QSOX sequences suffer from poor annotation
Many QSOX sequences available in the public databases are currently un- or mis-annotated. Due to the presence of a domain with homology to PDI proteins, the most common mis-annotation is the identification of the QSOX as a PDI. Although both QSOX and PDI act as catalysts of disulfide bond formation, they differ in their cellular localization, domain composition, and electron acceptor. Whereas PDI family proteins are defined by localization to the endoplasmic reticulum and the presence of one or more Trx domains [11], QSOX is defined by the presence of both a Trx domain and an Erv domain [7]. Hence, other annotations such as “Sulfhydryl oxidase like” or “Thioredoxin like,” are similarly insufficient. Another misleading annotation is the numbering of QSOX paralogs. In several instances, the numbering of a QSOX paralog in a given species does not correspond to its counterpart in a second, closely related species. We therefore suggest a coherent and phylogeny-based annotation and numbering for all QSOX sequences identified and used throughout this article. Accession numbers and corresponding annotations are provided as (Additional file 1: Table S1).
QSOX sequences are widespread, but not universal, in Eukaryota
QSOX sequences were previously identified within a range of eukaryotic organisms, including protists, plants, and animals, but not in fungi [7, 12] (Figure 2, Additional file 2: Figure S1). As genomic data continue to accumulate, more QSOX sequences are available for comparative and evolutionary analysis (Figure 3). Of particular interest are organisms that fall at the root of major eukaryotic taxa and enable a better understanding of the primary events leading to the diversity of contemporary QSOX enzymes (Additional file 2: Figure S1).
Stramenopiles, Euglenozoa, and Alveolata are three early branches of unicellular eukaryotes for which QSOX sequences were identified. Within Stramenopiles, QSOX sequences were identified in diatoms. QSOX sequences also appear in Oomycetes, which include several genera of plant pathogens: Phytophthora, Hyaloperonospora, and Albugo. In Euglenozoa, QSOX sequences were identified for Leishmania and Trypanosoma species, and among Alveolata for Perkinsus, Plasmodium, and Cryptosporidium.
In plants (Viridiplantae) QSOX is found in both green algae (Chlorophyta) and green plants (Streptophyta). Among green algae, QSOX sequences were identified for Chlorophyceae, Mamiellophyceae, and Trebouxiophyceae. Among Streptophyta, QSOX sequences were identified in mosses (Embryophyta) and vascular plants (Tracheophyta). A QSOX sequence was identified for S. moellendorffii, member of an ancient lineage at the root of vascular plants in the evolutionary tree. Among the vascular plants, QSOX sequences were found for both monocotyledons and eudicotyledons.
The earliest evidence of a QSOX sequence in Opisthokonta is of the marine choanoflagellate M. brevicollis. However, due to a gap in the data affecting the Trx domains, the M. brevicollis QSOX sequence could not be retrieved in its entirety, and the sequence was therefore excluded from subsequent analyses. Within Metazoa, a QSOX sequence was identified for the placozoan T. adhaerens, which has the smallest animal genome known to date. Among Eumetazoa, QSOX sequences were identified for species of Ctenophora, Cnidaria, and Bilateria (within Acoelomata in Trematoda, within Pseudocoelomata in Nematoda, and within Coelomata). Many QSOX sequences were identified for Arthropoda (both within Chelicerata and Mandibulata) and Chordata, in primitive small invertebrates such as Tunicata and the fish-like cephalochordate B. floridae, as well as in Craniata, from fish to humans.
In total, 228 QSOX sequences were identified within the genomes of 132 different species.
Local QSOX duplications in Viridiplantae contrast with a deep duplication in Craniata
Many organisms contain more than one QSOX gene (Additional file 2: Figure S2). In some species, such duplication events seem to be evolutionarily recent, whereas in others, the duplication occurred deep in the origin of the phylum. Two QSOX paralogs can be found in Craniata from fish to humans (Figure 3), with the exception of several species listed below. The branching at the origin of Craniata means that the human QSOX1 paralog is more similar to fish QSOX1 than to human QSOX2. Within the genomes of A. carolinensis, O. cuniculus, M. putorius, and S. scrofa, only the QSOX1 variant was found; in O. anatinus only the QSOX2 variant was found. The complementary QSOX paralogs may not have been identified due to poor coverage of the respective genomes.
