Human Paramyxovirus Infections Induce T Cells That Cross-React with Zoonotic Henipaviruses

Humans encounter multiple paramyxoviruses early in life. This study shows that infection with common paramyxoviruses can induce T cells cross-reactive with the highly pathogenic Nipah virus. This demonstrates that the combination of paramyxovirus infection history and HLA haplotype affects immunity to phylogenetically related zoonotic paramyxoviruses.

Effective live-attenuated vaccines against measles and mumps are available and incorporated into most national vaccination programs. Furthermore, multiple encounters with HPIV, HMPV, and HRSV occur within the first 10 years of life (3)(4)(5). These viruses are major causes of lower respiratory tract infections during childhood. Vaccine development is well under way, especially for HRSV, but no licensed vaccines against these respiratory viruses exist.
In addition to the paramyxo-and pneumoviruses largely restricted to humans, several members of the Paramyxoviridae that infect animals have zoonotic potential (6)(7)(8)(9). Nipah virus (NiV) is of particular interest, as it is associated with high fatality rates, classified as a biosafety level 4 (BSL-4) pathogen, and prioritized by the World Health Organization (WHO) as a pathogen for which vaccines are urgently needed (7,10,11) (https://www.who.int/activities/prioritizing-diseases-for-research-and-development-in -emergency-contexts). Large outbreaks of NiV infection have occurred in Bangladesh and Malaysia with severe neurological disease in humans, case fatality rates up to 75%, and considerable human-to-human transmission (7,8,12). NiV circulates in fruit bats, a reservoir host that is widely distributed over the southern hemisphere. Combining this with a possibility of viral spread via the respiratory route of certain NiV strains, suggested by epidemiological studies and supported by in vitro observations and animal experiments, it is not without reason that NiV is one of the viruses marked by the WHO with potential to cause a future pandemic through natural or deliberate exposures (13)(14)(15)(16). NiV is not the only paramyxovirus with zoonotic potential: the closely related Hendra virus (HeV) has caused lethal disease in humans in the past (17). In addition, virus discovery studies have identified an increasingly diverse range of henipaviruses and other paramyxoviruses in nonhuman reservoirs, suggesting potential for additional spillover events (18)(19)(20).
Vaccination against or infection with members of the Paramyxoviridae and Pneumoviridae induces virus-specific B and T cell responses. Virus-specific neutralizing (VN) antibodies are considered the main correlate of protection for most of the Paramyxoviridae (21). However, once a susceptible host is infected, viral clearance is predominantly mediated by cellular immune responses. This is probably due to the spread via cell-to-cell fusion and infected migrating cells (22), thus avoiding VN antibodies in the bloodstream. Whereas VN antibodies in these infections are exclusively targeted to membrane-exposed epitopes, T cells can target epitopes contained in any viral protein.
The importance of CD8 ϩ T cells in MeV clearance was corroborated in vivo; MeVinfected macaques depleted of B cells were able to normally clear virus infection (23), whereas CD8 ϩ T cell-depleted macaques presented with higher viral loads and a significantly longer duration of viremia (24). It has also been reported that children with deficits in cellular immunity suffer from more severe and prolonged measles (25).
Human and animal members of the genus Morbillivirus are phylogenetically closely related, and it is well known that MeV vaccination or infection can induce (partial) cross-protection from infection with heterologous morbilliviruses (26)(27)(28). In classical studies, MeV vaccines were used to immunize dogs against the closely related canine distemper virus (CDV), which normally causes lethal disease in dogs. MeV vaccination resulted in partial protection from clinical signs after CDV challenge infection, in the absence of cross-reactive antibodies (29)(30)(31). CDV was shown to be able to cross the species barrier into nonhuman primates (6,32,33). In experimental infections of nonhuman primates, MeV-vaccinated macaques also proved partially protected from CDV challenge in the absence of CDV-specific VN antibodies. Most importantly, vaccination resulted in strongly reduced levels of virus shedding from the upper respiratory tract. A rapid proliferation of white blood cells in MeV-vaccinated macaques upon CDV challenge was detected, reminiscent of a secondary cellular immune response (34). Combined, these data strongly suggest that cellular immune responses triggered by MeV infection or vaccination are an important correlate of cross-protection against infection with heterologous morbilliviruses.
