies (Stanevich, 2022).
During this extended infection period, the virus from this patient formed a single unique clade within the B.1.1 lineage but gained over 30 non-synonymous mutations at a rate that substantially surpassed the evolutionary rate of SARS-CoV-2 in the general population, suggesting prevalent intra-host viral adaptation (Stanevich, 2022). Many of these mutations reduced peptide binding to autologous HLA class I molecules, and two were experimentally validated to escape from the patient’s CD8 T cell response.
Another recent report demonstrated that mutations in the RBD that enhance ACE2 binding and viral infectivity concomitantly reduce T cell recognition through HLA∗24:02 (Motozono et al., 2021). A further study showed that some SARS-CoV-2 mutations in HLA-I-restricted epitopes resulted in lower binding to the HLA, leading to reduced recognition by CD8 T cells, but there was little evidence of these mutations being disseminated (Agerer et al., 2021).
It has been suggested that the L270F (YFQPRTFLL) Spike variant might be an escape mutant (Agerer et al., 2021), but this variant was only seen six times throughout our study period, in six different locations and five different PANGO lineages. From our data, this would suggest that while this mutation does arise, there is no evidence of spread of this variant following sporadic emergence across multiple different SARS-CoV-2 lineages.
A further study recently demonstrated the potential of SARS-CoV-2 mutations to escape from CD8 T cell responses to ORF3a and nucleocapsid (de Silva et al., 2021). Collectively, the above evidence demonstrates that SARS-CoV-2 can alter its sequence to escape from CD8 T cells within a given host, but there has been no proof of transmission of T cell escape variants (Tarke et al., 2021). We set out to look for evidence for dissemination of CD8 T cell escape in the wild.
We reasoned that if escape variants were to disseminate, then they would be most likely to first be seen in a prevalent response through a frequent HLA. HLA A∗02 is frequently carried by all human populations except those of recent African ancestry, where it is present at lower levels. HLA A∗02 is believed to be the most frequent HLA-I in the human population worldwide, occurring in ∼40% of individuals (Gonzalez-Galarza et al., 2020).
HLA A∗02 is hypothesized to have become so frequent in the population due to its success at presenting well-recognized epitopes from historically dangerous pathogens, a supposition consistent with this HLA having entered the population via interbreeding with Neanderthals and Denisovans after early humans migrated from Africa (Abi-Rached et al., 2011).
A previous study used overlapping peptides from the entire SARS-CoV-2 proteome to identify that the most common HLA A∗02-restricted responses in CP were to epitopes contained within peptides spanning residues 3,881–3,900 of ORF1ab and 261–280 of Spike (Ferretti et al., 2020). The Spike epitope was narrowed down to residues 269–277, sequence YLQPRTFLL.
A further study confirmed the prevalence of CD8 T cells that respond to YLQPRTFLL in CP that are absent in healthy donor blood taken before the spread of SARS-CoV-2 (Shomuradova et al., 2020). We confirmed the prevalence of responses to this epitope in our cohort where responses were observed in 12/13 HLA A∗02+ CP tested and all 7/7 SARS-CoV-2 vaccinees, including those who had received only the first dose of a double dose schedule.
We focused our attention on nonsynonymous mutations in the sequence encoding YLQPRTFLL. While at least 73 different amino acid changes have occurred in this region, 33 were seen in more than six sequenced cases in the period from the beginning of the pandemic to January 1st, 2022. The occurrences of mutations in this region that were basal to phylogenetically grouped clusters of 100+ cases all occurred at position 272 with the P272L mutation occurring in over 10,500 sequenced cases as of January 1st, 2022.
Over the course of the pandemic to date, the P272L variant has arisen independently in at least 10 different PANGO lineages during the study period but was most prevalent in the B.1.177 lineage that played a key role in establishing the “second wave” of SARS-CoV-2 in the UK and Europe during the autumn of 2020. P272L was the 4th most common variant seen in the B.1.177 lineage and its descendants and in the top 15 Spike mutations observed globally to January 31st, 2021.