In addition to the two QSOX paralogs, QSOX sequences have also diversified in Craniata through alternative splicing. In many Craniata, from rodents to humans, an alternative splice variant of QSOX1 has been identified (Additional file 2: Figure S2). This splice variant interrupts the final exon and eliminates the transmembrane segment but preserves all redox-active domains. The splice variation may thus affect processing or localization of the QSOX enzyme. Except for a single occurrence in M. musculus, splice variants of the QSOX2 paralog of Craniata have not been identified.
For most arthropods, no paralogs were identified. However, significant exceptions to this generalization are the coleopteran T. castaneum and Drosophila, with up to four paralogs in several Drosophila species. The QSOX paralogs of T. castaneum do not share the same duplication event as those of Drosophila. In Drosophila, a major duplication event occurred at the root of the genus, giving rise to the QSOX1 paralog present in all Drosophila and to a second branch of QSOX paralogs that underwent several further duplication events with no clear consistency between the species.
Among Nematoda, several duplication events seem to have occurred before the branching of Chromadorea and Enoplea, followed by inconsistent gene duplication and gene loss events.
Among Viridiplantae, QSOX paralogs were identified only within Streptophyta. The duplication events in plants are recent and seem to have occurred locally at the species or genus level. Accumulation of additional QSOX sequences might reveal slightly deeper branching between closely related plants. Nevertheless, as the QSOX paralogs in the two available Arabidopsis species genomes and those of G. max do not seem to have emerged from the same duplication event, the branching is not expected to be much deeper.
Among Alveolata, QSOX paralogs were identified for the P. marinus species. Due to a lack of genome sequences for closely related species, it is not known how deep this duplication event might be.
Among Stramenopiles, a duplication event may have occurred deep within Oomycetes, followed by multiple inconsistent gene duplication and gene loss events. However, it may be that several relevant paralogs were not identified due to incomplete genome coverage. Another duplication event among Stramenopiles seems to have occurred deep at the level of Bacillariophyta.
This extended catalog of QSOX sequences provides the platform for the following phylogenetic and comparative analyses.
A single evolutionary path within Viridiplantae contrasts with multiple branching within Metazoa
Several branches seem to emerge from the root of the QSOX phylogenetic tree and evolve apart (Figure 3). Although all neighbor-joining (NJ) and maximum likelihood (ML) trees branch according to the corresponding taxonomic affiliations of the examined species, low consensus scores are obtained at the roots of the major branches. It is only in more recent branches that the distinctions between QSOXs of different species are well supported. According to the consensus ML tree (Figure 3), the Viridiplantae cluster, beginning with the Mamiellophyceae branch, is well distinguished from other QSOX sequences. The distinction between QSOXs of monocotyledons and eudicotyledons is also well supported. Among Metazoa, somewhat reliable scores distinguish QSOXs of Arthropoda and Nematoda after the branching of the D. pulex and T. spiralis species, respectively. The distinction between QSOX1 and QSOX2 of Craniata is the first to be well supported within Chordata. Lastly, among protists, the Alveolata branch seems to cluster among the Stramenopile branch together with Oomycetes and apart from Bacillariophyta. From all of the above, it is clear that Viridiplantae evolved apart from the various Metazoan branches, although the relative positioning of the branches is poorly supported and hence does not reveal deeper affiliations. Removal of protists from the ML trees slightly improved the distinction between Viridiplantae and several of the Metazoan branches (Additional file 2: Figure S3). Overall, QSOX sequences of Viridiplantae display a continuous path from Chlorophyta to the higher plants in Streptophyta, with a few recent duplication events, whereas highly divergent forms of Metazoa QSOX sequences emerged from a very ancient ancestor.
To assess the evolutionary paths of the QSOX Trx and Erv domains independently, separate ML phylogenetic trees were constructed (Additional file 2: Figures S4 and Figure S5). Similarly to the original ML tree, the ML tree for each domain supports claims for branching order mainly for recent taxonomic affiliations. Nevertheless, important observations can be made upon closer examination. For example, the Trx ML tree shows notable differences from the Erv ML tree along the plant lineage. The Trx ML tree shows the branching of Viridiplantae following Mamiellophyceae, but the clear distinction between monocotyledons and eudicotyledons is missing. Conversely, the distinction between monocotyledons and eudicotyledons is strongly supported by the Erv ML tree, but this tree fails to show the early distinction of plants from other phyla. In other words, the ancient distinction of plant QSOX Trx domains, evident following Mamiellophyceae, was apparently followed by diversification of the Erv domain, particularly notable at the bifurcation of Streptophyta into monocotyledons and eudicotyledons. Therefore, it appears that the Trx and Erv domains of plant QSOXs were subjected to distinct selection events at different points in evolution.