In the present study, we explored if cellular cross-reactivity exists across the different genera of the Paramyxoviridae and Pneumoviridae. We investigated whether broadly reactive T cells are present in human donors and assessed whether these have the capacity to cross-react with endemic paramyxoviruses (like MeV, HPIV1 to -4, and MuV), pneumoviruses (HMPV and HRSV), and potentially zoonotic viruses (like CDV, NiV, and HeV). We now show that exposure to human endemic paramyxoviruses, by either vaccination or infection, can lead to cross-genus-reactive T cell immunity. We specifically show that these broadly reactive T cells also recognize NiV and thus may confer protection against the highly pathogenic zoonotic henipaviruses.
de Vries et al. CD8 MeV 1), but two MeV F-specific TCC from different donors cross-reacted with CDVinfected cells in an IFN-␥ enzyme-linked immunosorbent spot (ELISPOT) assay ( Fig. 2A and B). One of these was CD4 ϩ (CD4 Xreact 1), and the other was CD8 ϩ (CD8 Xreact 1). The cross-reactive cytotoxic CD8 ϩ TCC not only produced IFN-␥ upon recognition of MeVand CDV-infected cells (37) but also proved capable of clearing MeV and CDV infection from autologous B-lymphoblastic cell lines (B-LCL) in an in vitro virus suppression assay (38). In the same assay, CD8 MeV 1 could clear only MeV infection but not CDV infection (Fig. 2C). The CD4 ϩ TCC were also tested in the suppression assay, but neither TCC could clear either MeV or CDV infection (data not shown), confirming that these TCC were not cytotoxic. These data illustrate that morbillivirus-specific cross-reactive T cells exist in humans and that cytotoxic cross-reactive T cells can clear infection with heterologous morbilliviruses in vitro.
Cross-reactive TCC recognize epitope in the F protein. CD4 Xreact 1 and CD8 Xreact 1 were previously shown to respond to antigen (Ag)-presenting cells (APC) expressing the F protein (37). By using peptide-pulsed autologous B-LCL with overlapping peptides covering the F protein, and measuring IFN-␥ production via ELISPOT, we identified the minimal amino acid sequence required for activation of both TCC. Initially, overlapping 15-mer peptides (11 overlap) were used ( Fig. 3A and D), followed by fine-tuning with overlapping 10-mers (9 overlap) ( Fig. 3B and E). Finally, we determined that both CD4 Xreact 1 and CD8 Xreact 1 recognized the minimal 9-mer F 129 -137 (F AQITAGIAL for MeV, Fig. 3C and F). By performing cocultures with different B-LCL with known HLA haplotypes, we determined CD4 Xreact 1 to be restricted by HLA-DQ*06:03 and CD8 Xreact 1 to be restricted by HLA-B*15:01 (data not shown). When binding predictions were performed for 9-or 10-mer epitopes with the MeV F protein sequence for the HLA class I reference set (additionally including HLA-B*27:05 and HLA-B*39:01 [https://www.iedb.org/]), four different HLA class I molecules were predicted to bind F AQITAGIAL at high affinity (50% After finding that TCC CD4 Xreact 1 and CD8 Xreact 1 recognize the minimal epitope F 129 -137 , we determined the level of conservation of this epitope among all morbilliviruses. Interestingly, F 129 -137 was completely conserved between MeV, CDV, and rinderpest virus (RPV) but was also highly conserved among the other morbilliviruses (peste-des-petits-ruminants virus [PPRV], cetacean morbillivirus [CeMV], phocine distemper virus [PDV], and feline morbillivirus [FeMV]). The most obvious variation was at position F 135 , presenting as either isoleucine or valine, two similar amino acids with hydrophobic side chains (see Fig. S1 in the supplemental material).
F AQITAGIAL is part of the conserved fusion peptide. After determining conservation of F 129 -137 among morbilliviruses, we additionally aligned the F proteins of all Paramyxoviridae and Pneumoviridae. A phylogenetic tree was constructed on the basis of the F nucleotide sequence, revealing the different subfamilies and genera (Fig. S2A). Subsequently, having found CD4 Xreact 1 and CD8 Xreact 1 to be specific for F 129 -137 , we took a closer look at this specific region in the F protein. The F protein is normally formed as an inactive F 0 variant. For the F protein to become active, it must be cleaved by furin-like enzymes into F 1 and F 2 at the cleavage site. The N-terminal region of the F1 subunit adjacent to the cleavage site is called the fusion peptide (region F 113-137 ), a hydrophobic region that plays a critical role in the fusion process by insertion into the membrane of target cells (39)(40)(41). When aligning the fusion peptide of the relevant endemic and zoonotic paramyxo-and pneumoviruses (MeV, CDV, NiV, HeV, HPIV1 to -4, MuV, HMPV, and HRSV), we again found that the fusion peptide, and especially F 129 -137 , was highly conserved among the Paramyxoviridae (Fig. S2B), suggesting a functional constraint in virus evolution of this region in the protein.
