Absence of CD59 in Guinea Pigs: Analysis of the Cavia porcellus Genome Suggests the Evolution of a CD59 Pseudogene

The complement system plays an essential role in innate and adaptive immunity in all known vertebrate animals (1, 2). Complement immunity is initiated through Ag–Ab interactions (i.e., the classical pathway) (3), the activation and amplification of C3 in the presence of activating surfaces (i.e., alternative pathway or amplification loop) (4), or the recognition of sugar moieties (i.e., lectin pathway) (5), all ultimately leading to the cleavage and activation of C3 (6). This, in turn, initiates the formation of the membrane attack complex (MAC), which is composed of components C5b–C9 (7). Once formed, the MAC causes perforation and lysis of target cell membranes.

In vertebrates, complement regulators (CRegs) have been well characterized (8). CRegs are soluble or membrane-bound proteins that are responsible for the protection of host cells from complement-mediated damage (9, 10). One of the most ubiquitously expressed CRegs is CD59, an 18–20–kDa GPI-anchored membrane-bound protein that is expressed on almost all cell types in mammals and that prevents MAC assembly by inhibition of the interaction of the C5b-8 complex with C9 (11). Although initially characterized in humans, CD59 has since been identified across vertebrate species, including mammals, birds, reptiles, amphibians, and bony fish (8, 12–14). Structural conservation of this molecule across animal phyla, as well is its expression across all types, suggests a high degree of importance for CD59 in protecting the host from complement-mediated damage.

Early work using recombinantly expressed or native human CD59 suggested that the MAC-inhibitory activity was species specific (15, 16). However, subsequent work comparing the activities of human, rat, pig, and sheep CD59 analogs against different species sources of complement found a high degree of cross-species activity (17). Notable in this latter study was that guinea pig erythrocytes were uniquely sensitive to MAC-induced lysis, regardless of the species source of the MAC proteins, including homologous MAC proteins, and were uniquely protected from complement-mediated lysis by exogenously added CD59 from the various species. These observations provoked the suggestion that guinea pig erythrocytes lacked an endogenous CD59 analog (17).

Although this unique susceptibility of guinea pig erythrocytes to complement-mediated hemolysis has been known for several decades, there is still no definitive explanation as to why this is the case. Components of the guinea pig complement system have been purified and characterized from as early as the 1960s (18, 19), with all components of the activation pathways and terminal pathway characterized (20, 21). Guinea pig complement efficiently lyses classical targets, such as Ab-sensitized sheep erythrocytes, demonstrating that there is not an underlying complement deficiency. Other CRegs, including decay accelerating factor (CD55) and membrane cofactor protein (CD46), have been identified in and cloned from guinea pig (22, 23), but CD59 remains unidentified and an enigma.

In this article, we describe a combined immunochemical, phylogenetic, and genomic approach to elucidate whether guinea pigs express CD59. Our data further strengthen the contention that guinea pigs do not express CD59. We show that the gene encoding for CD59 in the Cavia porcellus genome has become a pseudogene. Together, the data demonstrate that there is no functional gene encoding CD59 in guinea pigs; these findings provide an explanation for the unique hemolytic properties attributed to guinea pig erythrocytes.

The expression of CD59 on guinea pig erythrocytes was assessed using six Abs: two rabbit polyclonal Abs generated against mouse CD59a and known to be cross-reactive across species, including human CD59, and two mAbs specific for human CD59 and mouse CD59a. Erythrocytes from normal mice (C57/Bl), CD59a-deficient mice, and humans were used as positive and negative controls. Erythrocytes from each source were washed three times in isotonic PBS, with centrifugation at 2000 × g for 10 min to pellet after each wash, and then resuspended to 0.1% (v/v) in flow cytometry buffer (FCB; PBS [pH 7.4] supplemented with 1% BSA [w/v]). Aliquots (100 μl) were incubated with 1 μg of the appropriate anti-CD59 Ab at 4°C for 1 h with the appropriate fluorescein (FITC)-conjugated secondary; donkey anti-mouse IgG (715-096-151; Jackson ImmunoResearch), goat anti-rabbit IgG (L320-NC51Z; Oxford Biotechnology), or rabbit anti-rat IgG (F-1763; Sigma) and washed twice at the recommended dilution in FCB. Stained erythrocytes were washed as above and analyzed by flow cytometry using a Becton Dickinson FACSCalibur. Erythrocytes from each species yielded a single population when analyzed for forward and side scatter; for each, the entire population was gated for analysis of fluorescent staining.

