Introduction

NLRP proteins (Nod-like receptors with a pyrin domain) have attracted recent attention because of their role in innate immunity and inflammation (Kufer and Sansonetti 2011; Strowig et al. 2012). A subset of NLRP genes are expressed in mammalian oocytes (Hamatami et al. 2004; Ponsuksili et al. 2006; Zhang et al. 2008) and maternal deficiency of some of these “reproduction-related” NLRPs (rNLRPs) have been shown to cause embryonic lethality in mice (Hamatami et al. 2004; Tong et al. 2000; Peng et al. 2012) and perturbations of genomic imprinting in human oocytes (Murdoch et al. 2006; Kou et al. 2008; Parry et al. 2011).

Tian et al. (2009) undertook a phylogenetic analysis of the NLRP genes from humans, chimpanzees, rats, mice, cattle, and dogs. These species had similar sets of NLRP genes, with the rNLRP genes forming a monophyletic group. Therefore, the major radiation of NLRP genes was already present in the most recent common ancestor of Boreoeutheria. These authors observed that rNLRPs were more evolutionarily labile, in both copy number and sequence, than other NLRPs.

Evidence that rNLRPs might play a role in the establishment or maintenance of genomic imprinting motivated us to re-investigate the evolutionary history and possible functions of rNLRPs. We extend the analysis of Tian et al. (2009) using NLRP sequences from an afrotherian (African elephant Loxodonta africana), marsupial (gray short-tailed opossum Monodelphis domestica), and monotreme (platypus Ornithorhynchus anatinus). These sequences allow inferences about the NLRP repertoire of the last common ancestor of eutherian mammals, the last common ancestor of marsupials and eutherians, and the last common ancestor of all extant mammals. Moreover, advances in knowledge of the effects of rNLRPs allows us to speculate about the role of these genes in genomic imprinting and reproductive disorders of mice and humans.

Results

Mammalian NLRP Repertoires

NLRP genes occur at eight locations in the human genome: 11p15.5 (NLRP6); 11p15.4 (NLRP10); 11p15.4 (NLRP14, 1 Mb from NLRP10); 19q13.42 (NLRP12); 19q13.42 (NLRP2, NLRP7); 19q13.43 (NLRP4, 5, 8, 9, 11, 13); 1q44 (NLRP3); and 17p13.2 (NLRP1). NLRP genes were found in the elephant genome at seven of these locations (based on conserved flanking markers). The exception was elephant NLRP6 which occurs on an unassembled fragment without flanking markers. The elephant genome contains an additional NLRP gene between MID2 and VSIG1 that we have provisionally named NLRPX (currently annotated as NLRP12-like in elephant, macaque, and marmoset). An NLRP pseudogene is located between these markers on the human X chromosome (NLRP3P). NLRPX-related sequences are sister to a clade containing NLRP10 genes in our phylogenetic tree.

Four NLRP genes are currently annotated in the opossum genome. NLRP10 is located on opossum chromosome 4 between GVIN1 (corresponding to the GVINP1 pseudogene at human 11p15.4) and CSMD2 (human ortholog at 1p34). Two other opossum NLRPs also map to chromosome 4: one appears orthologous to NLRP12 whereas the other (provisionally named NLRPa, currently annotated as NLRP12) appears in our phylogeny as sister to the rNLRP clade albeit with weak support. The fourth gene (provisionally named NLRP5-like, currently annotated as NLRP14-like) occurs next to EPN1 on an unassembled fragment. Human EPN1 neighbors the cluster of six rNLRPs at 19q13.43.

Five NLRP genes were found in the platypus genome. Four appear to be orthologs of human NLRP3 (currently annotated as NLRP12-like), NLRP6, NLRP10 and NLRP12. The fifth (provisionally named \(\psi\) NLRP10, currently annotated as NLRP3-like) occurs on the same fragment as NLRP10 and is possibly a monotreme-specific duplicate of that gene. None of the platypus genes group with the rNLRPs.

Figure 2 summarizes the NLRP repertoires of the mammalian species used in this study. We represent the data as a table of species against NLRP genes with phylogenies depicted as a reference for both. The cells of the table are shaded according to the presence or absence of a given NLRP gene in a particular species.