Novel F 129 -137 -specific CD8 ؉ TCC could be isolated from human donors. Since we know that CD8 ϩ T cells are crucial in the clearance of especially paramyxoviruses in vivo (23,24), and we had thus far relied on a single CD8 ϩ TCC that recognized a conserved region in the fusion peptide, we isolated novel F 129 -137 -specific CD8 ϩ TCC from peripheral blood mononuclear cells (PBMC) obtained from HLA-B*15:01-positive donors. To this end, we used phycoerythrin (PE)-labeled B*15:01-F AQITAGIAL tetramers, which could stain F AQITAGIAL -specific TCC (Fig. 4A). Three donors were selected, and PBMC were stained with the F AQITAGIAL tetramer (Fig. 4B). CD3 ϩ CD8 ϩ AQITAGIAL ϩ cells were single cell sorted, and F AQITAGIAL -specific TCC could be clonally expanded from a single donor. A single TCC was selected for subsequent experiments (CD8 Xreact 2) (Fig. 4C). Both CD8 Xreact 1 and CD8 Xreact 2 proved cross-morbillivirus-reactive and were shown to recognize both morbillivirus sequence F AQITAGIAL present in MeV, CDV, and RPV and the variant F AQITAGVAL (valine instead of isoleucine at 7th residue) present in PPRV, CeMV, and PDV, in a concentration-dependent manner (Fig. S3). CD8 Xreact 1 reacted strongly with F AQITAGIAL but also with F AQITAGVAL . CD8 Xreact 2 had a stronger affinity to F AQITAGVAL but still recognized F AQITAGIAL .
CD8 Xreact 1 and CD8 Xreact 2 are not identical TCC. Both CD8 Xreact 1 and CD8 Xreact 2 were HLA-B*15:01 restricted and reactive to F 129 -137 ; however, they did not recognize the same peptides. This suggested that the TCC were not identical. To confirm this, and to confirm clonality of the TCC used in different assays, we assessed the T cell receptor (TCR) variable (V) ␤ chain expression by flow cytometry. A panel of antibodies confirmed that the TCC were indeed different, as CD8 Xreact 1 expressed TCRV ␤ 14 and CD8 Xreact 2 expressed TCRV ␤ 8 (Fig. S4).
Reactivity to cells pulsed with different peptides (Fig. 5A), different V ␤ chain expression (Fig. S4), and variable affinity for F AQITAGIAL and F AQITAGVAL (Fig. S3) proved that CD8 Xreact 1 and CD Xreact 2 were different TCC, probably originally induced by different paramyxoviruses. We again performed peptide dilution series with pulsed autologous antigen-presenting cells and found that CD8 Xreact 1 recognized the epitope present in morbilliviruses and henipaviruses (F AQITAGIAL and F AQITAGVAL , respectively) at high af- finity and the epitope present in HPIV4 (F AQVTAAIGL ) to a lesser extent (Fig. 5B). CD8 Xreact 2 showed a different spectrum of reactivity: although it was reactive with the epitope present in morbilliviruses, henipaviruses, and MuV, affinity for HPIV2 and -3 (F AQITAAVAL and F AQITAAVAI , respectively) was much stronger (Fig. 5C).
As the peptide dilution experiments indicated strong affinity of particular peptides, we developed a novel flow cytometry-based assay to truly determine interaction strength (i.e., avidity) between peptide and TCC, by staining TCC with a saturating amount of HLA-B*15:01 tetramers conjugated with peptides of relevant paramyxoviruses. CD8 Xreact 1 and CD8 Xreact 2 were stained with the different F 129 -137 tetramers and briefly treated with a low concentration of urea, disrupting low-avidity binding between tetramers and TCC ( Fig. S5A and B). Percentage of binding and interaction strength at 2 M (based on mean fluorescence intensity [MFI] at 0 M) proved that CD8 Xreact 1 had the highest affinity for F AQITAGIAL (and was therefore probably induced by MeV), whereas CD8 Xreact 2 had the highest affinity for F AQITAAVAL and F AQITAAVAI (and was therefore probably induced by infection with HPIV2 or HPIV3) ( Fig. 5D and E).