Human CD59 gene (Ensembl transcript ID ENST00000351554.7) localized to chromosome 11 (bases 33,708,865–33,736,445, translated from the negative strand) was used as the basis for syntenic and codon-based alignments. The five exons from this region were used for alignment with the C. porcellus genome from the Ensembl annotation of the Broad Institute’s CavPor3 assembly, version 90 (24).

Because the human CD59 gene and other mammalian CD59 genes are flanked by genes FBXO3 and C11orf91, alignments were performed to a chromosomal region of the C. porcellus genome flanked by their respective gene homologs. This region, characterized by Ensembl as Scaffold DS562947.1 (bases 5,004,217–5,025,988), was used for alignment using individual human CD59 exons.

All five exons of human CD59 (Ensembl accession numbers ENSE00001326619, ENSE00001364419, ENSE00000824399, ENSE00000710249, ENSE00002187503) were aligned with the Ensembl annotated CavPor3 genome assembly, focusing on Scaffold DS562947.1 as described in the previous section. Each alignment was performed using the default parameters of the genomic alignment, changing only the default start/stop sequences to match the precise beginning and end of each exon (as indicated in Table I). Sequence similarity was calculated using the ClustalW-aligned sequences and the PRABI suite of the Pole Bioinformatique Lyonnais Web site (https://prabi.ibcp.fr) (25).

To determine the evolutionary relationship of the C. porcellus CD59 pseudogene with other vertebrate CD59 genes, a phylogenetic tree was constructed using the putative cDNA encoding CD59 from 43 species (Supplemental Table I). All cDNA sequences were aligned using ClustalW, and the phylogenetic tree was generated using the Neighbor-Joining method in the MEGA6 program suite (26), with bootstrap values n = 1000.

Another phylogenetic tree was constructed, based on the taxonomic organization of each of the species described in Supplemental Table I. In this case, the taxonomic tree was generated using the PhyloT (http://phylot.biobyte.de) web-based phylogenetic tree generator (27) and the taxonomic National Center for Biotechnology Information (NCBI) identifiers for each species.

To determine whether any transcriptional products were derived from the identified CD59 pseudogene, alignments were performed using RNA sequencing (RNA-Seq) data, acquired from the Broad Institute’s CavPor3 assembly (GenBank assembly no. GCF_000151735.1). The reads were analyzed using NCBI’s Genome Data Viewer, using the default aggregate settings of the C. porcellus annotation release 102. This setting included reads from brain, liver, lung, skeletal muscle, kidney, and cervix; tracks displaying RNA-Seq exon coverage were used as a measure of expression, compared with other previously characterized guinea pig CRegs: CD46, CD55, and complement receptor–related protein (Crrp; Crry homolog).

Erythrocytes from guinea pig, normal mouse, CD59a-deficient mouse, and human were stained with monoclonal and polyclonal Abs against mouse CD59a or human CD59. Human erythrocytes were strongly positive with the monoclonal anti-human CD59 MEM43 and moderately strongly positive with two polyclonal Abs raised against mouse CD59a (Fig. 1C), and mouse erythrocytes stained positively with the rat monoclonal anti-mouse CD59a and with the two polyclonal anti-CD59a Abs (Fig. 1D). CD59a-deficient mouse erythrocytes used as a specificity control were negative for all Abs tested (Fig. 1B). Guinea pig erythrocytes were also negative for all Abs tested (Fig. 1A).