Twelve of thirteen of our HLA A∗02+ CP cohort tested made a response to the YLQPRTFLL epitope, a response comprising >120 different TCRs in total for nine of the patients. Remarkably, no response was seen to the P272L variant. This finding was confirmed using four CP-derived T cell clones that included TCR α or β chains that have previously been described in other CP cohorts by independent research groups (Ferretti et al., 2020; Shomuradova et al., 2020).
Responses to the YLQPRTFLL epitope in a cohort of HLA A∗02+ healthy donors who had been vaccinated against SARS-CoV-2 but had never had symptoms or a positive test since January 2020 ranged from 0.01%–0.2% of CD8 T cells by peptide-HLA staining. However, all of the T cells raised against the YLQPRTFLL epitope across our vaccinee cohort (>50 TCRs) engaged the P272L very poorly and did not stain with P272L peptide-HLA tetramers or react to P272L peptide. The fact that >175 different TCRs from CP and vaccinees raised against the YLQPRTFLL epitope all fail to see the P272L variant suggests that either the proline at position 272 must form a major focus of TCR contact that a leucine residue cannot compensate for, or that substitution with leucine at this position must interfere with a commonly shared TCR binding mode (or both).
Comparison of the atomic structures of HLA A∗02:01-YLQPRTFLL and HLA A∗02:01-YLQLRTFLL showed a largely preserved fold, but with local geometry alterations at the epitope residues 3 to 5 (Spike residues 271–273) which presumably form the majority of contacts with TCRs. In order to study the interaction between a TCR and HLA A∗02:01-YLQPRTFLL, we attempted to make soluble forms of the TCRs from the four T cell clones we generated from CP. We were successful in manufacturing and crystallizing the YLQ36 TCR in complex with HLA A∗02:01-YLQPRTFLL.
The structure of this complex at 3.0 Å showed a canonical mode of TCR binding with a crossing angle of 51°. As the YLQ36 T cell failed to recognize the P272L variant or stain with P272L tetramers, we examined the effect substitution of proline 272 with leucine might have by superposing the P272L variant peptide (PDB 7P3E) into the TCR complex with the Wuhan epitope.
This model showed that the leucine at position 272 in the variant would protrude within 1 Å of the YLQ36 CDR3α loop that forms many important contacts with both peptide and HLA. This would be much closer than the allowed VdW contact distance and would result in repulsive forces. According to Figure 7F, accommodation of the P272L variant by the YLQ36 TCR would require a movement of the CDR3α loop of at least 1.3 Å, lengthening or breaking >25 molecular bonds with the peptide and 4 bonds with the MHC, to potentially result in a substantial loss in binding affinity.
However, the P272L mutant side chain may also adopt a rotamer that would render it clear of the TCR if no other conformational changes occur, though it is unclear whether such rotamers would be energetically favorable. A comparison of the Wuhan and P272L variant peptides may also suggest why the YLQ36 T cell does not recognize the P272L variant.
Structural analysis of HLA A∗02:01-YLQLRTFLL indicates there is a lack of intra-peptide bonds present in the P272L peptide compared to HLA A∗02:01-YLQPRTFLL Figure 7A. This may be due to the greater flexibility of leucine compared to proline, which would allow a reduction in the distance between the Gln271 and Arg273 Cα atoms, resulting in potential clashes between the Gln271 side chain and Gln271/Arg273 main chain, which would be relieved by breaking any bonds formed.
The lack of support in the P272L variant peptide resulting from a lack of intra-peptide bonds may cause the peptide to collapse under YLQ36 TCR approach, displacing the peptide and abolishing peptide:TCR interactions. While our structural data do suggest steric interference may be responsible for the lack of P272L variant recognition (Figure 7F), other factors may also be responsible for this loss of recognition, which would require further study to dissect.
Three other TCR complexes with HLA A∗02:01-YLQPRTFLL have been generated in recent months (Chaurasia et al., 2021; Szeto et al., 2021; Wu et al., 2022), albeit that two of these studies examined the same TCR (Szeto et al., 2021; Wu et al., 2022). The three unique TCRs, YLQ36, NR1C (PDB:7N6E), and YLQ7/YLQ-SG3 (PDB:7N1F/PDB:7RTR), all adopt roughly the same footprint and binding strategy atop HLA A∗02:01-YLQPRTFLL where TCR contacts with peptide are dominated by CDR3α with a strong supporting role from the germline-encoded CDR1α loop (Figure S7).