Phylogenetic Tree Root

Our tree places NLRP6 as sister to the other eutherian NLRP genes using chicken “NLRP3” to root the tree, similar to the analysis of Tian et al. (2009) who used the same root. A sister relation between NLRP6 and other mammalian NLRPs has been found repeatedly in phylogenetic analyses. These include analyses of NACHT (Hughes 2006; Laing et al. 2008) and LRR domains (Ng et al. 2011) that rooted the NLRP clade using non-NLRP proteins. A slightly different root was proposed in a phylogenetic analysis of PYD domains that placed the root between a clade containing NLRP6 and NLRP10 and the other NLRP genes (Kersse et al. 2011).

Diversification of NLRP2 and NLRP7

Figure 3 presents a Bayesian phylogenetic tree for current genes and pseudogenes labeled NLRP2 or NLRP7 in the databases as of May, 2012. The primate-specific duplication noted by Tian et al. (2009) is confirmed but independent duplications have also occurred in pigs, cattle, horses, and elephants. We suggest, as an interim measure, that use of NLRP7 be restricted to genes of that name in the primate clade.

Syntenic Analyses

Online Resource 3 shows the probable orthologies of all NLRP genes from humans against the other species of mammals, as well as chicken. These figures use the latest genome assemblies (as of May, 2012), which is why they differ slightly from the data-set used for the phylogenetic tree in Fig. 1.

Discussion

Evolutionary History of NLRP Genes

The similarity of the NLRP repertoires of humans and elephants shows that the major radiation of NLRP genes, including rNLRP genes, had already occurred in the common ancestor of Afrotheria and Boreoeutheria.

Distal chromosome 4 of opossum contains genes with orthologs on human chromosomes 11p and 19q. Thus, the data are compatible with a scenario in which the opossum genome maintains ancestral linkage of NLRP genes that have been dispersed onto the equivalents of human 11p and 19q in eutherian mammals. Unfortunately, the evolutionary inference is weak because opossum, platypus, and chicken genomes are only partially assembled for these regions.

The existence of NLRPa and NLRP5-like in opossum suggests that rNLRP genes evolved before the last common ancestor of Metatherian and Eutherian mammals. Our phylogenetic analysis (Fig. 1) confirms that opossum NLRP5-like belongs to the rNLRP clade. Whether it is expressed in opossum oocytes is unknown. Conservation of synteny suggests NLRP5-like is probably located on opossum chromosome 4 with the other NLRP genes. Four out of four opossum NLRP genes plausibly map to a region of chromosome 4 with orthologs on human chromosomes 11p and 19q (where 12 out of 14 human NLRP genes are located).

Platypus NLRP3 is located adjacent to RPS5 and SLC27A5, genes whose orthologs are located near the telomere of human 19q (close to the major cluster of rNLRPs), whereas human NLRP3 is located near the telomere of chromosome 1q as part of a small, recent addition to an otherwise ancient linkage group (Haig 2005). Therefore, NLRP3 may have been linked to rNLRP genes in the most recent common ancestor of monotremes and therian mammals. This leaves NLRP1 at human 17p13.2 as the only eutherian NLRP whose ortholog cannot be provisionally assigned to this ancestral linkage group.

However, a scenario that is consistent with our data is that rNLRPs were absent in the common ancestor of Monotremes and Metatherians. Then, a single rNLRP gene evolved in a common ancestor of Metatherians and Eutherians. This ancestral rNLRP is orthologous to possum NLRPa and it later diversified to give rise to the clusters NLRP2/7 and NLRP9/11/4/13/8/5, as well as NLRP14. NLRP2/7 may be still present in possum, but the current assembly does not contain them or their flanking markers (see Online Resource 3).

Lineage-Specific Duplications and Deletions

Our analysis provides evidence of lineage-specific duplications and deletions of NLRP genes in eutherian mammals. NLRP11 was previously proposed to be restricted to primates and NLRP4 to be restricted to Euarchontoglires (Tian et al. 2009) but we find orthologs of NLRP11 in pig and of NLRP4 in pig and horse (see Fig. 2 and Online Resource 4). Pig NLRP11 represents an interesting example, because were it absent, parsimony would indicate a single loss of NLRP11 in Laurasiatheria. Its presence indicates independent losses in mouse, dog, horse, and cattle.

The data also suggests an independent loss of NLRP8 and NLRP13 in marmoset, mouse, and elephant. In particular, the marmoset genome has NLRP11 and NLRP4 alone in an unplaced fragment, while conservation of synteny suggests these genes should be placed between NLRP9 and NLRP5. A future assembly, however, might reconstruct this area and place the fragment inside it, potentially also reconstructing NLRP8 and NLRP13.