Initial attempts to identify guinea pig CD59 cDNA used Basic Local Alignment Search Tool searches against the C. porcellus genome, using default parameters derived from mammalian CD59 cDNAs and amino acid sequences. The only homologies identified were members of the Ly6 protein family, leukocyte surface protein lacking the conserved residues, and motifs that characterize vertebrate CD59 (28, 29).

The second strategy to locate a putative CD59 gene in the C. porcellus genome involved syntenic alignment with the region of human chromosome 11 encoding the CD59 gene. The CD59 gene is located between the FBXO3 and C11orf91 genes in multiple mammalian genomes. Individual human CD59 exons were aligned on a chromosomal scaffold of C. porcellus, limited to a region between the FBXO3 and C11orf91 homologs. The results are shown in Table I. No DNA sequence homologous with human CD59 exon 1 was found, but homologs of human CD59 exons 2–5 were located within the 27.3-kb region, annotated by Ensembl as Scaffold DS562947.1, in the C. porcellus genome. Individual alignments of human exons 2, 3, and 4 showed nucleotide identities of 55.8, 62.3, and 57.8%, respectively. The homolog of human exon 5 showed only 25.8% identity, primarily because the exon 5 homolog was truncated compared with its human counterpart. A schematic diagram of this alignment is shown in Fig. 2.

To determine the degree of conservation at the protein level between human CD59 and a potential product of the identified guinea pig exon homologs, nucleotide alignments were performed between the open reading frame (ORF)-encoding regions of human CD59 (exons 3, 4, and 5) and the exon homologs identified in the C. porcellus genome. As seen in Fig. 3, the protein product of human exon 3, encoding the majority of the N-terminal leader peptide, shared a high degree of amino acid identity and similarity to the predicted product of the C. porcellus homolog, without any frameshifts (i.e., the sequence translated in a +1 ORF). In particular, the leucine-rich motifs and phenylalanine and cysteine residues were conserved in the predicted C. porcellus CD59 leader sequence.

Conservation at the (predicted) protein level was also observed when aligning human CD59 exon 4 and its C. porcellus counterpart without any frameshift (i.e., +1 ORF). In the human sequence, exon 4 encodes the last 3 aa of the leader peptide and the first 30 aa of the mature protein (starting LQCY); these 30 aa include five cysteine residues essential for protein stability, four of which are conserved in the predicted C. porcellus sequence. The first cysteine residue is displaced into the predicted leader sequence in the guinea pig.

Divergence between the two genomes was most pronounced when comparing the protein product of human CD59 exon 5 with the predicted C. porcellus exon 5 product. There was some sequence conservation through the first 24 aa (residues 31–54 in the mature human protein), with conservation of the two cysteine residues in this segment but not the tryptophan residue (W40 in mature human CD59), which is known to be a key functional residue in multiple species CD59 analogs; indeed, in the selected ORF, this (W40 in human) codon is replaced by a stop codon.

Taken together, these results demonstrate that the genomic sequence identified as the sole homolog of CD59 in the C. porcellus genome is a pseudogene incapable of producing a functional CD59 protein.

To determine whether the identified C. porcellus CD59 pseudogene shares a phylogenetic relationship with other vertebrate CD59 homologs, a phylogenetic tree was generated from CD59 cDNA sequences from diverse species (Fig. 4; Supplemental Table I), using a neighbor-joining method with bootstrap confidence intervals of n = 1000. These results were compared with a pruned phylogenetic tree, based on NCBI taxonomy (data not shown); the trees displayed a high degree of similarity between clades, with CD59 conservation distributed along phylogenetic classes (i.e., more similar CD59 sequences clustering along avian, reptilian, fish, and mammalian orders). Furthermore, within mammals, CD59 clusters were distributed consistently along taxonomical order (i.e., rodents, primates, and carnivores). Based on these results, a pairwise alignment was performed using the available cDNA sequence from the naked mole rat (Heterocephalus glaber), the species most closely related to C. porcellus in the analysis (Fig. 5).