Some differences are also observed between the Trx and Erv domain trees within Metazoa. For example, the Trx ML tree shows support for a distinction between the Craniata QSOX1 and QSOX2 variants well within Mammalia. In the Erv ML tree, however, the diversification of the QSOX2 variant is already well supported at the root of Craniata. QSOX1, in turn, is distinguished in the Erv ML tree from other animal QSOXs at the level of Mammalia. Therefore, it seems that animal QSOX variants underwent a significant evolutionary event that co-affected both their Trx and Erv domains prior to or during the emergence of mammals.
QSOX redox-active motifs differ between Metazoa and Viridiplantae and show enhanced diversity among paralogs
Differences in the redox-active CXXC motifs of the Trx and Erv domains of various organisms were described in a previous review on QSOX [12]. The identities of the intervening residues between the cysteines are expected to play a major role in modulating the redox-potential of the site [13] and may also contribute to interaction specificity. Here we extend the analysis of QSOX CXXC patterns by providing a phylogenetic perspective. Indeed, when we present CXXC motif sequences in the context of the QSOX phylogenetic tree, the motifs can be seen to define distinct groups of QSOX enzymes (Figure 4), and sequence patterns develop along the main branches.
The Trx-CXXC motif shows strong sequence conservation and characterizes major groups of organisms. Specifically, the CGHC sequence is nearly universal in Euglenozoa and Metazoa, with the exception of certain nematodes and arthropods of the Coleoptera, Diptera, and Hymenoptera lineages. Interestingly, among Drosophila, a QSOX paralog with a CGHC pattern coexists with an array of paralogs displaying a CG [DN] C sequence in their Trx-CXXC motifs. Similarly, in nematodes, QSOXs with a CGHC patterns are found together with a paralog exhibiting a CGAC pattern. To the extent possible, we assigned the designation QSOX1 to paralogs containing the CGHC pattern dominant in Metazoa QSOXs. Among Viridiplantae and the Apicomplexa branch of alveolates, the CPAC pattern is the most abundant, whereas Stramenopiles typically have a CPHC pattern. It is interesting to note that a CPXC pattern is rare among PDI family Trx domains, for which the CGHC pattern is prominent, even in PDIs from plants [14]. Hence, the distinction between the three major patterns (CGHC, CPHC, and CPAC) for Trx redox-active motifs in QSOX enzymes is independent of trends for CXXC motifs in PDI family proteins.
The Erv-CXXC patterns are more diverse than the Trx-CXXC patterns. For example, five organisms at the base of Metazoa, T. adhaerens, N. vectensis, H. magnipapillata, S. purpuratus, and B. floridae, all share the common Trx-CXXC sequence CGHC but have the Erv-CXXC sequences CQKC, CSYC, CRYC, CQNC, and CQEC, respectively. In Craniata, variation in the Erv-CXXC motif follows the taxonomic lineage. Specifically, both QSOX1 and QSOX2 of lower taxa often contain a CREC pattern, whereas the patterns diverge in higher taxa such that CRDC becomes dominant for QSOX1 and CKEC for QSOX2. In general, Erv-CXXC patterns delimit narrower groups of organisms than do Trx-CXXC patterns and vary in most instances of paralogs. Even in some recent duplications in which paralogs retain high sequence identity, the Erv-CXXC motifs differ. For example, the Erv-CXXC motifs have diverged following the independent duplications in Arabidopsis (CEEC vs. CEDC) and Trichocarpa (CDDC vs. CDEC).
Consideration of the various QSOX CXXC patterns in light of the phylogenetic tree allowed us to distinguish between diversity that implies the absence of constraints from diversity that suggests adaptation. In particular, the strong conservation of Trx-CXXC patterns within different lineages suggests tight constraints along these branches, despite the prominent differences in patterns between lineages. A phylogenetic perspective also aids in discrimination between primary and auxiliary paralogs. The primary paralog is present in closely related species and tends to exhibit the same CXXC patterns, whereas auxiliary paralogs greatly increase the sequence diversity of the motif. Finally, a phylogenetic approach allows speculation regarding the primordial patterns that diversified into those exhibited by contemporary organisms. For example, the occurrence of CPAC in the Trx-CXXC of plants may have arisen from a primordial CPHC pattern, still exhibited by Stramenopiles and even in two occurrences of Chlorophyta.
