The first two studies that experimentally investigated possible CCCoV-induced T-cell memory against SARS-CoV-2 were Grifoni et al. (2020) and Braun et al. (2020) ( Leslie, 2020). Braun et al. (2020) tested CD4 ⁺ T cell activation using S (spike) protein derived peptide pools, whereas Grifoni et al. (2020) investigated both CD4 ⁺ T and CD8 ⁺ T cell activations using peptide pools derived from various SARS-CoV-2 proteins. Both groups found SARS-CoV-2 peptide pools to activate T cells from healthy donors (HD), proposedly representing memory T cells primed by similar peptides during earlier CCCoV infections. Notably, T cells of HD were also activated if their peripheral blood mononuclear cells (PBMC) were incubated with pools of peptides derived from not very well conserved SARS-CoV-2 proteins such as S ( Braun et al., 2020; Grifoni et al., 2020), although—depending on the virus isolates—SARS-CoV-2 S protein does not share identical 9-mers with any of the CCCoVs but may share two identical 8-mer sequences ( Dijkstra & Hashimoto, 2020). Braun et al. (2020) assumed that even a <50% identity between the corresponding SARS-CoV-2 and CCCoV peptides might explain the assumed cross-virus CD4 ⁺ T cell memory. Similarly, the authors of the Grifoni et al. (2020) study—answering our question on how S-derived peptides could activate cross-virus CD4 ⁺ and CD8 ⁺ T cell memory despite not sharing identical 9-mers or (presumably) immunogenic identical 8-mers—explained that CD8 ⁺ T cell activation in their type of in vitro assay could be found for a substantial part of any peptides sharing only ≥70% identity and that for CD4 ⁺ T cell activation the requirements for peptide similarity were even lower (see comments section below the Grifoni et al., 2020 article at https://www.cell.com/cell/fulltext/S0092-8674(20)30610-3). Although it has been well established that T cells can be promiscuous in recognizing pMHC (MHC + peptide) complexes, the extents to which this is relevant for in vivo T cell memory and does or does not tend to involve peptides with very similar sequences are being debated (e.g., Grant et al., 2016; Petrova et al., 2012). As explained by Peng et al. (2020), in vitro observations that suggested the existence of CCCoV-primed anti-SARS-CoV-2 “cross-reactive” T cell memory might alternatively be caused by the activation of naive T cells or of T cells primed by not-so-similar peptides of non-related pathogens. Possibilities for explaining in vitro observations suggesting cross-virus T cell memory in the case of non-identical peptides are summarized in Figure 2. Naturally, the proportions of false negative and false positive outcomes depend on the sensitivity of the assay. The best chance for in vitro results truly representing stimulation of the same T cells (TCR-identical T cells) activated during CCCoV and subsequent SARS-CoV-2 infections—hence representing functional memory—is if the MHC-presented peptides are identical. Hence, the present article predominantly focuses on such identical peptides.

Table 1 summarizes the Grifoni et al. (2020) and Braun et al. (2020) studies, as well as later studies that experimentally investigated potential SARS-CoV-2 T cell epitopes. Section A of the table summarizes studies that only investigated peptide pools and/or intact proteins, and section B summarizes the studies that (also) investigated individual peptides. The table explains whether indications for CD4 ⁺ and/or CD8 ⁺ T cell activation were observed, whether MHC association was addressed and/or MHC binding tested, and whether indications for SARS-CoV-2 and CCCoV cross-virus T cell memory were investigated and observed. Importantly, the consensus was that anti-SARS-CoV-2 T cell memory in individuals that had only been infected with CCCoVs was weak and only detectable in a subset of individuals ( Sette & Crotty, 2020). It should be noted that none of the studies listed in Table 1 had specifically selected individuals with a known recent history of CCCoV infection, although presumably in such group a stronger T cell response against SARS-CoV-2 antigens would be found ( Sagar et al., 2021).

Table 2 lists the subset of the 9-mers that are identical between SARS-CoV-2 and at least one of the CCCoVs ( Dijkstra & Hashimoto, 2020) and additionally were predicted to bind representative MHC-I alleles or experimentally found to bind MHC-I, be part of MHC-II binding peptides, or to activate T cells. It also lists >9-mers for which evidence of T cell activation was reported. The experimental results summarized within Table 2 were collected from the articles listed in Table 1, from articles on SARS-CoV-1, and from reports uniquely deposited to the Immune Epitope Database and Analysis Resource (IEDB; http://www.iedb.org/; Dhanda et al., 2019). The 9-mer peptides listed in Table 2 are referred to in Table 1 by using the letter code of their N-terminal three amino acids.

