What defines a chemokine? – The curious case of CXCL17

In this article, we provide a critique of the literature regarding the discovery of CXCL17 and key aspects of CXCL17 biology which have shaped subsequent research. We have not focussed on the putative roles of CXCL17 in disease. For information on this, we direct you instead, to recent reviews by others covering these topics [1], [2], [3].

The first description of CXCL17 in the literature came late in 2005 when Pisabarro and colleagues at Genentech reported the identification of a putative chemokine secreted by dendritic cells. They named the protein Dendritic cell and Monocyte Chemokine-like protein, abbreviated to DMC [4]. The cDNA sequence encoding DMC had been identified earlier by a bioinformatics approach termed the secreted protein discovery initiative (SPDI), a large-scale effort to identify novel human secreted and transmembrane proteins which identified more than 200 novel proteins [5]. Using the ProHit fold-recognition alrogithm (ProCeryon Biosciences), the Genentech group screened 7,950 structures in the protein data bank and assigned a CXCL8-like structure to DMC [4]. Examination of the DMC cDNA sequence showed it to encode a protein of 119 amino acids with six cysteine residues, the first four of which are arranged as CXC motifs, a feature observed in a subset of chemokines [6].

Pisabarro and colleagues generated a His-tagged version of DMC which they expressed as a recombinant protein in E. coli and SF9 cells and observed it to be chemotactic in Transwell® assays for CD11c+ dendritic cells and CD14+ monocytes, albeit with low potency, optimal at 500 nM concentrations and above [4]. Chemotaxis of CD11c DCs was abolished by treatment of the cells with pertussis toxin (indicative of coupling to a Gαi coupled GPCR) and by incubation of cells with LPS. Northern blot analysis coupled with immunohistochemistry using an in-house mouse monoclonal (3H8) showed expression of CXCL17 in mucosal tissues such as the lung, duodenum and colon [4].

Independently, in mid-2006 Weinstein and colleagues at Pfizer identified a novel cDNA by analysis of genes within the human genome whose expression was tied to vascular endothelial growth factor (VEGF) [7]. This cDNA they named VEGF Coregulated Chemokine-1 (VCC-1) and is identical to the CXCL17/DMC cDNA. Realtime-PCR analysis found overexpression of VCC-1 in human colon and breast cancer tumours suggestive of a role in tumorigenesis [7]. Supportive of this, transformation of NIH-3T3 cells with adenoviral-vectors encoding CXCL17 and subsequent introduction into nude mice showed CXCL17 to significantly enhance cell growth when compared to vector controls.

The CXCL17 gene was reported to be encoded by four exons mapping to human 19q11, a locus not previously reported to encode human chemokines. Subsequent phylogenetic analysis by Zlotnik and colleagues showed both the human and mouse forms of CXCL17 to be distinct from known clusters of CXC chemokines such as the IP-10 and GRO clusters, suggestive of having evolved independently of other chemokines and likely to bind to receptors distinct from those of known chemokines [8]. Fig. 1A shows an alignment of the primary sequence of human CXCL17 alongside 18 other mammalian, marsupial and monotreme orthologues. Of note, genes encoding CXCL17 have not been detected in the genomes of avians or reptiles. What is immediately apparent is the high level of homology between human CXCL17 and the other orthologues (38.1%-98.3%), and also the high proportion of basic residues within the sequence. Phylogenic analysis of the evolution of CXCL17 (Fig. 1B) suggests that CXCL17 evolved from an ancient common ancestor before mammalian diversification, but after the differentiation of sauropsida (reptiles and avians) from synapsids in the carboniferous period, thus explaining the absence of orthologues in reptiles or avians. In silico analysis of the mouse gene encoding CXCL17 suggests the existence of an alternatively spliced variant encoding a slightly larger molecule of 128 amino acids with a substantially different C-terminus (E5QUW37; UniProt), resulting from a 1 base pair frame-shift at the 5′ boundary of exon 4 (SPG and JEP, unpublished observations). As this isoform is only detected at the transcript level, and has not experimentally verified in protein assays, the relevance of this CXCL17 variant is uncertain and remains to be explored.