A pairwise alignment between the C. porcellus CD59 pseudogene and the protein-encoding CD59 cDNA of its most taxonomically related species, the naked mole rat, demonstrated a high degree of conservation throughout the sequence. However, a 15-bp gap and two 3-bp gaps were observed in the C. porcellus CD59 pseudogene, all consistent with keeping the sequence in-frame.

To test whether the C. porcellus CD59 pseudogene was transcribed, the predicted sequences from all four characterized exon domains were aligned with RNA-Seq data acquired from GenBank, based on the Broad Institute’s CavPor3 assembly, and included reads from brain, liver, lung, skeletal muscle, kidney, and cervix. Expression was visualized using the NCBI Genome Browser. This analysis was also performed with other known C. porcellus CRegs: CD46, CD55, and Crrp. As seen in Fig. 6, the three characterized CRegs had significant RNA expression throughout the gene, with signal intensities reaching nearly 17,000 for CD59, 530 for CD46, and 300 for Crrp. For each of the CD59 pseudogene exons, there was essentially no signal for expression, with the exception of a weakly defined 80-bp region (corresponding to Ensembl scaffold DS562947.1: 5065147–5065227) that reached a signal intensity ∼ 10; this segment corresponds to an intronic area between the regions homologous to exons 4 and 5 of human CD59.

The propensity for guinea pig erythrocytes to be lysed by a variety of animal sera, as well as to be effectively protected from lysis by incorporation of CD59 from other species, led us to propose that guinea pig erythrocytes lacked a functional CD59 protein, rendering them susceptible to lysis by homologous or heterologous complement (17). Although we had no data on other tissues and cell types, we further speculated that guinea pigs might have a genetic lack of CD59, an anomaly given the wide species distribution of CD59 analogs.

In this study, we set out to address this anomaly and provide evidence of the fate of CD59. First, we stained guinea pig erythrocytes with a panel of anti-mouse CD59a and anti-human CD59 Abs, including two polyclonal Abs raised against mouse CD59a that showed broad species cross-reactivity, staining mouse and human erythrocytes; specificity was demonstrated by showing that CD59a-deficient mouse erythrocytes were negative for all Abs. Guinea pig erythrocytes showed no staining, supporting the contention that they lacked a CD59 analog.

We then conducted an extensive genomic study using an annotated C. porcellus genome. Basic Local Alignment Search Tool alignments using other mammalian CD59 genes identified only limited homology to genes of the Ly6 superfamily of molecules; no gene possessing the conserved domains of CD59 was found in the C. porcellus genome (28, 29). Syntenic alignments were then performed, looking for the C. porcellus genomic region corresponding to the region of human chromosome 11 containing the CD59 gene. In mammalian species, the CD59 gene is flanked by the FBXO3 and C11orf91 genes. The C. porcellus FBXO3 and C11orf91 homologs were identified; however, no intact CD59 gene was located between these two genes. Individual alignments of the five human CD59 exons across this region identified regions of homology with exons 2–5. Exons 2–4 share a high degree of sequence similarity between human and guinea pig with putative in-frame ORFs. In contrast, alignments with human exon 5, which codes for the majority of the mature protein sequence, identified putative regions of homology consisting of two adjacent truncated regions (together, ∼60% the size of its human counterpart) that did not contain a complete ORF but contained non-sense mutations and deletions. In humans, this region encodes for a cysteine–asparagine combination that appears to be highly conserved across vertebrate Ly6 family proteins (30). The loss of this highly conserved region and the presence of non-sense and deletion mutations support the assertion that this segment of chromosomal DNA is no longer under functional constraint, but instead represents an inactive pseudogene prone to genetic drift.

