Herpesviruses have accompanied humankind since the dawn of evolution. Herpesvirus infections date back at least 6 million years, even before evolutionary split between hominids and chimpanzees (1). From that time on, viral strategies to ensure survival and dissemination coevolved together with the immune system that continuously developed new measures to clear viral infections.

To date, nine different herpesviruses capable of infecting humans have been identified: HSV-1, HSV-2, varicella zoster virus (VZV), Epstein–Barr virus (EBV), human cytomegalovirus (HCMV), HHV-6A, HHV-6B, HHV-7, and Kaposi’s-sarcoma-associated herpesvirus (KSHV) (2).

Despite their different life cycle and growth properties, cellular tropisms and although they cause different diseases, all herpesviruses share common features. They are enveloped and contain a linear, double-stranded DNA genome, ranging from 125 kb (for VZV) to 235 kb (for HCMV) (2). Furthermore, all have the ability not only to infect lytically but also to establish life-long infection in their host, a status called latency, which is a dormant infection lacking pathology and viral replication (2, 3).

Most herpesviruses are widely spread in human populations. Serological tests reveal that HSV-1, VZV, EBV, HHV-6, and HHV-7 have the highest prevalence of the herpesvirus family and infect about 90% of the adult population (4–10). Notably, the prevalence of herpesviruses varies geographically and is influenced by socioeconomic status (2, 11, 12). HCMV prevalence can therefore vary between 50 and 100% dependent on the population studied (13). Some herpesviruses reactivate symptomatically and frequently in healthy individuals for as yet unknown reasons, while others only cause symptomatic reactivation in immunodeficient patients (3). However, research over the past few years revealed that all herpesviruses use common strategies during primary infection, reactivation, and sometimes even during latency, in order to evade the immune surveillance during the different phases of herpesvirus life cycle. The interactions between adaptive immunity and herpesviruses are described elsewhere (14–17). In this review, we will focus on the interaction of herpesviruses and natural-killer group 2, member D (NKG2D)-expressing immune cells. The human-activating receptor NKG2D is expressed on all natural killer (NK) cells as well as on most T cells including γδ T cells and NK T cells (18). Its importance was shown for tumor surveillance (19) and inflammatory diseases (20). The significance of NK cells in herpesvirus immune surveillance becomes clear by looking at NK cell-deficient individuals who suffer from recurrent, severe, potentially life-threatening herpesvirus infections (21, 22).

In the course of human evolution, eight different, functional ligands for the NKG2D receptor emerged: MHC class I polypeptide-related sequence A and B (MICA and MICB, respectively) and the unique long 16 binding protein 1–6 (ULBP1–6) (23). Also known as “stress-induced ligands,” they are barely found on healthy cells in order to avoid auto-reactivity toward normal tissues. These ligands, however, are upregulated and expressed on the cell surface following various stresses including genotoxic stress, oncogene activation or hypoxia that are commonly seen in tumorigenesis, or following viral infection (24, 25).

All NKG2D ligands belong to the MHC class I-like protein family. ULBP family members have an α1/α2 domain structure, whereas the MIC proteins possess an α1/α2/α3 domain structure (26). Interestingly, classical MHC class I proteins serve mainly as inhibitory ligands for NK cells, whereas the NKG2D ligands activate NK cells (27, 28).

Up until now, 16 different allelic variants were identified for the 6 members of the ULBP family (29). More than 100 different MICA alleles and more than 40 MICB alleles were identified to date; a finding that demonstrates the striking superior evolutionary plasticity of the MIC family [http://hla.alleles.org/alleles/classo.html; (30)] (Figure 1). The reason behind this enormous diversity of the MIC family is still unknown. MIC genes lack hypervariable regions; point mutations and genetic shuffles occur over all three domains (31). Comparing amino acid sequence homology, MICA and MICB are very similar (about 85% identity), whereas the similarity to ULBP family proteins is comparatively low (only about 20–25% identity between MIC and ULBP proteins). ULBP family members shares about 60% amino acid sequence identity with each other (32–34). Interestingly, MICA, MICB, ULBP4, and ULBP5 contain a transmembrane domain and a cytoplasmic tail, whereas ULBP1, ULBP2, ULBP3, ULBP6, and one particular allelic variant of MICA (allele *008) are glycosylphosphatidylinositol (GPI) anchored (30).

Only recently, post-transcriptional cellular mechanisms that control stress-induced ligand expression by RNA-binding proteins (35–37) and microRNAs (miRNAs) (38, 39) began to be unraveled; however, the regulatory circuits and expression patterns in normal cells remain incompletely understood. By contrast, much information was gathered about the suppression of NKG2D ligands during herpesvirus infection, emphasizing the importance of the receptor NKG2D for anti-viral immunity.

