In order to increase the reliability of the alignment for the phylogenetic reconstruction, we selected sequences with >85% coverage and >60% similarity with the human PPP1R2 CDS. By doing this, 81 sequences were included in the tree that represented all the pseudogenes with the exception of PPP1R2P7 (Additional file 1: Table S1). The unused sequences encompassed pseudogenic fragments and sequences where identity with PPP1R2 was detected mostly outside the CDS (e.g. PPP1R2P7) or presented truncated CDS (e.g. some PPP1R2P8 and PPP1R2P9).

From the ML tree, four major clusters can be distinguished, generally supported by high bootstrap values (Figure 1). One of the clusters includes most mammalian PPP1R2 sequences, the exceptions being Primates PPP1R2, Glires PPP1R2, PPP1R2-like sequences (rabbit, Oryctolagus cuniculus, Orcu; rat, Rattus norvegicus, Rano; and mouse, Mus musculus, Mumu), and the elephant PPP1R2 (Loxodonta africana, Loaf). The other cluster comprises PPP1R2P8 and PPP1R2P8-like primate sequences. Mammalian PPP1R2P9 sequences compose a third cluster and a fourth cluster includes all PPP1R2 and related pseudogene sequences from Primates (PPP1R2P1/P2/P3/P4/P5/P6/P10). These sequences are clustered with the Glires PPP1R2 sequences. PPP1R2 is also present in the gray short-tailed opossum (Monodelphis domestica, Modo) which is consistent with the presence of PPP1R2 in eukaryotes, being indeed an ancient and well conserved gene 34 .

Two major retroposition events can be inferred, the retroposition that originated PPP1R2P9 and the retroposition that gave origin to PPP1R2P1/P2/P3/P4/P5/P6/P10 (Figure 2). Retroposition of PPP1R2P9 occurred before the split of Eutheria (placental mammals) from Metatheria (marsupial mammals) at ~163.9-167.4 millions of years ago (Mya), as suggested by the presence of this pseudogene in the marsupial gray short-tailed opossum and in all other mammals, making PPP1R2P9 the most ancient pseudogene still present in humans (Figure 2). In the X chromosome we found, close to PPP1R2P9, more PPP1R2-like copies that seem to have arisen by PPP1R2 gene duplication: marmoset (one copy), rat (two copies), mouse (two copies) and pig (one copy). The phylogenetic analysis shows that these copies are related to parental PPP1R2 gene suggesting that this gene has been retroposed to the X chromosome more than once independently and at different time points in these species. We checked for gene conversion events and we did not find any evidence for it. In the phylogenetic tree the PPP1R2P9 genes are clearly apart of these PPP1R2-like that are clustered in the PPP1R2 gene group. PPP1R2P7 is also non-primate specific. Indeed, PPP1R2P7 was present in all mammalian orders included in this study, with the exception of Glires and opossum, suggesting that it was originated ~94.4-163.9 Mya (Figure 2). Retroposition of PPP1R2P1/P2/P3/P4/P5/P6/P10 is more recent and occurred in the ancestor of Primates and during Primates’ evolution since these pseudogenes occur only in primate species (Figure 2). Other retroposition events of PPP1R2 gene have also occurred in some mammals (pig, Sus scrofa, Susc; dog, Canis lupus familiaris, Cafa; giant panda, Ailuropoda melanoleuca, Aime; marmoset, Callithrix jacchus, Caja; and mouse; shown in the tree as R2-like) and appear to be species-specific events since these fragments are not widespread in mammals and both copies present in each species cluster together. Clustering of Glires PPP1R2 and R2-like pseudogenes along with PPP1R2P1/P2/P3/P5/P6/P10/P4 from Primates is consistent with the grouping of these species within the Euarchontoglires (or Supraprimates) superorder 44 . PPP1R2P1 was originated before the separation of New World monkeys (Platyrrhini) and Catarrhini that occurred 43.4-65.2 Mya (Figure 2). A 70 bp deletion seems to have occurred in Hominidae after the divergence from Hylobatidae, ~20.6 Mya. Also, an Alu repeat was inserted after the radiation of the Hominoidea, ~29.4 Mya, in the middle of the sequence disrupting it, but without affecting the open reading frame (ORF) (Figure 3). Interestingly, in chimpanzee, PPP1R2P1 suffered a recent duplication event that gave rise to a second locus separated by two Alu repeats flanking a LINE1 (long interspersed nuclear element, family L1) element (Figure 3, not included in the ML tree). Concerning PPP1R2P3, we found that it clusters along with Primates’ PPP1R2 suggesting that this is the most recent retroposed pseudogene originated after the separation of Hominoidea from Cercopithecoidea (old world monkeys), ~0.6-29.2 Mya, since no copy was found in rhesus monkey and marmoset (Figures 2 and 3). Clustering of PPP1R2P2 and PPP1R2P10/P4 might indicate that these pseudogenes arose by duplication. Our analysis shows that PPP1R2P10 is the ancestral, being originated before the division of Platyrrhini and Catarrhini (42.6-65.2 Mya), while PPP1R2P4 is a duplication that occurred only in humans, being therefore a duplicated pseudogene (Figures 1, 2 and 4). Also, in orangutan, a duplication occurred very close to PPP1R2P10 (~8.8 kb) that is not related with human PPP1R2P4, and was hence here named PPP1R2P10-like (Figures 1 and 4). The other pseudogenes (PPP1R2P5, PPP1R2P6 and PPP1R2P8) were originated at the same time as PPP1R2P10 (Figures 2, 4 and 5). PPP1R2P2 was originated in Catarrhini after its separation from the Platyrrhini ~29.2-42.6 Mya (Figures 2 and 4). The PPP1R2P7 (Glires) and PPP1R2P8 (gibbon) sequences were not retrieved from the databases, which suggest the later deletion of these pseudogenes (Figure 5). The fact that some genome annotations are early assemblies, might explain the missing of these and other sequences. However, the good quality of Glires (Mus, Rattus and Oryctolagus) genome assemblies reinforces the absence of PPP1R2P7 sequence and suggests that it occurred in the common ancestor. The absence of gibbon PPP1R2P8 sequence could also be explained by the several insertions present, similar to what happens in other species, virtually dismantling it and making the retrieval impossible (Figure 5). Moreover, the conserved linkage confirms the results of the phylogenetic analysis, being all pseudogenes flanked by the same respective genes in all species analyzed (Figures 3, 4, 5, 6).