RNA-Seq analysis confirmed that no portion of any of the CD59 exon homologs was transcribed. Therefore, from an evolutionary perspective, it is likely that elements associated with the promoter have been lost. The loss of this region, along with the loss of a human exon 1–homologous region, would suggest that a deletion event at the 5′ end of the gene may have contributed to the inactivation of a previously functional gene. In contrast, expression of other C. porcellus CReg proteins was readily detectable from RNA-Seq, indicating that other CRegs were still functional and that these components are likely sufficient to prevent MAC-associated autoimmune damage.

The absence of a region homologous to exon 1 of human CD59 in the guinea pig genome may imply high genetic drift, eliminating any significant similarity to human CD59 exon 1, or a deletion event removing the exon 1–like region. The second possibility is of particular interest, because it may imply an ancestral region prone to genetic recombination. This might help to explain gene-duplication events associated with the CD59 gene; CD59 gene duplication has been described in mice (31), and a putative CD59 gene duplication event is reported in Chinese softshell turtle (Ensembl access number ENSPSIG00000007164). In both cases, the gene-duplication products are adjacent to each other while still being syntenic to the FBXO3 gene. In the mouse, the two CD59 genes are differentially expressed, with CD59a broadly distributed in all tissues and CD59b expressed only in male genital tissue (31). The CD59 gene-duplication product in Chinese softshell turtle lacks the conserved regions associated with a functional CD59. Nevertheless, these two instances of gene duplication, combined with the putative gene deletion event in guinea pigs, suggest a dynamic genetic element that may exist in CD59 genes across all vertebrates.

The fact that intact and functional CD59 genes have been found in all other available mammalian genomes suggests that loss of CD59 expression was caused by a mutation unique to guinea pigs (or an ancestral species). From an evolutionary standpoint, the lack of a functional CD59 in guinea pigs adds to the debate on the importance of this molecule in protecting vertebrates from autoimmune damage. CD59 is reported to be ubiquitously expressed on (almost) all cells of mammals (32) and is considered to be indispensable for immune homeostasis. Studies using CD59a-knockout mice demonstrate that CD59 deficiency is compatible with life; the mice showed increased erythrocyte turnover and a compensated anemia and increased pathology in a number of complement-mediated disease models (33–36). CD59 deficiency in humans has serious consequences; isolated CD59 deficiency in humans was first described >25 y ago in a Japanese patient presenting in his 20s with symptoms resembling paroxysmal nocturnal hemoglobinuria (37). More recently, CD59 deficiency was reported in several families of North African Jewish ethnicity; all cases presented in infancy with chronic hemolysis and episodes of peripheral neuropathy resembling Guillain-Barré syndrome (38). All cases were homozygous for a missense mutation in the CD59 gene, pCys89Tyr, and had no detectable CD59 on cell surfaces. Disease course was severe with high mortality. Other cases with remarkably similar clinical features, but different CD59 mutations, were subsequently described (39, 40); importantly, treatment with the anti-C5 Ab eculizumab markedly improved hemolytic and neurologic symptoms.

The relatively mild phenotype in CD59a-knockout mice has been ascribed to the presence in rodents of another widely expressed CReg, Crry; this was shown to be essential for murine erythrocyte protection from complement attack, whereas CD59 was dispensable (41). No Crry homolog has been reported in guinea pigs, although a truncated homologous protein, Crrp, has been described (42). The gene encoding for Crrp aligns with murine Crry and may serve a similar function, reducing the impact of the absence of CD59 in guinea pigs. Overall, the redundancy of membrane-bound CRegs in rodents likely compensates for the lack of functional CD59 in guinea pigs. Nevertheless, from an evolutionary perspective, our results represent the first example, to our knowledge, of a species in which the CD59 gene has become a pseudogene and demonstrate the capacity of a rodent species to regulate complement activation on self-cells independently of this molecule.

We thank Dr. Ana M. Aransay (Center for Cooperative Research in Biosciences bioGUNE) for providing technical advice on the preparation of this manuscript.

The authors have no financial conflicts of interest.