A second phylogenetic reconstruction without any of the reconstructed pseudogenes is available in Online Resource 5. There are three minor differences between the phylogeny in Fig. 1 (which includes pseudogenes) and Online Resource 5. First, the reversal of the divergence of the NLRP8-13 and NLRP5-14 gene pairs. Second, the NLRP10-X gene pair changing from being a monophyletic group with NLRP3 and NLRP12, to being its outgroup. Third and finally, possum NLRP5-like clustering with NLRP8 instead of NLRP5.

Maternal-Effect Lethality in Mice

Inactivation of Nlrp5 (also known as Mater) in mouse mothers causes arrested development of embryos at the two-cell stage whether or not an embryo inherits a functional copy of Nlrp5 from its father (Tong et al. 2000). Nlrp5 protein is associated with the cytoplasmic lattice of mouse oocytes and appears to be essential for the formation and/or stability of the lattice (Kim et al. 2010). Mitochondria of Nlrp5-deficient oocytes are scattered throughout the cytoplasm, rather than concentrated in the subcortical layer (Fernandes et al. 2012). Knockdown of Nlrp2 or Nlrp14 mRNA in mouse oocytes similarly causes arrested development of embryos during early cleavage (Hamatami et al. 2004; Peng et al. 2012).

Nlrp5 protein associates with Ecat1/Filia protein in mouse oocytes and early embryos (Ohsugi et al. 2008; Zheng and Dean 2009). Mutations in Filia cause maternal-effect embryonic lethality with apparent defects in the assembly of mitotic spindles (Zheng and Dean 2009). Filia belongs to a family of genes expressed in oocytes and early embryos (Pierre et al. 2007) that includes C6orf221 (see below).

Perturbations of Genomic Imprinting in Humans

During normal embryonic development, the maternal allele is methylated and paternal allele unmethylated at most imprinting control regions (ICRs). The H19 ICR is an exception to this generalization, with an unmethylated maternal allele and methylated paternal allele (Reik and Walter 2001; Schulz et al. 2010). DNA in the sperm pronucleus of recently fertilized mammalian oocytes undergoes conversion of most 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC) whereas 5mC is protected from this activity in the egg pronucleus (Iqbal et al. 2011; Wossidlo et al. 2011). Paternally derived chromosomes then progressively lose 5hmC during early embryonic development (Inoue and Zhang 2011). Taken together these results suggest that most ICRs are methylated in the maternal germ line with paternal ICRs unmethylated because of demethylation of the sperm pronucleus after fertilization, with the methylated paternal H19 ICR somehow protected from this process.

Women homozygous for mutations of NLRP7 exhibit maternal-effect embryonic lethality in the form of the repeated conception of biparental complete hydatidiform moles (biCHMs) (Murdoch et al. 2006; Kou et al. 2008; Parry et al. 2011). These “embryos” exhibit biparental non-methylation of most imprinting control regions (ICRs), with the notable exception of the H19 ICR (Kou et al. 2008; El-Maarri et al. 2003). Women with mutations of both alleles of C6orf221 also produce biCHMs with loss of differential methylation at most ICRs except H19 (Parry et al. 2011; Judson et al. 2002). Thus, loss-of-function of NLRP7 and C6orf221 cause similar perturbations of imprinting. C6orf221 belongs to the Ecat/Filia gene family whose members are expressed in oocytes and early embryos and are known to interact with NLRP proteins (Zheng and Dean 2009).

The methylation pattern of biCHMs would be explained if maternal ICRs become demethylated in embryos because of a failure to protect the maternal pronucleus from conversion of 5mC to 5hmC. One possibility is that maternal deficiency of NLRP7 or C6orf221 results in a disruption of spatial relations in the fertilized oocyte causes the egg pronucleus to be subject to the same epigenetic modifications as the sperm pronucleus.

Human NLRP7 is the product of a recent duplication of NLRP2. Maternal homozygosity for a frame-shift mutation in NLRP2 has been associated with loss of “maternal” methylation at 11p15.5 and a clinical diagnosis of Beckwith–Wiedemann syndrome (BWS) in offspring (Meyer et al. 2009). Finally, two individuals with phenotypic features of BWS have been reported with independent breakpoints in ZNF215 (ZKSCAN11), a gene immediately adjacent to NLRP14 (Alders et al. 2000).