Intron positions do not reveal a common imprint between Viridiplantae and Metazoa
The differing numbers of Trx-fold domains in Metazoa vs. other QSOXs, the absence of QSOX from fungi, the different motif sequences, and the ambiguities at the roots of major branches in QSOX phylogenetic trees prompted us to consider whether contemporary QSOX enzymes may be descended from more than a single fusion event linking a Trx-fold to an Erv domain. To search for evidence to the contrary, we inspected QSOX intron positions through various taxa (Figure 5). It has been previously observed that H. sapiens share intron positions with A. thaliana in certain orthologous genes. Even if the ancestral sequence endured several intron gains and losses, a primitive imprint could still be distinguished [15]. Inspection of QSOX intron positions revealed diversification between phyla, with apparently no common intron positions among Viridiplantae and Metazoa. Some closely situated or ambiguous positions will be discussed for clarification. In one case, an intron marks the end of the α2 helix of the Trx domain among many Metazoa, whereas an intron occurs at the beginning of the helix close to the Trx-CXXC active site in Viridiplantae. The following intron in Viridiplantae is at the beginning of the α3 helix, whereas in Metazoa the intron is at the end of the β4 strand. In the ψErv domain, an intron is located before the α1 helix in both Viridiplantae and Metazoa, but the lack of the Trx2 domain as well as any sequence conservation makes it difficult to conclude from this one coincidence that plant and animal QSOX have a common origin. Further on, two introns are located before and after the α2 helix in Viridiplantae, whereas in Metazoa the intron is located within the α2 helix. In summary, no residual signature of a common evolutionary origin between plant and animal QSOX was detected in the pattern of intron positions. This result does not imply that QSOX enzymes from different kingdoms lack a common ancestor, only that evidence for a common ancestor is lacking in the intron patterns.
The ψErv/Erv module, strongly characteristic of QSOX, contrasts with a Trx module only weakly differentiated from PDI family domains
We then sought more insight into the potential origins of the Trx and Erv domains to assess the likelihood of various scenarios that may have produced the current QSOX family. A comparison of the QSOX Trx and Erv domains with other proteins containing only one or the other of these domains demonstrated that the QSOX ψErv/Erv module is clearly distinct from all other known Erv enzymes. The most widespread member of the Erv family is Erv1, a mitochondrial enzyme found broadly throughout eukaryotes, including yeast. Erv1 is a symmetric dimer, in which the two copies of the Erv domain interact. This arrangement differs significantly from QSOX, in which the Erv domain forms a pseudodimer with the ψErv domain and is therefore inaccessible for true dimerization. The presence of the ψErv region immediately upstream of the Erv domain in QSOX, as well as the existence of a third, carboxy-terminal CXXC motif (CT-CXXC), are unique characteristics of the QSOX subfamily of Erv enzymes that distinguish them definitively from other Erv domains. Remarkably, the CT-CXXC motif, despite being dispensable for enzyme assays in vitro[8], is as highly retained as the two catalytic CXXC motifs of QSOX. In summary, the available data suggest that the ψErv/Erv modules in all contemporary QSOX enzymes share a common ancestor.
In contrast to the clear distinction between the ψErv/Erv module of QSOX enzymes and Erv domains of other proteins, the distinction between Trx domains of QSOXs and PDI family proteins is less evident. The strong structural similarity among Trx domains from PDI proteins, and between PDI and QSOX, makes the identification of defining features of Trx domains in one protein family vs. the other quite challenging. Unlike the extension containing the CT-CXXC motif of QSOX, which is a distinguishing characteristic of its Erv domain, the CX6-8C disulfide in the QSOX Trx1 domain is found also in PDI proteins. No secondary structure element, loop, or other structural feature appears to differentiate QSOX Trx domains from PDI Trx domains. Furthermore, the large and ancient PDI family provides a rich source for both Trx and juxtaposed Trx1/Trx2 domains from which the Trx modules in plant and animal QSOXs may have been derived. Focusing on amino acid sequences rather than structural features, the CGHC and CPAC patterns, found in the Trx domains of major QSOX lineages, are observed in various PDI proteins as well. Interestingly, the CGHC pattern is strongly dominant among PDI proteins that contain non-catalytic Trx domains downstream of a catalytic one, just as the CGHC pattern is characteristic of QSOXs that contain a catalytic/non-catalytic arrangement of Trx domains. In turn, CPXC patterns are frequently observed in PDI family proteins that lack non-catalytic Trx domains, analogous to plant and protist QSOXs.