Features such as the existence of transcriptional related data, presence of regulatory elements, mRNA stability (e.g. UTRs, polyA signals), translation initiator sequence and complete ORFs (no truncations or disabling mutations) are indicators of the putative functionality of genes. A search for such features was conducted in order to verify the potential functionality of the PPP1R2 pseudogenes.

Considering the other pseudogenes sequences, many insertions in PPP1R2P8 lead to a completely disrupted ORF and missense mutations in PPP1R2P4/P10 and PPP1R2P6 lead to premature stop codons (Figures 4 and 5). Also, since these pseudogenes showed low coverage to the parental gene, most of the 3’UTR is missing and so, regulatory elements such as polyadenylation signals that are important for the transcript cleavage and stability are absent. This indicates that no transcription or translation should be expected from these pseudogenes, which corroborates with the fact that no expression was found in ESTs and high-throughput databases with the exception of PPP1R2P4. Considering the pseudogenes with the highest coverage in relation to the parental gene, PPP1R2P2 and PPP1R2P5, no ORF disruptions were found but many missense mutations were found in all species analyzed that lead to premature stop codons (Figure 4). All four polyadenylation signals present in the parental PPP1R2 mRNA are conserved in PPP1R2P2. Although protein expression is unlikely, PPP1R2P2 message was found by qPCR in human testis but not in peripheral blood leukocytes 39 . Also, two experiments from ArrayExpress, report the up/down regulation of this pseudogene in prostate adenocarcinoma and in a prostate transcriptomic study performed in a Caucasian population 50 . These results might be artifacts or could be due to other PPP1R2 pseudogenes/parental gene since this pseudogene is located in chromosome 21 that has low density (~232 genes, only surpassed by the Y chromosome with 130 genes), and as expected, the processed pseudogene density is also low, 34 51 , making the transcription highly unlikely.

PPP1R2 forms a stable and high affinity complex with PPP1C by blocking the active site. The reactivation of the complex is triggered by phosphorylation at Thr72 of PPP1R2 through several kinases, including glycogen synthase 3 (GSK3) 52 53 54 . PPP1R2 is also phosphorylated at the residue Ser86 by casein kinase 2 (CK2) that accelerates the subsequent phosphorylation at Thr72 by GSK3 16 . The comparison of human PPP1R2P1, PPP1R2P3 and PPP1R2P9 with PPP1R2 amino acid sequences (Figure 7) shows that PPP1R2P9 is the most divergent (41%) and PPP1R2P3 the most similar (95%). Regarding the PPP1 binding motifs, SILK and KSQKW, they are conserved in all PPP1R2-related proteins, and KLHY is conserved in PPP1R2P3 but a substitution of the first residue to Thr or Arg is observed for PPP1R2P1 or PPP1R2P9, respectively 55 . The C-terminal acidic stretch (DDDEDEE) required for GSK3 phosphorylation 55 56 is maintained in PPP1R2P3 although the GSK3 phosphorylation site Thr73 is substituted to Pro. The other two pseudogenes maintain the GSK3 phosphorylation site but the acidic stretch has several changes particularly in PPP1R2P9 (Figure 7). Finally, the CK2 phosphorylation site Ser87 is conserved in PPP1R2P1 but is substituted by an Arg in PPP1R2P3 and PPP1R2P9. Overall, the analysis shows that these PPP1R2-related proteins should maintain the ability to bind to PPP1C, as was already demonstrated for PPP1R2P3 and PPP1R2P9 40 48 , and the ability to regulate the holoenzyme activity by GSK3 phosphorylation is compromised in PPP1R2P3 48 , and may also be but in a lesser extent in PPP1R2P9, due to the change Ser87 to Arg.