The NR1C TCR is comprised of TRAV12-1/TRAJ43 CVVNRNNDMRF and TRBV19/TRBJ2-2 CAGQVTNTGELFF chains and binds to HLA A∗02:01-YLQPRTFLL with a KD of 2.71 μM (Chaurasia et al., 2021). Mariuzza and colleagues generated the YLQ7 TCR by pairing TCR α and β chains seen in HLA A∗02:01-YLQPRTFLL tetramer+ CD8 T cells in CPs. While our YLQ36 and the NR1C TCRs use the most common TRAV12-1 V gene seen in HLA A∗02:01-YLQPRTFLL-specific T cells, YLQ7 uses TRAV12-2 (TRAV12-2/TRAJ30 CAVNRDDKIIF; TRBV7-9/TRBJ2-7 CASSPDIEQYF) (Wu et al., 2022). YLQ36 TRAV12-1-encoded CDR1α Ser32 makes 4 contacts with Gln155 of HLA A∗02:01 and 4 contacts with Spike Arg273, including a hydrogen bond.
The CDR1α Ser32 of the NR1C TCR makes 2 contacts with Gln155 (both of which are hydrogen bonds), and the TRAV12-2-encoded CDR1α Ser32 in the YLQ7/YLQ-SG3 TCR makes 3 contacts with Gln155 of HLA A∗02:01, including 1 hydrogen bond, and 2 contacts with Spike Arg273. The common binding mode of these TCRs (Figure S7) and importance of Ser32 provides a molecular explanation for the observed V gene bias towards TRAV12-1 and TRAV12-2 seen in HLA A∗02:01-YLQPRTFLL-specific T cell populations in CPs and vaccinees.
The NRIC TCR was shown to bind to HLA A∗02:01-YLQPRTFLL with a KD of 2.71 μM, very similar to that observed with three other HLA A∗02:01-YLQPRTFLL-specific TCRs in the same study (Chaurasia et al., 2021). The TRAV12-2 YLQ7/YLQ-SG3 TCR was also included in this study but named NR1F. The three studies that included the YLQ7/YLQ-SG3 TCR reported that it bound to HLA A∗02:01-YLQPRTFLL with KD 1.8–6.6 μM (Chaurasia et al., 2021; Szeto et al., 2021; Wu et al., 2022).
These results are of the same order of magnitude and fall firmly within the normal range seen for most interaction with viral peptides (Dolton et al., 2018). The four TCRs examined by Chaurasia et al. bound to the HLA A∗02:01-YLQLRTFLL variant antigen with an average of >68-fold weaker affinity than to the Wuhan HLA A∗02:01-YLQPRTFLL antigen in parallel experiments (Chaurasia et al., 2021).
These findings are in accordance with those of Mariuzza and colleagues who showed that the P272L variant antigen bound to the YLQ7 TCR with >70-fold lower affinity and are consistent with the potential structural rearrangements required to accommodate the longer leucine residue at position 272. Such large reductions in TCR binding are expected to drastically reduce T cell activation and are consistent with our finding that HLA A∗02:01-YLQPRTFLL-specific T cells from CP or vaccinees fail to respond to physiological levels of the P272L variant.
SARS-CoV-2 variants have been rapidly emerging throughout the current pandemic with mutants that enhance transmissibility, evade host immunity, or increase disease severity being of particular concern (Tao et al., 2021). Current systems for identifying variants of public health concern involve risk assessments that identify mutations that are present, and then look for mutations that have a known biological effect in order to assess their significance. While extensive amounts of work have been undertaken to predict and characterize the likely effects of mutations in Spike on antibody recognition, the same is not true for T cells.
Current risk assessments that only consider antibody affinity and factors associated with transmissibility (e.g. mutations that improve cellular infectivity) are clearly incomplete. The range of potential effects of mutations that affect recognition by T cells is significant and urgently requires further study. It is clear that there was a significant SARS-CoV-2 outbreak that spanned Europe of a lineage carrying a mutation that we have shown has a detectable impact on T cell recognition, which went unrecognized at the time.