The only identical T cell epitope repeatedly found to be immunogenic by independent research groups was peptide “VYI” (VYIGDPAQL) (highlighted in bold in Table 1; details in Table 2). The VYI peptide is shared between SARS-CoV-2, HCoV-HKU1, and HCoV-OC43 viruses and part of a larger identical stretch AKHYVYIGDPAQLPAPR in their helicase protein (aka nonstructural protein 13 or NSP13). Prachar et al. (2020) showed that the VYI peptide bound to HLA-A*24:02, although not very stable. To our knowledge, all studies that investigated the VYI peptide—three in total—found it to stimulate T cells in an HLA-A*24:02 context ( Ferretti et al., 2020; Kared et al., 2020; Nelde et al., 2021). Ferretti et al. (2020) investigated the entire SARS-CoV-2 proteome by a “T-scan” assay measuring activation of CD8 ⁺ memory T cells from COVID-19 convalescent donors after incubation with HEK293 cells engineered to express a single HLA allele and one of a set of overlapping 61 aa stretches. By this method, Ferretti et al. (2020) detected only three SARS-CoV-2 stretches with “dominant” epitopes that activated CD8 ⁺ T cells from multiple HLA-A*24:02 ⁺ COVID-19 convalescent donors; one of these three stretches encompassed the VYI peptide and stimulated two of five investigated samples. Involvement of the VYI peptide was confirmed by activation of the HLA-A*24:02 ⁺ T cells upon coculturing with HLA-A*24:02 ⁺ target cells pulsed with VYI peptide. If for the T-scan screen 61 aa stretches of CCCoV proteomes were used instead, Ferretti et al. (2020) appear to have found a weak but noticeable response by HLA-A*24:02 ⁺ memory CD8 ⁺ T cells in the cases of HCoV-HKU1 and HCoV-OC43 (our interpretation of their Figure 5A). Kared et al. (2020) performed a binding assay, testing 94 peptides from across the SARS-CoV-2 proteome—predicted by software to bind HLA-A*24:02— using pHLA-A*24:02 tetramers for labeling of CD8 ⁺ T cells from five HLA-A*24:02 ⁺ COVID-19 convalescent donors. They found positive reactions for only eight of the 94 peptides. One of these eight peptides was the VIY peptide, which detectably labeled T cells of only one of the five donors. Regarding possible CCCoV-induced anti-SARS-CoV-2 T cell memory, Kared et al. (2020) mentioned “Notably, SARS-CoV-2 specific CD8 ⁺ T cells were not detected in any of the healthy donors recruited before the official SARSCoV-2 pandemic;” their number of HD controls, however, was low, and for their HLA-A*24:02-matched experiments seems to have been between one and four. Nelde et al. (2021) tested ten predicted HLA-A*24:02 SARS-CoV-2 epitopes and found the VYI peptide to be one of the three “dominant T cell epitopes” as it elicited activation of CD8 ⁺ T cells from seven of ten HLA-A*24:02 ⁺ COVID-19 convalescent donors upon incubation with their PBMC. On the other hand, for PBMC of 17 healthy HLA-A*24:02 ⁺ donors a stimulation of CD8 ⁺ T cells could not be observed. To summarize these three studies, the VYI peptide is among the most immunogenic SARS-CoV-2 T cell epitopes in HLA-A*24:02 ⁺ individuals, although there is no evidence yet that this was primed by previous CCCoV infections. Presumably, because of the latter, none of the three studies mentioned that VYI peptide is shared between SARS-CoV-2, HCoV-HKU1, and HCoV-OC43 ( Ferretti et al., 2020; Kared et al., 2020; Nelde et al., 2021). We assume that not finding anti-VYI T cells in HLA-A*24:02 ⁺ HD was only a matter of assay sensitivity, because CCCoV-induced T cell memory is expected as the VYI peptide is embedded in a longer identical AKHYVYIGDPAQLPAPR stretch shared between SARS-CoV-2, HCoV-HKU1, and HCoV-OC43; thus, the MHC-I pathway processing of the peptide is expected to be similar in each viral background, and software predicts that the immunoproteasome efficiently generates the VYI peptide from all three viruses ( https://imed.med.ucm.es/Tools/pcps/; Gomez-Perosanz et al., 2020). An additional reason for assuming the involvement of CCCoV-induced T cell memory is that the helicase is not one of the more abundant (structural) viral proteins ( Davidson et al., 2020) whereas nevertheless VYI is consistently found as one of the dominant T cell epitopes. Future experiments selectively investigating HLA-A*24:02 donors with a recent HCoV-HKU1 or HCoV-OC43 infection will probably be more sensitive in detecting anti-VYI CD8 ⁺ T cells primed by CCCoV infection.