PPP1CC2 is a sperm-specific protein phosphatase involved in spermatogenesis and sperm motility 32 33 57 . Its inhibition in vivo, was associated with a PPP1R2-like activity since GSK3 was able to reverse the process 32 . Recently, a report identified the PPP1R2 protein in heat-stable extracts of bull testis and mouse testis and sperm where it may account for this PPP1R2-like activity 58 . It is well known that testis is one of the organs where most pseudogenes are expressed and their gene products were shown to have important roles in spermatogenesis and other germ cell related functions 52 53 54 . This might be due, in part, to the hyper-transcription state of the autosomal chromosomes in the meiotic and post-meiotic germ cells due to chromatin modifications 13 54 59 . A recent study done by GENCODE has revealed that 64% of all validated expressed pseudogenes are expressed in testis 3 . PPP1R2 is one of the PPP1C regulators with more related pseudogenes 34 . We have previously identified PPP1R2P3 message and protein, in testis 43 45 48 . We hypothesized that from the other pseudogenes, only PPP1R2P1 and PPP1R2P9 are capable of being also translated. In fact, the two pseudogenes, PPP1R2P3 and PPP1R2P9, were present in the mass spectrometry data obtained from a human sperm immunoprecipitation (Table 1).

Peptides were identified by Orbitrap Velos mass spectrometry, from human sperm heat stable extracts and immunoprecipitates using rabbit anti-PPP1R2 antibodies. aa, amino acids; pI, isoelectric point.

This analysis was based on the fact that the molecular weight of these PPP1R2-related proteins should be similar to the parental one (PPP1R2-23.0kDa), being therefore present in the same region where the band was extracted to mass spectrometry analysis. The antibody used to immunoprecipitate PPP1R2-related proteins was raised against a peptide containing amino acid residues 134–147 from the mouse PPP1R2 sequence (Figure 7). This antibody was used previously to detect PPP1R2 48 58 . In the 14-residue region, PPP1R2P1 and PPP1R2P9 have two and three substitutions respectively, when comparing to PPP1R2 sequence. We predicted that using this antibody, we were also able to detect the other PPP1R2-related proteins. Mass spectrometry data identified 23 MSMS (tandem mass spectrometry) spectra corresponding to 8 different peptides matching unequivocally to PPP1R2P9 (Figure 7 and Table 1) and 3 MSMS spectra corresponding to one peptide matching unequivocally to PPP1R2P3 48 .

The sequence coverage obtained for PPP1R2P9 was 36.5% and the mascot score levels were 623.41 (in addition, spectra were manually evaluated). This is the first time that PPP1R2P9 protein is detected, being clearly recovered from human ejaculated sperm. Additionally, these results also indicate that native PPP1R2-related proteins are indeed heat stable and migrate at the same position as the parental PPP1R2. Lastly, no peptides were recovered for PPP1R2P1 using this method, which might suggest the absence of the protein, at least from the sperm cells.

Pseudogenes have been regarded as being derived from functional-encoding genomic DNA sequences that have accumulated disabling mutations (frameshifts and premature stop codons) that make them non-coding protein genes. This lack of function predicts that pseudogenes are not under selective pressures and thus evolve neutrally (reviewed in 1 ). Nevertheless, this view keeps being challenged by the accumulation of examples of transcribed pseudogenes with several acknowledged functions (e.g. regulation of the expression of paralogous genes through the generation of small-interfering RNAs) 1 . Signatures of selection, in addition to sequence conservation, have been considered as obvious indicators of the functional importance of pseudogenes 60 .