F 129 -137 -specific TCC clear NiV-infected cells. Thus far, we had only characterized the cross-reactive TCC for their capacity to bind peptides or tetramers based on HLA-B*15:01 complexed with F 129 -137 and their potential to secrete IFN-␥ upon antigen (Ag)-specific stimulation. Both CD8 Xreact 1 and CD8 Xreact 2 were identified as TCC induced by endemic paramyxoviruses that could potentially play a role in cross-immunity against zoonotic paramyxoviruses but do not likely affect infections with distantly related pneumoviruses. To confirm the capacity of CD8 Xreact 1 and CD8 Xreact 2 to actually suppress NiV replication, we performed the aforementioned virus suppression (shown in Fig. 2C) (38). Autologous or HLA-matched B-LCL were infected with green fluorescent protein (GFP)-expressing NiV and cocultured with a concentration series of TCC. The absolute number of TCC required to suppress NiV replication was determined (Fig. 6A to C). Simultaneously, the TCC were tested for their capacity to kill cells infected with MeV, MuV, CDV, and HPIV3 ( Fig. 6A to C) and HMPV and HRSV (data not shown). HLA-mismatched antigen-presenting cells were included as a negative control. CD8 Xreact 1 efficiently suppressed MeV and CDV replication (Fig. 6A), corresponding with reactivity in IFN-␥ ELISPOT (Fig. 6A). CD8 Xreact 2 suppressed MeV and CDV replication but could additionally suppress HPIV3. Surprisingly, MuV replication was not suppressed, although this TCC did react with the F 129 -137 epitope present in MuV (Fig. 6A). Importantly, both MeV-specific TCC CD8 Xreact 1 and HPIV-specific TCC CD8 Xreact 2 killed NiV-infected cells (Fig. 6B), confirming functionality against highly pathogenic henipaviruses. Corresponding to peptide reactivity, none of the TCC killed cells infected with the Pneumoviridae member HRSV or HMPV.
Conserved regions are present throughout the paramyxovirus proteome. After identifying F 129 -137 as a cross-reactive epitope to which HLA-B*15:01-restricted T cell ® responses can be induced by multiple paramyxoviruses, we performed a systematic search for conserved regions throughout the proteome of the Paramyxoviridae and Pneumoviridae, containing potential cross-reactive T cell epitopes. Since we were mostly interested in T cells induced by endemic paramyxoviruses that cross-react with henipaviruses, we chose NiV as a reference virus and aligned the six common proteins to MeV, HPIV3, and HPIV2. Conserved regions, defined as stretches of 10 homologous amino acid residues allowing 4 mismatches, were found in all comparisons (NiV versus MeV, 119 regions; NiV versus HPIV3, 93 regions; NiV versus HPIV2, 59 regions) and were most abundant in the L protein ( Fig. 7A and Data Set S1). Alignments for all the endemic and zoonotic paramyxo-and pneumoviruses were prepared for these 271 regions of interest in total, and the average homology percentage was determined. When only regions with an average of more than 50% homology among all viruses were selected, we identified 27 regions in total that potentially contain T cell epitopes that are conserved throughout the paramyxovirus proteome (Data Set S1E).
Using these definitions to screen for conserved regions of interest, we identified one region in N, one region in F, and 25 regions in L (Fig. 7B). Confirming the validity of this unbiased approach, F 117-134 (NiV numbering, contains the full fusion peptide and F 129 -137 in MeV numbering) was indeed identified as a region of interest. Using NiV as a reference, we determined how conserved all regions of interest with potential epitopes were in the endemic and zoonotic paramyxo-and pneumoviruses. All epitopes were almost completely conserved in the closely related HeV, and Ͼ50% conservation was found for HPIV2, HPIV4, and MuV; Ͼ60% conservation was found for HPIV1 and -3; and Ͼ70% conservation was found for MeV and CDV (Fig. 7B). Altogether, this suggests that additional cross-reactive CD8 ϩ T cells targeting evolutionarily conserved peptide sequences can be induced by heterologous infections.