Conclusions

Our analysis shows that all the major NLRP genes are present in eutherian mammals and that rNLRP-related genes are present in marsupials. The rNLRPs have a complex history of independent duplications in several eutherian lineages. While strong inferences about the origin of rNLRP genes remain elusive, the following scenario is consistent with the data. The appearance and diversification of rNLRP genes in the common ancestor of Metatherians (where opossum chromosome 4 maintains ancestral linkage of NLRP genes) and Eutherians, followed by breakage of linkage and consequent dispersal to their present locations in eutherian mammals.

Mutations in NLRP2 and NLRP7 in humans have been associated with disorders of genomic imprinting, Beckwith–Wiedemann syndrome and biparental complete hydatidiform moles. A plausible explanation is that deficiencies of NLRP2 and NLRP7 result in the maternal pronucleus being subject to the programmed demethylation that is normally restricted to the paternal pronucleus. Because human NLRP2 and NLRP7 are the product of a primate-specific duplication, mice are perhaps not the best model organisms to study these reproductive disorders.

Methods

Data-Set and Syntenic Analyses

A total of 116 amino-acid sequences for NLRP proteins were obtained from GenBank (Benson et al. 2011) and used in the phylogenetic analyses. We used a combination of genes annotated as NLRP or NLRP-like, identified as protein coding or as pseudogenes. Online Resource 2 contains the accession numbers of the sequences used in our analyses.

We focused on genes from human (Homo sapiens), macaque (Macaca mulatta), marmoset (Callithrix jacchus), cattle (Bos taurus), pig (Sus scrofa), dog (Canis lupus familiaris), horse (Equus caballus), mouse (Mus musculus), elephant (Loxodonta africana), opossum (Monodelphis domestica), platypus (Ornithorhyunchus anatinus), and chicken (Gallus gallus). Blastp (Altschul et al. 1997) was used with the 14 human NLRPs as queries to the non-redundant protein sequences (NCBI: nr) to assure no other NLRP proteins were undetected from these taxa. The sequence data used in the main phylogeny (Fig. 1) was collected on or before September 2011, using the current genome assemblies at the time. While new assemblies for dog, cattle, and pig are available at the time of this writing, there are no new assemblies for elephant, possum, or platypus. Therefore, we chose not to redo our analyses, for the new data do not pertain the initial diversification of NLRP genes.

A total of 112 protein sequences were available for NLRP genes from the species listed above and identified as protein coding. A gene annotated as NLRP1-like in chicken [GenBank:XP_422818.3] and one of the two NLRP5s annotated in pig [GenBank:NP_001156879.1] were outliers in our preliminary alignments and phylogenetic trees and were excluded from further analyses. The NLRP1-like gene of chicken is on chromosome 9 in a region of conserved synteny with human chromosome 3 from which no NLRP or NLRP-like has been reported. The results of blastp using chicken NLRP1-like as query shows that the most likely ortholog is an uncharacterized protein with IFT80 and IL12A as flanking markers in mammalian species. The characterization of this protein as NLRP1-like is likely due to the presence of FIIND and DEATH domains, which are also present in mammalian NLRP1 genes.

Six “pseudogenes” were included in our analysis: two from elephant, and one each from macaque, marmoset, opossum, and platypus. Pseudogene sequences were transformed into amino acid sequences by identifying exon boundaries and removing frame shifts. These reconstructed “proteins” aligned well with proteins encoded by orthologous NLRP genes.

Blastn was performed to look for other pseudo-genes or gene traces in the genomes of possum and platypus, but no significant hits were found. Additionally, orthologous and paralogous relationships were established by inspecting flanking markers of all NLRP genes in the genome assemblies for all species used. We considered enough evidence for the absence of an NLRP gene/pseudogene when neither Blast nor syntenic analyses could find suitable candidates (see Online Resource 3).

When a gene was suspected absent, we performed a tblastn search with the corresponding human NLRP gene(s) as query. A cut-off E-value of \(10^{-10}\) was used throughout. The only non-NLRP hits in non-human mammals were CARD4, CARD15, CIITA, NOD1, NOD2, NOD4, NLRC3, NLRC5, NLRX1, and RNH1. All of these hits are due to the presence of shared domains with NLRP genes. Furthermore, all these non-human hits were found in synteny with the corresponding human genes and their flanking genes.

Sequence Alignment and Phylogenetic Analyses

We used UGENE as our working platform (Okonechnikov et al. 2012). Multiple sequence alignments were performed with T-Coffee (Notredame et al. 2000) and are available upon request. The phylogenetic trees were reconstructed with Maximum Likelihood (ML) methods using Garli (Zwickl 2006) and Bayesian estimation (BA) using MrBayes (Ronquist and Huelsenbeck 2003; Huelsenbeck and Ronquist 2001).