Here, by using six ML methods, signatures of both positive and negative selection were detected in the PPP1R2P9 pseudogene, as well as in the parental gene PPP1R2 (Additional file 2: Table S2). Signatures of negative selection were far more evident than those of positive selection, for both genes. Four methods, REL, FEL, SLAC and FUBAR, showed sites negatively selected, with most being detected by more than one method. Signatures of positive selection were principally detected by FEL and MEME methods. The codons 92 and 120, for PPP1R2, and the codons 6, 208 and 211, for PPP1R2P9, were detected by at least two separate methods. No detection was obtained for the PAML method. It is known that sperm-expressed genes present in chromosome X tend to be positively selected when compared with X-linked non-sperm genes and with sperm-expressed autosomal genes 61 62 . This evolutionary pressure is due to their hemizigous expression in males that will favor advantageous mutations and remove any deleterious one. PPP1R2P9 is not evolving neutrally and may thus be expressed, further supporting a functional role for this pseudogene.

The human PPP1R2 mRNA sequence (GenBank accession number NM_006241.4) was used to detect orthologs and pseudogene-related sequences by performing a BLAST search on GenBank, from National Center for Biotechnology Information (NCBI, http://BLAST.ncbi.nlm.nih.gov/) and Ensembl (http://www.ensembl.org/Multi/blastview) databases against all available mammalian reference genomic sequences. Only sequences with more than 60% of sequence similarity and with query coverage of more than 35% were recovered. Genomic sequences flanking the retrieved sequences were also manually inspected for missing parts, especially at the 3’UTR.

The retrieved sequences (Additional file 1: Table S1) were visually inspected and aligned using ClustalW implemented in BioEdit 7.0.9.0 65 . For phylogenetic reconstruction, and to improve accuracy, only sequences encompassing >85% coverage of the human PPP1R2 CDS (nucleotide positions 377–994 of the mRNA sequence) and with >60% of sequence similarity were included in the alignment. In order to determine the phylogenetic relationships between the PPP1R2 gene and related pseudogenes, the best-fit model of nucleotide substitution was first assessed using the program jModelTest v0.1.1 66 under the Akaike Information Criterion (AIC). A maximum likelihood (ML) phylogeny was inferred using the software GARLI v1.0 67 by indicating the best nucleotide substitution model. No starting topology was defined and the program was set to run until no significant topology improvement (as defined by the default settings) was found after 1000000 generations. Five independent runs were performed to check the consistency of the estimates. The support of each node was assessed using 1000 bootstrap replicates. For each bootstrap replicate, the number of generations was set at 100000, above the generation where the last topological improvements were found for each of the five independent replicates. A 50% majority-rule consensus tree of the 1000 bootstrap replicates was created using PAUP* 68 . The support values at each node of the consensus tree were added to the best tree found by GARLI.

Divergence times from the other species in relation to Homo sapiens in millions of years ago (Mya) were obtained from TimeTree (http://www.timetree.org/) 69 .

Sequences obtained from the BLAST queries were analyzed in terms of presence of intronic regions, polyA traits (PolyApred, http://www.imtech.res.in/raghava/polyapred/), truncation of the 5’UTR and chromosomal location. Chromosomal locations were obtained from the GenBank database (Additional file 1: Table S1). Pseudogenes located in the same chromosome and nearby and/or with intronic regions were classified as duplicated pseudogenes. Pseudogenes that were located in different chromosomes and had polyA traits, truncation of the 5’UTR and no introns were classified as processed pseudogenes. Furthermore, genes flanking each human PPP1R2 pseudogene and conserved among mammals were selected. Conserved linkage, meaning conservation of synteny and also conservation of the gene order, was then searched for in order to provide insights regarding their orthology.

The distance of each pseudogene to the closest neighboring gene, not taking into account the presence of nearby pseudogenes, was calculated. Repeated sequences were detected by submitting each pseudogene sequence to the program RepeatMasker from Institute for Systems Biology, Seattle, Washington, USA (http://www.repeatmasker.org/).

Coding sequences evolving neutrally present a ratio (ω) of non-synonymous (dN) over synonymous substitutions (dS) that do not significantly deviate from one. An excess of non-synonymous substitutions over synonymous substitutions (dN > dS) might indicate positive selection, suggesting that the replacement might be advantageous, while negative selection results from the scarcity of non-synonymous substitutions (dN < dS), indicating that a particular mutation most likely is deleterious and is being removed from the gene pool. Pseudogenes are considered to evolve neutrally (reviewed in 1 ).