DISCUSSION
Humans are repeatedly exposed to members of the Paramyxoviridae and Pneumoviridae, either by natural infection or by vaccination. Most of these viruses are encountered in the first years of life, including the parainfluenza viruses, HRSV, and HMPV (3)(4)(5)42). Furthermore, MeV and MuV are incorporated into most national vaccination programs. Collectively, it can be concluded that development of immunity to paramyxo-and pneumoviruses starts shortly after birth and is formed during the first years of life. Here, we show that immunity induced by endemic paramyxoviruses cross-reacts with NiV and may confer protection from highly pathogenic zoonotic paramyxoviruses, like NiV and HeV.
MeV is considered a target for global eradication by the WHO. Although measles eradication would save many lives, it will result in reduced compliance with or even cessation of MeV vaccination. It is already known that measles vaccination or infection induces cross-protection against other morbilliviruses. We have shown here that MeV infection can also lead to the induction of cross-genus-reactive T cells, broadly reactive with most members of the Paramyxoviridae. In a scenario of reduced vaccine compliance or cessation of vaccination in the absence of measles circulation, many children would grow up without morbillivirus-specific immunity, creating a niche for zoonotic morbilliviruses to cross the species barrier. CDV is already capable of infecting a wide range of carnivores (43) and even noncarnivorous species (44,45). The CDV outbreaks in nonhuman primate colonies indicate that CDV and potentially other animal morbilliviruses are more than a theoretical risk for primates lacking morbillivirus immunity (6,9). Combined with the continuous circulation of the lethal NiV, discovery of novel zoonotic paramyxoviruses with the potential to cause lethal disease in humans indicates that a future with loss of morbillivirus-specific immunity calls for better understanding of immunity to paramyxoviruses.
Although NiV infection remains rare in humans, this virus is studied intensely because of its high fatality rate, which reportedly is around 40% for the Malaysia strain and over 90% for the Bangladesh strain (46,47). Because the virus is so lethal, little is known about adaptive immune responses to viral infection. A small recent outbreak in India, in which NiV killed 16 out of 18 infected individuals, provided the opportunity to study adaptive T and B cell responses in two survivors (48). A marked increase in the absolute number of CD8 ϩ T cells was detected, expressing Ki67, granzyme B, and PD-1, a profile of acute effector cells. Although NiV-specific antibodies were also detected, ® viral clearance coincided with the appearance of CD8 ϩ effector cells. This is reminiscent of infection with MeV, where it is already known that virus-specific CD8 ϩ T cells are crucial for viral clearance (21,(23)(24)(25)49). Interestingly, studies with NiV in a nonlethal swine model have also shown that cytotoxic effector cells are crucial in clearance of infection virus (50). Altogether, it appears that although NiV infection is frequently lethal, virus-specific CD8 ϩ T cells contribute to surviving NiV infection. Cross-genusreactive T cells induced by endemic paramyxoviruses could therefore be an important factor in the outcome of an NiV infection.
Here, we describe that the F, N, M, and L proteins are conserved between the members of the Paramyxoviridae and Pneumoviridae. Interestingly, F, H, and N have already been described as major targets for CD8 ϩ T cell responses in acute measles patients (51,52). The novel cross-reactive epitope in F that we present here is located in the fusion peptide, a highly conserved region among paramyxoviruses due to functional constraints (53). The fusion peptide is directly involved in fusion between the viral envelope and target membrane, anchoring into the target cell membrane (54,55). Interestingly, this region was previously reported to be immunogenic, as a CD4 ϩ T cell epitope was identified in dogs immunized with CDV (56). To our knowledge, fusion peptide-specific CD8 ϩ T cells have not been previously described.
The N protein has a crucial role in packaging viral RNA (57), forming a structure known as the ribonucleoprotein (RNP). The RNP is essential in the viral life cycle, as it functions in viral assembly, budding, protecting the genome from innate immune responses, and preventing RNA degradation by host nucleases. The paramyxovirus N protein is divided into two regions, the conserved N-terminal core (N CORE ) and hypervariable C-terminal tail (N TAIL ) (58). The N CORE is important for assembly with the viral RNA, and several studies have shown that the central conserved region of N (CCR, amino acid N 258 -357 ) is mainly responsible (59)(60)(61). It was previously shown that N 321-350 evokes CD4 ϩ responses in measles convalescent-phase donors (62). However, HLA-A*0201-restricted CD8 ϩ T cells to the CCR have also been demonstrated (63). Combining our regions defined as cross-reactive with a previously identified TCC recognizing N 331-339 (LLWSAMGV), we now hypothesize that this TCC is cross-reactive and may recognize many other paramyxo-and pneumoviruses. Similar to F 129 -137 , N 331-339 is present in the conserved F-X4-Y-X3--S--A-M-G motif that is essential in assembly of N with viral RNA and therefore under functional constraints (64). Viral escape to T cell immunity by mutating these regions is highly unlikely, as this results in nonviable viruses.