Curiously, different signal peptide cleavage sites have been predicted for human CXCL17. The original description of DMC/CXCL17 reported that the SignalP 3.0 algorithm predicted the presence of a signal peptide spanning amino acid positions 1-23, which was likely cleaved at serine-23 to generate the species CXCL17 (24-119) [4]. In contrast, Weinstein and colleagues used the same algorithm and suggested that both human and mouse CXCL17 were cleaved at serine-22 to generate the slightly larger variant CXCL17 (23-119) [7]. Current commercial sources of recombinant CXCL17 produced in E. coli favour the CXCL17 (24-119) variant of the human chemokine and the CXCL17 (23-119) variant of the mouse chemokine. Given these discrepancies, it may be worthwhile to purify the chemokine from biological fluids and undertake N-terminal Edman sequencing and mass spectrometry (MS) to clarify this point.

Using a gene expression microarray database, coupled with immunohistochemistry, Burkhardt and colleagues subsequently examined the expression profiles of all known human chemokines and confirmed and extended earlier findings regarding CXCL17, showing expression to be restricted to mucosal tissues including the tongue and stomach [9]. Such findings are supported by proteomic analysis revealing the presence of high levels of CXCL17 in mucosal tissues of the respiratory system, pancreas and lymphoid tonsil tissue [10], while genomic analysis indicates CXCL17 is constitutively expressed at high levels in specific epithelial cell types within the respiratory system [11]. Given this expression pattern, the authors hypothesized that CXCL17 may possess antimicrobial properties, which was supported by evidence of bactericidal and fungicidal activity in vitro [9]. The group also developed a CXCL17-specific ELISA and showed elevated expression of CXCL17 in the bronchoalveolar lavage fluid recovered from individuals with idiopathic pulmonary fibrosis (IPF) compared with an absence of CXCL17 following lavage of the lungs of healthy volunteers [9]. Given the role for VEGF in the pathogenesis of IPF, this fits with the earlier discovery that VEGF and CXCL17 expression is co-regulated [7].

As outlined earlier, the initial description of CXCL17 by Pisabarro and colleagues described chemotactic activity for CD14+ monocytes and CD11c+ dendritic cells [4]. The next publication to examine the chemotactic activity of CXCL17 followed on from the study of Weinstein et al., [7] and genetically modified NIH-3T3 cells to express the full length CXCL17 protein [12]. Cells transformed in this manner showed enhanced oncogenic potential when introduced into both nude mice and SCID mice, producing larger tumours more quickly than control cells. Immunostaining of tumours using a CD31-specific antibody revealed CXCL17 expression to be associated with increased vascularisation and an influx of CD11b+ Gr-1+ cells of immature myeloid origin [12]. Boyden chamber chemotaxis assays with mouse splenocytes and recombinant CXCL17 revealed pertussis toxin-sensitive chemotactic responses, peaking in the 20–50 nM range which microscopically, resembled immature neutrophils. Thus, although agreeing with the report of Pisabarro that a GPCR was likely involved in CXCL17-myeloid cell recruitment, the responses would appear to be 10-fold more potent for mouse CD11b+ Gr-1+ cells.

The group of Albert Zlotnik proceeded to further explore the expression profile of CXCL17 in mice, noting that it was expressed at basal levels in mucosal tissues such as the lung, colon and tongue of pathogen-free mice but at raised levels in the same tissues of normally housed mice exposed to pathogens [13] giving rise to the notion that CXCL17 is a dual homeostatic/inflammatory chemokine. In the same study, recombinant mouse CXCL17 was administered i.p. to mice and the resulting inflammatory exudate was found to be enriched for F4/80+ macrophages but not CD11c+ DCs or Gr-1+ neutrophils, in contrast to earlier reports regarding chemotactic activity for human [4] and mouse [12] myeloid cells. In mice deficient for CXCL17, a small but significant reduction in the number of lung resident macrophages was observed compared with wild type mice. Collectively, this led the authors to conclude that CXCL17 is a chemoattractant involved in the recruitment of macrophages to the lung. Using cell suspensions produced from lung homogenates, in vitro Transwell chemotaxis assays using a single concentration of 200 ng/ml recombinant CXCL17 (approximately 20 nM) was found to recruit macrophages in a pertussis toxin-sensitive manner. As with the earlier study of human myeloid cells by Pisabarro and colleagues, this suggested the involvement of a GPCR although the chemokine was active at concentrations that the earlier study had failed to show recruited human CD11c+ dendritic cells or CD14+ monocytes [4]. Curiously, the in vitro chemoattractant properties of CXCL17 for mouse macrophages was severely blunted in cells isolated from CXCL17-deficient mice when compared with wild type mice.