Maximum-likelihood codon-based tests were used to test for statistically significant signatures of selection in PPP1R2 and related-pseudogenes. Nevertheless, only PPP1R2P9 sequences were analyzed since at least 10 sequences are required to robustly detect signatures of selection 70 . Signatures of positive and negative selection were searched for in Datamonkey webserver (http://www.datamonkey.org) that uses the HyPhy package 71 . The best-fitting nucleotide model (GTR + G) was determined using the automated tool provided by Datamonkey. Five models were used: single likelihood ancestor counting (SLAC), fixed-effect likelihood (FEL), random effect likelihood (REL), fast unbiased bayesian approximation (FUBAR) and mixed effects model of evolution (MEME). SLAC is based on the reconstruction of the ancestral sequences and the counts of dS and dN at each codon position of the phylogeny. FEL estimates the ratio of dN/dS on a site-by-site basis, without assuming an a priori distribution across sites while REL fits a distribution of rates across sites and then infers the substitution rate for individual sites. FUBAR detects selection much faster than the other methods and to leverage Bayesian MCMC to robustly account for parameter estimation errors. Finally, MEME is capable of identifying instances of both episodic and pervasive positive selection at the level of an individual site. Sites with P values <0.1 for SLAC, FEL and MEME, posterior probability of >0.9 for FUBAR, and Bayes Factor >50 for REL were considered as being under selection. CODEML (PAML version 4, 72 ) was also used to detect positive selection by comparing a null model and a model that allows positive selection (M1 vs. M2 and M7 vs. M8). The contrasting models were compared by computing twice the difference in the natural logs of the likelihoods (2ΔlnL). In the site-specific models that allow the ratio ω to vary among codons, we performed Likelihood Ratio Tests (LRTs) with 2 degrees of freedom to compare the following models (NS sites): M1 (nearly neutral evolution ω0 = 0, ω1 = 1) with M2 (neutral and positive selection: ω 0 = 0, ω 1 = 1, ω 2 > 1) and M7 (beta-distributed negative selection: 0 # ω # 1) with M8 (beta-distributed negative selection and positive selection: 0 # ω1 # 1, ω2 >1) 2 73 . Only amino acids identified in M8 by using the Bayes Empirical Bayes (BEB) approach and with posterior probability >95% were considered as evolving under positive selection. For the initial working topology, ML trees were constructed using MEGA5 74 with substitution nucleotide models determined by the software: TN93 + I and partial deletion (95% cut-off) for PPP1R2P9 and K2 + G with G = 4 and partial deletion (95% cut-off) for PPP1R2.

Since testis is one of the organs where most pseudogenes are expressed 75 and spermatozoa are the final product of spermatogenesis, the presence of some of the studied pseudogenes was tested in human sperm. Ejaculated sperm was collected from healthy donors by masturbation into an appropriate sterile container. Spermograms were performed by experienced technicians and only samples with normal parameters were used 76 . Informed consents were signed allowing samples to be used for scientific purposes. The study was conducted in accordance with the guidelines of the “Helsinki Declaration”. In brief, sperm was lysed in 1 × RIPA buffer (radioimmunoprecipitation buffer, Millipore Iberica S.A.U., Madrid, Spain) supplemented with protease inhibitors (10 mM benzamidine, 1.5 μM aprotinin, 5 μM pepstatin A, 2 μM leupeptin, 1 mM PMSF), sonicated 3 × 10 sec and centrifuged at 16000 g for 20 min, at 4°C. RIPA supernatant sperm extract was immunoprecipitated using Dynabeads® Protein G (Life Technologies S.A., Madrid, Spain) and 1 μg of rabbit anti-PPP1R2 (against a mouse PPP1R2 peptide, amino acids 134–147) with standard direct immunoprecipitation procedure 48 . Also, an independent RIPA supernatant sperm extract was prepared, boiled in a water bath for 30 min, chilled on ice for 2 min and centrifuged at 16000 g for 20 min, 4°C to obtain a heat stable extract.

For mass spectrometry analysis, the immunoprecipitate and the heat stable extract were resolved by 10% SDS-PAGE along with purified positive controls. Gels were stained with Coomassie blue colloidal (Sigma-Aldrich Química, S.A., Sintra, Portugal) using standard procedures 48 . Bands were then excised from the gel using commercial PPP1R2 band as control and destained. An overnight digestion with trypsin (Promega, Madison, Wisconsin, USA) was performed and resulting peptides were extracted and prepared for mass spectrometry analysis using an Orbitrap Velos mass spectrometer as described elsewhere 48 . Subsequent generated data were imported to ProteinScapeTM (Bruker Daltonik GmbH, Bremen, Germany, 77 ) and analyzed using MASCOT (version 2.2.0, Matrix Science, London, UK, 78 ) search algorithm. Proteins were considered to be identified if the Mascot score (ProteinScapeTM) was higher than 65.