As emphasized above, herpesvirus family members developed numerous mechanisms to interfere with the expression of the stress-induced ligands. However, most of these studies still leave unanswered questions. More mechanisms and viral effectors are still waiting to be discovered. The viral protein repertoire is probably much larger than currently known; by using ribosome profiling of HCMV and KSHV, numerous new open reading frames (ORFs) have been identified (76, 77). The functions of many viral proteins and ORFs are yet unknown and we are just on the verge of understanding the importance of viral non-coding RNAs, including long non-coding RNAs (78, 79).

While the NKG2D receptor itself is conserved among species, its ligands are not. Interestingly, having eight functional ligands of two different families (MIC and ULBP) and various alleles, the human NKG2D ligand repertoire is more complex than that of other species. Mice possess even nine functional ligands (MULT1, Raet1α–ε, H60a–c) (80). However, their domain structure reveals them to be ULBP family homologs with low allelic diversity. Non-human primates were shown to have homologs of the MIC proteins (81, 82). Still, compared to humans with more than 100 allelic variants, even great apes seem to possess lower allelic variation (83).

Herpesviruses might be a major driving force for diversification of stress-induced ligands and further mutagenesis within alleles leading to allelic variations. None of the described viral mechanisms is capable of eliminating the expression of all stress-induced ligands, the evolutionary pressure rendered these ligands so diverse that no single viral protein or RNA is sufficient to regulate all of them.

As described earlier, viral miRNAs of HCMV, EBV, and KSHV target MICB mRNA at different sites of its 3′-UTR and suppress protein translation (53, 56). Despite the high degree of sequence homology in their 3′-UTRs, MICA is not targeted by any of these miRNAs. The binding sites for the viral miRNAs of HCMV (miR-UL112) and EBV (miR-BART2-5p) are modified by a single-nucleotide insertion, thus abolishing miRNA-induced translation repression. The sequence that is targeted by the KSHV encoded miR-K12-7 is completely absent due to a major deletion in the MICA 3′-UTR.

A similar mutagenesis apparently occurred in the MICA protein to escape UL16 binding. UL16 binds to an α-helical structure in the α2 domain of MICB. By substituting single amino acid residues in MICB with their MICA equivalents, Spreu et al. could show that a single substitution (at two different positions) is sufficient to abolish UL16 binding (84); hence, MICA is spared from UL16-mediated intracellular retention by virtue of very few mutations. Additionally, Klumkrathok et al. suggested that even different allelic variants of MICB are bound with different affinities by UL16 due to amino acid substitutions in the α2 domain (85).

A third piece of evidence for a herpesvirus-driven coevolution is the emergence of MICA*008, a highly prevalent, GPI-anchored MICA variant. MICA*008 is not targeted by UL142, by US18, or by US20. Only recently, it was discovered that the evolutionary relatively novel US9 is capable of targeting solely this distinct allele, but none of the full-length alleles containing a transmembrane domain (68).

These few examples illustrate well the human capability to adjust to viral immune evasion strategies. Accordingly, this strong selective, coevolutionary pressure necessitates modification of viral effector molecules targeting the immune surveillance system as well. By comparing different isolates from HCMV-infected individuals, Renzette et al. and Sijmons et al. indeed showed on a global level that genes involved in immune evasion within the HCMV genome are strongly diversified and contain high numbers of single-nucleotide polymorphisms (86, 87). Among others, one particular mutable gene was found to be UL142, which interacts with NKG2D ligands as pointed out before (87).

It seems obvious to conclude that herpesviruses and ligands for NKG2D continuously shape each other during coevolution, whereas the NKG2D receptor itself remains conserved.

Particularly MICA took the lead in this race on the human side, its 3′-UTR became shortened and modified and numerous allelic variations emerged to withstand herpesvirus infection. The diversity of MICA alleles might thereby even create a “population level resistance” by making it difficult for newly emerging viral mechanisms to successfully target all MICA variants at once.

However, in contradiction to this theory and the supposed importance of stress–ligand evolution, several reports showed a wide distribution of a MICA–MICB null haplotype (also described as MICA-del–MICB-null), a phenotype that occurs mainly, but not exclusively, in East Asia (88–91), apparently with no major evolutionary disadvantage or clinical manifestations. In fact, there are several known MICA-null alleles also independent of this haplotype. If and how MICA and MICB functions are compensated in these individuals, e.g., by the redundancy of the other NKG2D ligands that are still present, has yet to be elucidated; however, this phenomenon teaches us that we are still far from a complete understanding of the complex families of NKG2D ligands.

DS outlined, wrote, referenced the manuscript, and prepared the figure and table. OM supervised and carefully edited the work.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.