HLA-A*23:01 belongs to the same “supertype” as HLA-A*24:02 ( Lund et al., 2004) meaning that they tend to bind similar peptides, and in the IEDB database it is reported that HLA-A*23:01 also binds VYI peptide (IEDB Reference:1000425). It has not been described yet whether VYI peptide can stimulate T cells in an HLA-A*23:01 context.

Although not identical throughout the sequence, as an exception, Table 2 also lists the 9-mer SPRWYFYYL (“SPR”) for SARS-CoV-2 and its matching LPRWYFYYL (“LPR”) for HCoV-HKU1 and HCoV-OC43. Several independent studies ( Table 1; SPR indicated in bold) suggest that SPR is highly immunogenic and involved in cross-virus immune memory (summarized in Table 2). The only amino acid difference between the SPR and LPR peptides is at the P1 position, which is not necessarily important for peptide conformation in pMHC-I complexes or for binding T cell receptors (TCRs) (e.g., Sewell, 2012). Thus, it seems plausible that a fraction of the T cells primed by pMHC/LPRWYFYYL may also recognize pMHC/SPRWYFYYL. It was convincingly shown that SPR peptide binds HLA-B*07:02 alleles and in that context can stimulate CD8 ⁺ T cells from COVID-19 convalescent donors ( Ferretti et al., 2020; Kared et al., 2020; Peng et al., 2020; Schulien et al., 2021; Snyder et al., 2020). CD8 ⁺ T cells from HLA-B*07:02 ⁺ HD could also be stimulated by SPR, suggesting cross-virus CD8 ⁺ T cell memory induced by the LPR peptide of HCoV-HKU1 or HCoV-OC43 ( Schulien et al., 2021), which is consistent with in vitro results showing that CD8 ⁺ memory T cells from HLA-B*07:02 ⁺ COVID-19 convalescent donors were activated by HLA-matched cells expressing 61 aa fragments of HCoV-HKU1 or HCoV-OC43 encompassing the LPR peptide ( Ferretti et al., 2020). However, SARS-CoV-2 peptides encompassing or even only partially overlapping the SPR peptide also induced responses—remarkably strong in some cases—of CD4 ⁺ T cells from COVID-19 convalescent donors and, to a lesser extent, from HD (also summarized in Table 2; Le Bert et al., 2020; Nelde et al., 2021; Peng et al., 2020; Tarke et al., 2020). The combined results suggest an overlap of highly immunogenic MHC-I and MHC-II epitopes, and—although such overlap is not impossible—some caution for the possibility that SPR peptide might (additionally) cause non-MHC-restricted immune stimulation seems to be warranted. Except for HLA-B*07:02, SPR peptide was also found to bind the MHC-I alleles HLA-B*51:01, -*53:01, and -*54:01 (IEDB database Reference:1000425).

Other than VYI, we did not find any other identical peptides shared between SARS-CoV-2 and CCCoVs for which current publications convincingly indicate a high immunogenicity. However, the 9-mers FVDGVPFVV (“FVD”) and LPYPDPSRI (“LPY”) could be promising, although they were only reported as immunogenic CD8 ⁺ T cell epitopes in single publications. Snyder et al. (2020) appear to have identified FVD peptide as an immunodominant SARS-CoV-2 T cell epitope in the HLA-A*02:01 context, although their descriptions of this matter could be more detailed. Tarke et al. (2020) found that CD8 ⁺ T cells from a few HLA-B*51:01 ⁺ COVID-19 convalescent donors could be stimulated by several peptides that encompassed the LPY 9-mer peptide sequence, namely peptide DYVYLPYPDPSRI (a 13-mer shared between SARS-CoV-2 and HCoV-HKU1), its shorter versions VYLPYPDPSRI (an 11-mer) or YLPYPDPSRI (a 10-mer), or peptide LPYPDPSRIL (a 10-mer); software predicts that the encompassed LPYPDPSRI (a 9-mer) is the best HLA-B*51:01 binder and that of the other lengths only the 10-mers YLPYPDPSRI and LPYPDPSRIL are expected to bind this allele ( Table 2), suggesting that some processing of the longer peptides may explain the combined results. Future analysis of the immunogenicity of the LPY 9-mer peptide would be interesting.