Despite the similarities between QSOX and PDI Trx modules, in all phylogenetic trees constructed to include both sets, QSOX Trx domains clustered separately, although with very low scores at the root (Additional file 2: Figure S6). Further inspection of QSOX sequences compared to PDIs revealed an amino acid position that differs consistently between the two protein families. Two residues upstream of the CXXC motif in the Trx domain, a proline that is found nearly universally in PDI proteins is replaced by, most often, a serine or threonine in metazoan QSOXs or a histidine in plant QSOXs (Additional file 2: Figure S7). A proline is found only in select QSOX enzymes, such as those from O. tauri, M. pusilla, and P. infestans. The phylogenetic trees and the pinpointing of particular residues suggests that subtle sequence differences do distinguish QSOX and PDI Trx domains, consistent with either divergent evolution of QSOX Trx domains from a single precursor or convergence of distinct QSOX Trx domain sequences due to shared pressure to function in conjunction with the ψErv/Erv module.
In either scenario, the question remains whether the fusion of Trx and Erv domains has a common purpose across all organisms that display this fusion. In particular, is the fusion indicative of a shared biochemical mechanism? The availability of QSOX crystal structures [9] allowed for an analysis of QSOX sequences against the backdrop of the three-dimensional structures. The ensuing analysis of conservation at the level of QSOX domains was conducted to shed more light on the evolution of fundamental structural and mechanistic elements of the QSOX enzyme.
Conservation at the Trx-Erv domain interface suggests a conserved electron transfer mechanism
As described above (Figure 1), QSOX enzymes are characterized by the fusion of a Trx domain to an Erv domain bridged by non-catalytic structural elements. The two catalytic domains show the strongest sequence similarity. Excluding core secondary structure elements such as the central Trx1 β-strands, six regions of particularly notable similarity are detected in QSOX, as highlighted in a sequence logo built from all available QSOXs (Figure 6A). Comparison of these regions with the QSOX structure allows the function of each to be assessed and emphasizes the importance of the interaction between domains.
The first two segments of conservation correspond to the Trx1 redox-active site and the CX6-8C disulfide-bonded loop adjacent to it in the tertiary structure. Together, these two regions constitute the interface between the Trx1 and the Erv domains when they are packed against one another in the electron-transfer intermediate [9]. To better appreciate the conservation at the contact site, conservation scores were mapped onto the QSOX structure (Figure 6B). Apart from the Trx-CXXC motif, the most conserved residue in these two segments is a proline at the beginning of the β4 strand. This proline, present in many Trx-fold proteins and found structurally in the cis configuration, was previously suggested to contribute to substrate binding [16] or shown to facilitate substrate release [17] or inhibit metal binding by the reactive thiolate-based active site of thioredoxins [18]. Following the cis-proline, there is poor conservation at the sequence level in QSOX, though at the structural level the domain is predicted to retain an α4 helix.
The third region of conservation (Figure 6C) is another disulfide bonded loop at the amino-terminal junction of the Erv domain (Cys449-Cys452 in the H. sapiens QSOX1 sequence). Within the eleven residues between the cysteines in this loop is a basic residue (Arg401 in the H. sapiens QSOX1 sequence) that makes an electrostatic interaction to the phosphates in the FAD cofactor. The remainder of the loop projects out from the Erv helical bundle to contribute, together with the sixth region of conservation detailed below, to a structural element that has been described as a “backstop,” since it seems to constrain the Trx1 domain sterically as it docks against the Erv domain during interdomain electron transfer. The presence of this backstop is one of the major differences between QSOX Erv domains and other Erv enzymes. This difference may reflect the structural contexts of the di-cysteine motifs that interact with the Erv-CXXC in QSOX compared to other Erv enzymes. Specifically, the QSOX Trx domain, from the opposite end of the multi-domain protein, interacts with the QSOX Erv domain [9], whereas di-cysteine motifs on flexible regions of polypeptide tethered locally interact with the FAD-proximal disulfide of stand-alone Erv domains [19, 20]. The sequence conservation in QSOX extends from the backstop loop through the first three turns of the α1 helix in the Erv domain, where a tryptophan and histidine project from the same side of the α1 helix to form part of the FAD binding site.