Repeated exposure to antigenically related members of the Paramyxoviridae and Pneumoviridae in the first years of life potentially provides a selective advantage for cross-reactive T cells. Preferential utilization and boosting of immunological memory based on a previous infection when exposed to a related virus is known as "original antigenic sin." This has previously been described for influenza virus and dengue virus, mostly in the context of serological imprinting (65). Antibody production during an initial exposure to an infection or vaccination dominated the repertoire in subsequent exposures (66,67). This can be a double-edged sword: in the case of influenza, if vaccines are developed that induce memory B cells that make broadly neutralizing antibodies, it is possible that these will persist even as the vaccine recipients age and lose their ability to mount new responses. However, for CD8 ϩ cytotoxic T cells, it was shown that during a secondary infection with a different strain of dengue virus, virus-specific cytotoxic T cells do not kill infected cells but rather release cytokines that exacerbate damage of endothelial cells (68,69). Something similar may apply to the paramyxoviruses. As mentioned before, humans are repeatedly exposed and develop immunity to these viruses in the first years of life. We speculate that it may be crucial in which order certain paramyxo-and/or pneumoviruses are encountered for subsequent cross-protection against other endemic or zoonotic viruses (Fig. 8). Early infection with a certain paramyxovirus (for example, HPIV2 or HPIV3) could commit the immune system to mounting cross-genus-reactive T cells, whereas other infections could lead to a relatively narrow reactivity. Although this is speculative, the possibility of this scenario emphasizes the importance of studying T cell responses to paramyxoviruses.
Our study is limited to several T cells targeting a single epitope, restricted to a specific HLA haplotype. The origin of the cross-reactive CD8 ϩ T cells in this study is known; CD8 Xreact 1 was isolated from a convalescent-phase blood sample obtained from a primary unvaccinated measles patient. CD8 Xreact 2 was isolated from a blood sample obtained from an adult health care worker repeatedly exposed to infants with respiratory tract infections. Whether the induction of cross-reactive T cells by MeV or HPIV infections here is one example of a broader effect requires further study and analysis of T cell specificities in paramyxovirus convalescent-phase donors. Although we show that the TCC identified and characterized here can kill MeV-, CDV-, HPIV3-, and NiV-infected cells in vitro, it remains to be determined whether these clonotypes would be protective when faced with a paramyxovirus infection in vivo.
Highly pathogenic emerging infectious diseases pose a significant threat to human welfare all over the world. The family Paramyxoviridae includes several viruses with the potential to emerge and cause future pandemics, like the henipaviruses. Here, we show that humans could already be at least partially protected against zoonotic henipaviruses by cross-genus-reactive T cell clonotypes induced by previous infections. We argue that it is critical to study T cell responses to both endemic and zoonotic paramyxoviruses and pneumoviruses and that these studies are crucial in the development of novel vaccines. Identification of specific conserved targets that are immunogenic and at the same time under functional restraints, therefore not allowing escape mutations, is the first immunological basis for future universal paramyxovirus vaccines.

MATERIALS AND METHODS
Ethics statement. CD4 MeV 1, CD8 MeV 1, CD4 Xreact 1, and CD8 Xreact 1 were established from PBMC from children 4 weeks after acute measles and were previously published (35)(36)(37)49). The TCC previously described have been renamed in this study; information regarding the designation can be obtained from the corresponding author. Human PBMC used for the generation of novel TCC (CD8 Xreact 2) were obtained from Dutch blood donors in accordance with the Declaration of Helsinki and under the approval of the Dutch ethics committee, and informed consent was obtained (METC, permit MEC-2015-095). Blood samples were collected in heparin Vacutainer tubes. PBMC were isolated from peripheral blood by Ficoll gradient. Three adult male individuals were selected for isolation of novel TCC.