In addition to the above, Mateus et al. (2020) and Tarke et al. (2020) found some other SARS-CoV-2 peptides that encompassed ≥9-mers shared with CCCoVs and stimulated CD4 ⁺ T cells ( Table 2), but whether those identical stretches formed the immunogenic epitope has not been determined yet.

For speculation on the global implications of the ability of different MHC-I alleles to bind conserved peptides, their global distributions must be appreciated. The online Allele Frequency Net Database (AFND; http://allelefrequencies.net/; Gonzalez-Galarza et al., 2020) comprises information of MHC-I allele frequencies in different populations, and Figure 3 includes their visual summaries of global distributions of the alleles HLA-A*24:02 (binder of VYI peptide), HLA-B*07:02 (binder of SPR peptide), HLA-A*02:01 (binder of FVD peptide), and HLA-B*51:01 (predicted binder of LPY peptide). Table 3 lists populations with the highest frequencies of HLA-A*24:02 according to AFND. For all studies summarized in Table 3 the allele frequencies and for a subset also the percentage of individuals carrying the allele had been determined (note that if randomly distributed, an allele frequency of >0.29 would correspond to a prevalence of >50%). Data for populations in Japan and for Japanese in the USA are highlighted in shades of gray ( Table 3). In Japanese populations the HLA-A*24:02 allele frequencies are ≥0.33 except for in the Ainu which are a racial minority indigenous to Hokkaido in Northern Japan. Ikeda et al. (2015; the “Japan pop 16” study in Table 3) investigated 18604 individuals from “all parts of Japan” and the authors concluded that the HLA-A*24:02 allele is distributed in ~60% of the Japanese population. Other populations in which HLA-A*24:02 is present in >50% of the investigated members are various indigenous populations in Asia, Oceania, and the Americas ( Table 3). Those non-Japanese populations are not further discussed in this study as it is harder to collect their data relevant to COVID-19.

HLA-B*07:02 is common in Northern Europe ( Figure 3) and is carried by approximately a third of the Irish population (reports at AFND). HLA-A*02:01 is common in Europe, and, for example, found in 49% of the Polish population (report at AFND). HLA-B*51:01 is common in Southern Europe and the Middle East ( Figure 3) and was found at an allele frequency of 0.19 in Saudi Arabia (report at AFND).

A year into the pandemic, the accumulated number of COVID-19 deaths per million inhabitants in Japan is 45, which is >30 times fewer than in countries such as Italy (1465), the UK (1559), and the USA (1362) (on February 1st, 2021, according to https://www.worldometers.info/coronavirus/). At least for the initial wave of the disease in the first half of 2020 the apparent low prevalence of COVID-19 was supported by finding specific antibodies in only ~0.1% of citizens of Tokyo in June 2020 (governmental report https://www.mhlw.go.jp/content/000648706.pdf) and minimal or even negative excess mortality rates until at least July 2020 ( Yorifuji et al., 2021). These low numbers are quite surprising since the stringency of behavioral regulations to protect against COVID-19 have been less severe in Japan than in most Western countries ( https://www.bsg.ox.ac.uk/research/research-projects/coronavirus-government-response-tracker; Hale et al., 2020). The surprise about the relatively low incidence of COVID-19 in Japan was captured well in the title of a BBC article on July 2020 “Coronavirus: Japan's mysteriously low virus death rate” ( https://www.bbc.com/news/world-asia-53188847). Most of the difference with Western countries can probably be explained by voluntary behaviors such as the willingness of the Japanese to wear masks (despite absence of obligation) and by better individual health status such as a relatively low prevalence of obesity. As a note, however, the Japanese are not per se better protected against respiratory viruses, as mortalities per capita resulting from the 2009 influenza pandemic were higher than in European countries ( Simonsen et al., 2013).