For the Bayesian reconstructions we used 2 runs, each with 4 chains and a burn-in fraction of 0.25. The runs were carried for an initial one million iterations, with increments of one million iterations until convergence. We declared a run to converge when the differences in the Log-Likelihood were less than 0.1 % for the last million iterations and the Potential Scale Reduction Factor convergence criterion was within 1 % of 1. For the Maximum Likelihood reconstruction we used 2 search repetitions running for 5 million generations with 100 bootstrap iterations from an initially random tree. The JTT+G+F model of amino-acid evolution using the observed aminoacid frequencies (Jones et al. 1992) was selected based on the best Log-Likelihood score provided by ProtTest (Darriba et al. 2011).

The consensus tree (Online Resource 1) was constructed by merging the results from ML and BA using DendroPy (Sukumaran and Holder 2010). We chose to present only the Bayesian tree in Fig. 1 because it is more resolved.

In contrast to Tian et al. (2009), we included all amino-acids available from all sequences into the analysis, which included several partial and low-quality sequences. It is generally accepted that using partial sequences diminishes the accuracy of the inferred phylogeny. This detrimental effect is, however, still not well understood, with researchers suggesting detrimental effects ranging from insignificant to severe Wiens (2003); Hartmann and Vision (2008); Burleigh et al. (2009); Kück et al. (2010). Several approaches have been suggested to increase phylogenetic inference accuracy, ranging from masking (i.e., removal) of problematic data to the statistical simulation of missing data Hartmann and Vision (2008); Kück et al. (2010). However, Bayesian and, to a lesser extent, Maximum Likelihood approaches seem to be particularly resilient to missing data, with consistently high accuracy being achieved even in the absence of up to 95 % of some of the samples (Wiens and Moen 2008; Wiens and Morrill 2011). In our case, only 11 out of the 116 sequences (less than 10%) are partial (NLRP6 of elephant and platypus, NLRPb of possum, NLRP3 and NLRP11 of marmoset) or low quality (NLRP5 and NLRP10 of elephant and horse, respectively, NLRP1 of elephant, and NLRP11 of pig). Also, the missing data of partial sequences is presumably small and is therefore unlikely it has a strong effect on accuracy. In particular, the shortest partial sequence corresponds to the NLRP6 of platypus which has a length of 417 amino-acids. The longest NLRP6 is in macaques and has 1045 amino-acids, while the longest NLRP sequence overall is NLRP1, also in macaques, which has 1475 amino-acids. Therefore, we can conservatively estimate less than 30 % missing data in less than 10 % of the sequences.

In Online Resource 5, we performed a phylogenetic recontructions of the NLRP genes in Fig. 1 leaving out all reconstructed pseudogene sequences.

Fig. 1
figure 1

Consensus phylogenetic tree of the NLRP gene family. The phylogenetic tree was inferred using MrBayes for 10,000,000 time steps. We used 115 amino-acid sequences of NLRP genes and “pseudogenes” from 11 mammalian species (human, macaque, marmoset, mouse, pig, cattle, horse, dog, elephant, possum, and platypus) as well as one sequence from chicken as an outgroup

Full size image
Fig. 2
figure 2

Presence and absence of NLRP genes in mammalian species. This table summarizes the known information about the distribution of NLRP genes across the mammalian species included in our analyses. The cells of the table are shaded according to the presence or absence of a given NLRP gene in a particular species. Light gray is used to indicate the presence of a single NLRP gene identified in the databases as “protein coding.” Light gray with a wave pattern is used if multiple copies of the gene are present, either identified as “protein coding” or “pseudogene.” Dark gray indicates that a single gene is present and identified as “pseudogene.” Black is used to denote the probable absence of the gene; in general, we consider a gene absent if it is not currently annotated in the species, absent from BLAST searches, and syntenic examination on the other mammalian species. We use white to denote that there is not sufficient evidence based on flanking markers to consider a gene absent

Full size image
Fig. 3
figure 3

Consensus phylogenetic tree of NLRP2 and NLRP7. The phylogenetic tree was inferred using MrBayes for 1,000,000 time steps. We used 19 amino-acid sequences of all available NLRP genes and “pseudogenes” denoted as NLRP2 or NLRP7 in the mammalian species used in our other analyses, as well as human NLRP14 as an outgroup. Potentially independent duplications of NLRP2/7 can be seen in Laurasiatheria, primates, and elephant

Full size image