In a cutting-edge manuscript published in the Journal of Immunology, Maravillas-Montero and colleagues described a series of experiments which led to the identification of the orphan receptor GPR35 as receptor for CXCL17 [14]. THP-1 cells were shown to migrate in a modified Boyden chamber assay to approximately 20 nM of recombinant CXCL17 and to respond with intracellular calcium flux to 200 nM of CXCL17. In both cases, the efficacy was enhanced by culture with prostaglandin E2 (PGE2), and in the case of chemotaxis, was sensitive to pertussis toxin. Having shown that CXCL17 did not bind to a collection of known CXC chemokine receptors, the authors examined the expression of other GPCRs in monocytes that showed the same tissue-selective expression pattern. GPR35 was identified as a candidate receptor and expression of the GPR35 ORF in the pre-B line BaF/3 resulted in transfectants which responded with an intercellular calcium flux to CXCL17 in a dose-dependent manner. Perhaps surprisingly, no transfectant chemotaxis data were provided to complement the calcium flux data. GPR35 was shown to be expressed by THP-1 cells at the mRNA level which was enhanced by treatment with PGE2. Conversely, GPR35 mRNA expression was reduced in the lungs of CXCL17-deficient mice when compared to WT mice. This, they argued, was because the distinct subset of macrophages recruited to the lung by CXCL17 specifically expressed GPR35. Thus, partially on the basis of indirect evidence, CXCL17 was claimed to be a specific GPCR for CXCL17 and the authors took to renaming GPR35 as CXCR8 [14].

Our own group was collaborating with the group of Graeme Milligan who had a long-standing interest in the pharmacology of GPR35 and its interaction with a group of drugs used in the treatment of asthma, including cromolyn disodium, zaprinast and lodoxamide [15], [16]. GPR35 had been originally identified by Brian O’Dowd as an ORF predicted to encode a GPCR based on hydrophobicity plots [17]. Subsequent studies had shown GPR35 mRNA to be expressed by eosinophils, basophils and mast cells [15] and to respond to products of tryptophan metabolism such as kynurenic acid [18] and various lysophosphatidic acids [19] albeit at micromolar concentrations. This has led some to question the physiological relevance of this finding [20]. The human genome encodes two distinct GPR35 isoforms known as GPR35a and GPR35b generated by variations in N-terminal splicing, with GPR35b differing from GPR35a by an additional 31 amino acids at the N terminus [21].

Following the publication by Maravillas-Montero and colleagues, we transiently expressed N-terminally tagged GPR35 at the cell surface of the mouse pre-B L1.2 cell line but were unable to show specific chemotaxis to a wide range of CXCL17 concentrations [22]. Notably, at micromolar concentrations, CXCL17 induced the chemotaxis of both mock-transfected and GPR35 transfected cells, suggesting an endogenous receptor on mouse pre-B cells with low affinity for CXCL17. A stable cell line expressing GPR35 was also unresponsive to CXCL17 in calcium flux assays and endocytosis assays. In a BRET assay, human, mouse and rat orthologues of CXCL17 were unable to induce β-arrestin recruitment or G-protein coupling over a wide concentration range, unlike the control agonists zaprinast and lodoxamide, which are well-documented synthetic agonists of GPR35 [15], [23]. Turning to a monocyte line, we found that PGE2-treated THP-1 cells were weakly chemotactic in response to gradients formed by micromolar concentrations of CXCL17, in contrast to control chemokines such as CCL2. ML-145, a potent antagonist of GPR35 was unable to block the CXCL17-mediated chemotaxis. Thus, we concluded that a receptor distinct from GPR35 was likely responsible for the previously described chemotactic activity [22]. Independently, the group of Im had arrived at a similar conclusion [24]. Using HEK293-GPR35 transfectants, they had observed that lodoxamide but not CXCL17 induced G protein-coupling and that this activation was sensitive to the GPR35-selective antagonist, CID2745687. Chemotactic responses of THP-1 cells to CXCL17 were insensitive to either CID2745687 or siRNA-mediated knockdown of GPR35, again suggestive of an alternative receptor being responsible for the actions of CXCL17 [24].

More recently, the serotonin metabolite 5-hydroxyindoleacetic acid (5-HIAA) has been shown by the group of Jason Cyster to be a high potency ligand for GPR35, driving the chemotaxis of GPR35 transfectants and GPR35+ neutrophils in a dose-dependent manner, optimally chemotactic at concentrations of 3–10 nM [25]. 5-HIAA is released from activated platelets and mast cells and appears to drive neutrophil diapedesis into infected tissues where activated neutrophils have upregulated GPR35 [25]. Thus, it appears that the hunt for a bona fide CXCL17 receptor continues.