Ultimately, it is unclear whether SARS-CoV-2 incorporates enough genetic plasticity to allow it to escape from humoral immune responses. However, the area targeted on the RBD by neutralizing antibodies is large enough, and the range of epitopes targeted in polyclonal human sera broad enough, to ensure that no single mutation should allow complete escape from neutralization in the majority of individuals (Rees-Spear et al., 2021).
Likewise, the wide array of HLA across the population (Gonzalez-Galarza et al., 2020) and the broad range of epitopes responded to in COVID-19 patients (Grifoni et al., 2020) combine to make it unlikely that SARS-CoV-2 will completely escape from T cell surveillance in the near future. Over the course of the pandemic to date, it is increasingly clear that mutations that enable the adaptation of SARS-CoV-2 to bind to the human ACE2 receptor or gain entry into human cells have been strongly selected for, with the virus exhibiting significant jumps in fitness with the evolution of each new variant of concern. While a range of mutations conferring resistance to antibody binding to specific parts of the Spike protein have been identified (e.g. E484K), there has been little evidence to date of mutations that may confer an advantage at an epitope targeted by T cells being selected for on a population-wide level.
Examining our data, it is clear that P272L in B.1.177 was on an upward trajectory in the autumn of 2020, but this lineage was ultimately displaced by B.1.1.7/Alpha, which had a much higher growth advantage due to a range of mutations selecting for cell entry and ACE2 binding. It is important to note that although P272L in B.1.177 was clearly increasing in frequency in late 2020, the case numbers concerned, despite being relatively large, would not be sufficient to detect a signal of a growth advantage compared to B.1.177 lineages lacking the P272L variant at the time.
When combined with the analysis of the impact of the P272L change characterized here, it is likely that P272L conferred some advantage, but it was also clear that at the time that advantage was not on the same scale as the selective advantage that was conferred by mutations increasing transmissibility in B.1.1.7/Alpha.
It is important to note, however, that as the pandemic progresses, and SARS-CoV-2 reaches an optimal position with regards to ACE2 binding and cell entry, it is likely that other selective advantages (such as T cell escape) will become more important in the mix of potential advantages that the virus could develop to enable it to persist as a human pathogen.
It is also instructive that P272L has continued to emerge on a local level since its first emergence in early 2020, including in all variants of concern identified to date. Amongst these, P272L emerged and showed local spread in both B.1.1.7/Alpha and B.1.617.2/Delta, including significant numbers of cases in Australia, Italy, and the US. Based upon these observations, we believe that P272L, and other similar mutations that will impact T cell epitopes, will become increasingly important as a route for SARS-CoV-2 lineages to increase their competitiveness, particularly as global vaccination rates increase.
Our data suggest that the 269–277 epitope of Spike is one region that should be monitored, and its impact considered as part of the development of next generation vaccines. Furthermore, the potential of T cell escape presented by mutations in this region would argue for the inclusion of mutations in the 269–277 epitope of Spike in risk assessments by public health agencies going forward.
In summary, we demonstrate that SARS-CoV-2 can readily alter its Spike protein via a single amino acid substitution so that it is not recognized by CD8 T cells targeting the most prevalent epitope in Spike restricted by the most common HLA-I across the population. While it is not possible to directly attribute the emergence and propagation of the Spike P272L SARS-CoV-2 variant in parts of the world where HLA A∗02:01 is frequently expressed to CD8 T cell-mediated selection pressure, specific focusing of immune protection on a single protein (e.g. SARS-CoV-2 Spike favored by all currently approved vaccines [Krammer, 2020]) is likely to enhance any tendency for escape at predominant T cell epitopes like YLQPRTFLL. Our demonstration that mutations that evade immunodominant T cell responses through population-frequent HLA can readily arise and disseminate, strongly suggests that it will be prudent to monitor such occurrences and to increase the breadth of next generation SARS-CoV-2 vaccines to incorporate other viral proteins.