Sequence homology between paramyxoviruses. Paramyxovirus and pneumovirus sequences were selected (see Table S1 in the supplemental material), and protein alignments for N, P, M, F, G/H, and L were prepared using ClustalW multiple alignment in the BioEdit software. Homology percentages of the different proteins were calculated for the endemic and zoonotic paramyxo-and pneumoviruses. Furthermore, stretches of homologous amino acids between MeV and CDV for all proteins were identified, counted, and divided into stretches of Ͻ5, 6 to 10, 11 to 20, or Ͼ20 homologous amino acid residues.
Paramyxo-and pneumovirus phylogenetic tree. An F gene nucleotide alignment was prepared using ClustalW multiple alignment in the BioEdit software, and an unrooted maximum likelihood In this model, we suggest that it is crucial in which order paramyxo-and pneumoviruses are encountered to develop cross-immunity to the highly pathogenic henipaviruses (or other paramyxoviruses). (B) Initially committing T cell immunity to epitopes present in pneumoviruses or MeV could lead to a relative narrow immunity and therefore no protection and death upon encountering a highly pathogenic virus. (C) Encountering HPIV2 or -3 before MeV vaccination or infection could lead to committing to broadly reactive T cell epitopes and therefore (partial) immunity to highly pathogenic viruses.
de Vries et al. ® phylogenetic tree was estimated under the general time-reversible model using PhyML software version 3.0 (GTR ϩ I ϩ G model) (70).
IFN-␥ ELISPOT assays. TCC were screened for (cross) reactivity with paramyxo-and pneumoviruses by IFN-␥ ELISPOT assay as previously described (38,71). Briefly, 5 ϫ 10 3 TCC (responder cells) and 2 ϫ 10 4 autologous, HLA-matched, or HLA-mismatched peptide-pulsed or virus-infected B-LCL (antigenpresenting cells) were cocultured for 1.5 h at 37°C in a 96-well V-bottom plate. Cells were transferred to nylon-membrane-bottom ELISPOT plates (Millipore) coated with a monoclonal antibody (MAb) against IFN-␥ and incubated for an additional 5 h at 37°C. Spots were stained with a secondary MAb against IFN-␥ and counted using an automated ELISPOT reader. For affinity assays, B-LCL were pulsed with a 3-fold dilution series of peptides.
Epitope mapping cross-reactive TCC. The minimal epitope recognized by CD4 Xreact 1 and CD8 Xreact 1 was mapped by performing IFN-␥ ELISPOT assays as described above. HLA-matched B-LCL were initially pulsed with overlapping 15-mer peptides (11 overlap) covering the entire F protein. On the basis of this first assay, a region of interest was defined and the epitope was fine-tuned using overlapping 10-mers (9 overlap). Finally, several 9-mers were tested to determine the minimal epitope.
Urea-based avidity assays. To determine the strength of the binding interaction between representative F 129 -137 tetramers from different paramyxo-and pneumoviruses and HLA-B*1501, we performed urea-based avidity assays. TCC were stained with the different F 129 -137 tetramers (kindly provided by the NIH Tetramer Core Facility) for 15 min at room temperature, washed, and subsequently incubated with increasing concentrations of cold urea (2, 4, 6, and 8 M). After a 15-min incubation period on ice, the amount of tetramer still bound to HLA-B*1501-expressing cells was determined by flow cytometry. The mean fluorescent intensity (MFI) in the PE channel was determined, and ratios compared to 0 M urea were calculated as a representative for the strength of the interaction.
Systematic screen for potential paramyxovirus-specific cross-reactive T cell epitopes. The NiV (NC002728) sequence was selected as a reference virus from the genus Henipavirus, and the six common proteins were aligned to representative paramyxoviruses from the genus Morbillivirus (MeV), genus Respirovirus (HPIV3), and genus Rubulavirus (HPIV2). Conserved regions were defined as stretches of 10 homologous amino acid residues allowing 4 mismatches and subsequently identified in the alignments. Two hundred seventy-one regions of interest were defined, and alignments for all the endemic and zoonotic paramyxo-and pneumoviruses were prepared for these regions. Conservation percentages were calculated when comparing all the individual paramyxo-and pneumoviruses (Data Set S1B, C, D, and E), and a single average conservation percentage was calculated from these. All regions with an average percentage over 50% were selected, defining 27 regions that potentially contain T cell epitopes that are conserved throughout the paramyxo-and pneumovirus proteome. Using NiV as reference, conservation percentages for the 27 regions of interest were calculated against all other endemic and zoonotic paramyxo-and pneumoviruses.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.