The current paradigm of chemokine receptor activation by chemokines points to a pivotal role of the chemokine N-terminus in driving conformational changes by binding to an intrahelical pocket within the chemokine receptor [26], [27], [28], [29]. Notably, extension or truncation at the chemokine N-terminus has been shown to greatly influence the potency and efficacy of chemokine receptor signalling [30], [31], [32]. Given the uncertainty regarding the native N-terminus of secreted CXCL17, this demands further analysis. Alignment of human CXCL17 (24-119) with the other mature CXC chemokines is possible (Fig. 2) and conserves the relative positions of three of the six cysteine residues with cysteine residues found in the other human CXC chemokines. Upon alignment, it becomes readily apparent that CXCL17 has a comparatively long N-terminal domain of more than 40 amino acids which contains the first of the two CXC domains.

Similar to CXCL17, the human CC chemokines CCL6 [33], CCL9 [34], CCL15 [35] and CCL23 [34] all have six conserved cysteine residues and have N-terminal domains of around 30 amino acids upstream of the CC domain. This group of CC chemokines have weak potency in chemotaxis assays, which in several instances has been shown to be increased several fold by N-terminal truncation either with proteases such as mast cell chymase or via recombinant protein engineering [33], [36], [37]. We hypothesized that given the expression of CXCL17 at mucosal surfaces where mast cells reside, cleavage of CXCL17 by mast cell chymase may generate a more potent form of CXCL17. Incubation of recombinant human CXCL17 (24-119) with recombinant mast cell chymase was observed to induce cleavage of CXCL17 at an unidentified location, but failed to increase its potency in assays of THP-1 cell chemotaxis [22]. Use of other enzymes such as neutrophil elastase to cleave and activate CXCL17 was not explored.

Lee and colleagues reported that the rat orthologue of CXCL17 was expressed at the mRNA level predominantly in the rat glandular stomach, analogous to the human stomach [38]. Immunohistochemistry of the same tissue using an anti-mouse CXCL17 found CXCL17 to be intensively expressed in the mucosal layer in keeping with earlier findings describing a mucosal expression pattern [9]. Using the same antibody in Western blotting of conditioned media from the KATO III rat gastric cell line, two bands with approximate molecular weights of 8 kDa and 22 kDa were reported, the lower band approximating to the predicted molecular weight (11 kDa) of the CXCL17 (24-119) species. The same two band pattern was also observed when the conditioned media of HEK-293T cells transfected with a C-terminally FLAG-tagged version of CXCL17 (1-119) was probed, which presumably undergoes signal peptide cleavage. The authors concluded that the higher molecular weight species of around 22 kDa was a pro-peptide which was cleaved by endoproteases to reveal a shorter active form in which an N-terminal fragment containing the first CXC motif was removed [38]. Analysis of the rat CXCL17 primary sequence by the authors suggested cleavage around the basic trio of Lysine-61-Lysine-63, producing a mature species CXCL17 (64-119). The authors went on to mutate the corresponding trio of Arginine-61-Lysine-63 of the human CXCL17 ORF to alanine and observed that HEK-293T cells transfected with this construct produced only the higher molecular weight band of 22 kDa. Thus, they concluded that Arginine-61-Lysine-63 of human CXCL17 was a recognition site for proteolytic cleavage. N-terminal sequencing (Edmann degradation) of immunoprecipitated CXCL17 from rat lung homogenates yielded a fragment with N-terminal XAVL residues, which was found to match the sequences of Alanine-66 to Leucine-68 of rat CXCL17. In rat CXCL17 but not in human CXCL17, a threonine is found at position 65 which could potentially undergo O-linked glycosylation and which would be undetectable via Edman degradation (accounting for the “X”). The authors went on to generate recombinant forms of human CXCL17 (24-119) and CXCL17 (64-119) in E. coli which they named 6-Cys-CXCL17 and 4-Cys-CXCL17 respectively. The 4-Cys-CXCL17 form was found to have enhanced activity compared to the 6-Cys-CXCL17 form in chemotaxis and VEGF-induction assays employing THP-1 cells when used at a single concentration of 300 nM [38].