In Japan, as in other countries, CCCoV infections are poorly monitored, but available data indicate that from the winter 2014–2015 until November 2019 (we were not able to find more recent data) the most common CCCoV species in Japan has been HCoV-OC43. Especially during the winters of 2014–2015 and 2018–2019, this virus species appears to have been prevalent ( Komabayashi et al., 2020; Kubota-Koketsu et al., 2020). Thus, many Japanese individuals probably received a relatively recent immune boost with the VYI peptide.

Currently, in the winter 2020–2021, there has been a surge in the number of COVID-19 cases and deaths in Japan ( https://www.worldometers.info/coronavirus/; https://www.aljazeera.com/news/2021/1/4/japan-weighs-state-of-emergency-amid-severe-covid-19-surge), and it is unclear whether the factors that during the first half of 2020 protected the Japanese better than the populations of many other countries are still in place. Possibly, the warm winter season of 2019–2020 (Japan Meteorological Agency reports https://www.data.jma.go.jp/obd/stats/data/en/smp/index.html), and the associated early hay fever season ( https://global.weathernews.com/news/13178/; http://kafun.taiki.go.jp/), may have helped to protect the Japanese population during the first half of 2020. In Japan, largely because of aging monoculture coniferous forests, depending on the definition approximately half of the population may be considered to suffer from pollen-induced allergic rhinitis, which has been called a national affliction (e.g., Minami et al., 2019; Nakamura et al., 2019; Yamada et al., 2014). Although speculative, the associated inflammation of the respiratory tract might non-specifically elevate both innate and specific (possibly anti-VYI T cells) immunity against COVID-19. Although each of these individual biological factors probably is not very protective against COVID-19 (see also the below paragraph), at the population scale a combination of such factors might significantly impact the virus reproduction number (R).

In short, (i) most Japanese individuals possess the MHC-I allele HLA-A*24:02 that presents a highly immunogenic SARS-CoV-2 T cell epitope VYI, (ii) in recent years many of them have been exposed to this epitope by HCoV-OC43 infection, and (iii) at the time of the first COVID-19 wave many of them had an elevated immune status of their respiratory tracts because of pollen allergy.

MHC polymorphism is believed to be driven by differences in immune responses conferred by the different alleles, but actual evidence for this to cause differences in disease resistance is close to absent ( Kelly & Trowsdale, 2017; Yamaguchi & Dijkstra, 2019). Presumably, this is caused by each set of MHC alleles having enough choice within a pathogen proteome for presenting some peptides efficiently to the immune system. Theoretically, the MHC allelic effect on differences in disease resistance should become larger if the choice of possible immunogenic epitopes becomes more limited. Hence, the effect of MHC polymorphism on immune memory induced against a related virus should be larger than on immune memory against the same virus. However, several studies have investigated the association of MHC polymorphism with differences in COVID-19 resistance, and—arguably—no convincing associations have yet been presented ( Iturrieta-Zuazo et al., 2020; Littera et al., 2020; Lorente et al., 2021; Wang et al., 2020). From these combined within population studies, however, it probably follows that HLA-A*24:02 does not necessarily have a detectable positive or negative impact on the two commonly analyzed COVID-19 parameters, which are the frequency of contracting COVID-19 and the severity of the disease. As far as we know, there has not been a thorough investigation yet of a possible association between MHC polymorphism and the virus titers in the upper respiratory tract. The severity of COVID-19 is predominantly determined by whether the virus infects the lower respiratory tract ( Cyranoski, 2020), which, curiously, has been described as largely disconnected from the level of virus replication in the upper respiratory tract ( He et al., 2020; Lavezzo et al., 2020). He et al. (2020), for example, stated “There was no obvious difference in viral loads across sex, age groups and disease severity.” Therefore, if cross-virus memory T cells would not be sufficient to help block an infection but instead help to expedite the end of upper respiratory tract infections, the HLA-A*24:02 allele might give protection at the population level by limiting spread without having an impact on the COVID-19 parameters typically studied in the HLA-association studies (contraction and severity at the level of individuals). We speculate that HLA-*24:02-restricted anti-VYI CD8+ T cell immune memory—especially if recently boosted by HCoV-OC43 or HCoV-HKU1 infections or nonspecific immune stimulations—can reduce the total number of virus particles secreted by COVID-19 patients and so the replication number (R) of the virus at the population level.