The fourth conserved region in QSOX comprises the Erv redox-active di-cysteine motif. In addition, a highly conserved histidine and phenylalanine follow the CXXC motif in the α3 helix of the Erv domain. The imidazole side chain of the histidine is surface-exposed and interacts with the polar edge of the FAD isoalloxazine. The phenylalanine, together with the α1 helix tryptophan mentioned above, contributes to the hydrophobic pocket occupied by the non-polar edge of the isoalloxazine.
The fifth conserved region is the α4 helix in the Erv-domain helical bundle, which contains a set of side chains that contacts the adenine portion of the FAD. Specifically, a histidine and two asparagines are hallmarks of Erv domains [21]. The α4 histidine interacts with the conserved histidine from the α1 helix, and the two imidazole rings, in a planar arrangement, stack against the FAD adenine ring.
Finally, the sixth conserved segment (Figure 6C) comprises the CT-CXXC motif. The CT-CXXC disulfide is immediately downstream of another basic residue (Lys500 in the H. sapiens QSOX1 sequence) in the vicinity of the FAD phosphates and an aromatic residue that packs against the outer face of the FAD adenine. Non-QSOX Erv domains have these same two functional residues, but in a simpler structural context: they are present in a KXXF/Y motif on the approximately ten-residue linker that directly connects the fourth and fifth helix in the Erv bundle. In QSOX, there are insertions both upstream and downstream of the basic/aromatic residue pair. Upstream is a stretch of about ten residues that constitute the second half of the backstop by packing against the third conserved segment (Figure 6C). Downstream of the basic/aromatic pair is the CT-CXXC and a weakly conserved loop preceding the fifth helix in the domain. The presence of the loop between the CT-CXXC and the fifth helix is conserved, though its composition less so. The function of this loop in the QSOX structure or mechanism is still unclear.
In summary, regions of highly conserved sequences in QSOX correspond to 1) important redox-active or structural disulfide bonds, 2) residues in direct contact with the FAD cofactor, and 3) features that appear to contribute to the interaction of the Trx and Erv domains but do not appear to be essential for the folding or fundamental catalytic activity of each module in isolation. In particular, the backstop in the ψErv/Erv module, contributed by regions #3 and #6 described above, is a prominent feature that appears to promote domain docking for electron transfer.
Evidence of functional constraints within regions of high sequence variability
Inspection of sequences that are poorly conserved may also shed light on aspects of QSOX function. An apparently poorly conserved region that nevertheless is likely to be crucial for QSOX activity is the linker between the Trx domain(s) and the ψErv region. The sequences of this linker are highly variable, but the lengths fall into a narrow range: about 20–35 amino acids between the predicted ends of the flanking secondary structure elements. A rough conservation of linker length is consistent with the proposed role of the linker in providing flexibility for relative domain rotations in QSOX function, whilst tethering the two interacting domains with high effective concentration [9]. Although some outliers appear to contain linkers of up to a hundred amino acids, these may arise from erroneous predictions of splice sites based on genome sequences.
Other positions in QSOX, in particular the N- and C-terminal extremities, appear variable in overall sequence alignments of the enzyme family. Indeed, there appear to be few constraints on the sequence upstream of the QSOX Trx domain. The apparent variability near the C-terminus, however, masks a certain underlying homogeneity. In particular, strong sequence conservation within taxonomic groups is observed in the region between the Erv domain and the transmembrane segment, and though QSOX paralogs differ from one another here, the differences are often retained across species. For example, Craniata QSOX1 and QSOX2 show a characteristic difference in the number of amino acids linking the Erv domain to the membrane (Figure 7). An extra segment present in QSOX1 contains a predicted helix followed by a set of basic residues that may be a protease processing site (Ilani T, Alon A, Grossman I, Horowitz B, Kartvelishvily E, Cohen SR, Fass D: A secreted disulfide catalyst controls extracellular matrix composition and function, submitted). The conserved differences between QSOX1 and QSOX2 suggest that the membrane-proximal region may contribute to differing function, localization, or processing of the QSOX1 and QSOX2 paralogs.