We have wrestled with these interpretations and find them to be problematic. Firstly, the band of around 22 kDa seen on western blotting tallies with neither the predicted molecular weight of 13.8 kDa for CXCL17 (1-119) nor with the predicted weight of 11.4 kDa for the CXCL17 (24-119) species. Secondly, a CXCL17 (64-119) species as postulated by the authors has a predicted molecular weight of 6.5 kDa but appears to run considerably higher than the 6 kDa molecular weight marker used in their analysis [38]. We have conducted similar western blotting approaches on both the commercially available human CXCL17 (24-119) and lysates from CHO cells expressing a C-terminally His-tagged version of CXCL17 (1-119). In these studies, we have observed the same two band pattern for both recombinant and mammalian expressed CXCL17 preparations (SPG and JEP manuscript in preparation). Since the commercially available preparation is particularly unlikely to undergo proteolytic cleavage, we conclude that the two species, in all likelihood, represent monomers and dimers of CXCL17 (24-119) running at the predicted weights of 11 and 22 kDa. We also found that expression of human CXCL17 (1-119) in CHO cells in the presence or absence of Benzyl GalNAc (an inhibitor of O-liked glycosylation) did not result in a predicted shift in apparent molecular weight which would be indicative of glycosylation and could have accounted for differences in apparent molecular weight (SPG and JEP, manuscript in preparation). An alternative explanation of the apparent failure of the Lysine-61-Argine-63 triple alanine mutant form of human CXCL17 to undergo cleavage [38] is that it has a propensity to exist as a dimer rather than a monomer. The fact that truncated forms of rat CXCL17 were observed in lung homogenates is perhaps unsurprising, given that such fluids are likely to be enriched for a variety of proteases. The relevance of CXCL17 truncations in the human setting remains to be established.

To date, the structure of CXCL17 remains unresolved, although several efforts have been made to model the structure in silico based on those of well characterised CXC chemokines. In the first attempt at CXCL17 modelling, the primary sequence was threaded against a small PDB library containing 7,950 entries. From this alignment, of the top 20 hits, those ranked 6th and 12th (CCL3: 16.2%; and CXCL8: 14.3% identity), were used to justify the generation of a new library of chemokines with CXCL8-like folds [4]. Notably, within this library, the alignment with the highest level of homology was generated with the CXCL8 mutant E38C/C50A (1ICW) which was initially engineered to rearrange the disulphide linkages [39]. As might be anticipated, these mutations were reported to result in a 500–2,500 fold decrease in binding to the CXCL8 receptors when compared to wild type CXCL8 [39]. Pissabaro et al., also reported CXCL17 homology to CCL4 and CCL5 based on CD spectroscopy findings [4] although crucially, comparisons to CXCL8 were not performed. Secondary structure predictions for CXCL17 by using the PHD [40] and DSC [41] servers were reported to be in agreement with a CXCL8-like fold although these data were not reported in the paper [4].

Recently, however, the notion that CXCL17 has a chemokine fold has been challenged by modelling by others [42], [43], [44]. Denisov used the DSC server to ascertain secondary structural predictions and came up with a structure which was predominantly α-helical, a prediction with which our own efforts concur. Using the favoured mature sequence of CXCL17 (24-119) we obtain a similar prediction of four helices bundle via the DSC server (Fig. 3A). We subsequently employed the I-TASSER server to carry out our own modelling. I-TASSER carries out multiple threading of sequence alignments followed by iterative fragment assembly simulations [45], [46], [47]. Using this approach, we obtained five models of the CXCL17 structure (Fig. 3B). All five models agreed with the DSC predications, being composed of 3–4 α-helices and without resemblance to the classical chemokine fold. These look to be in broad agreement with I-TASSER modelling reported by Sun et al., [44], and presented in a supplementary figure by Niiya and colleagues [43], although in both of these manuscripts, no discussion of the lack of apparent chemokine-like structure for CXCL17 was made.

With the benefit of hindsight, we can reflect upon the initial folding prediction made by Pisabarro and colleagues for CXCL17 using ProHit [4]. Notably, by their own confidence criteria, other proteins in the initial folding library with greater homology to CXCL17 were overlooked, such as the RXR-α DNA-binding domain (1dsz.B). The structure of this protein has been solved and comprises a zinc-stabilized bundle of 3 α-helices with a short length of β-sheet [48], the helical portions of which are perhaps more consistent with the in silico models of CXCL17 generated by ourselves and others.

留言 (0)

沒有登入
gif