Fig 1.
Sequence alignments of the endogenous human chemokine CXCL8 and the 14 highly divergent genotypes of the HCMV-encoded UL146 chemokine.
Alignment generated with ClustalW 2.1 and adapted to illustrate differences in secondary structure. White on black illustrates identical residues, and black on grey similar residues. Important differences are color coded. The ELR motif is marked white on red and residues 100% conserved in UL146 are marked with a star (*). Genotypes 5 and 6 are shown with the longer N-terminus predicted by the SignalP 4.1 server [16]. Predicted glycosylation sites are marked black on green (N-glycosylation) and black on yellow (O-glycosylation) based on the NetNGlyc 1.0 [17] and NetOGlyc 4.0 [18] server predictions (threshold = 0.5). The two additional cysteine residues in the vCXCL1GT4 sequence (suggesting a third disulfide bond) are marked black on orange. The secondary structures of the chemokines are shown above the sequences as determined by NMR and X-ray structures for CXCL8 [23–25] and determined by Rosetta modelling for vCXCL1GT1–GT14. The proposed fourth and fifth β-strands unique to several vCXCL1 genotypes of the UL146 encoded chemokines have been marked by light blue (for the exact location of the α-helix and the β-strands for a particular UL146 genotype see S1 Fig).
Fig 2.
Purification and mass spectrometry analysis of recombinant vCXCL1GT1 and vCXCL1GT5 expressed in COS-7 cells.
(A) Cartoon representation of predicted glycosylation sites in the 14 genotypic vCXCL1 proteins (B) HPLC elution profiles of conditioned media from eukaryotic cells transfected with UL146GT1 (top panel) and UL146GT5 (bottom panel). Minor background peaks are marked numerically. mAU; milliabsorbance units. (C) MALDI-TOF mass spectrometry analysis of untreated (top chromatogram) and PNGase F treated (bottom chromatogram) samples of vCXCL1GT1 and vCXCL1GT5 (red and blue peaks isolated in panel B).
Fig 3.
Structural analysis of vCXCL1 GT1 and vCXCL1 GT5 via molecular modeling and Nuclear Magnetic Resonance (NMR) Spectroscopy.
Cartoon representations of the top ten models (lowest Rosetta total score) of vCXCL1GT1 are overlaid and viewed from the left (A) and front (B). Each polypeptide chain is rainbow colored from N- (blue) to C-terminus (red). The extended C-terminus is hidden to highlight the convergence in the modeling (core Cα-RMSD: 3.9Å) and formation of the expected C-terminal α-helix. (C) Right-side view of three of the top vCXCL1GT1 models with the wild-type C-terminus, which forms a β-hairpin in all cases, but with no consensus location. (D) The ten top models of vCXCL1GT5 are shown overlaid and viewed from the left (core Cα-RMSD: 1.8Å) and front (E). (F) Right-side view of three of the top vCXCL1GT5 models with the wild-type C-terminus, which forms a β-hairpin in all cases, but with no consensus position. (G) NMR 1H-15N HSQC spectrum of the C-terminus of vCXCL1GT1 (residues 51–95). Backbone NH assignments are indicated by the one letter amino acid code and residue number. (H) Predicted secondary structure for vCXCL1GT1 based on secondary chemical shifts for residues 51–95 using TALOS+. The modeled secondary structure of vCXCL1GT1 using RosettaCM is displayed above. (I) Plot of {1H}-15N Heteronuclear NOE values reflecting the picosecond-nanosecond motions of the vCXCL1GT1 C-terminus. Proline residues are indicated with a red P, a gap at position 82 indicates inconclusive assignment information for this residue.
Fig 4.
Screening of non-ELR vCXCL1GT5 activity on G protein signaling of an array of all 18 classical human chemokine receptors and the HCMV-encoded US28.
(A) N-terminal sequences of the wild-type and truncated versions of vCXCL1GT5. (B) Agonism of 1 μM wild-type vCXCL1GT5, vCXCL1GT5 (4–97), and vCXCL1GT5 (6–97) on chemokine receptor G protein signaling in an IP accumulation assay compared to the response of the endogenous ligand for the receptor. The following endogenous chemokines were used as a positive control (concentrations shown in brackets): CCR1;CCL5 (B: 100 nM, C: 10 nM), CCR2;CCL7 (B: 100 nM, C: 10 nM), CCR3;CCL11 (B: 100 nM, C: 10 nM), CCR4;CCL17 (B: 100 nM, C: 10 nM), CCR5;CCL5 (B: 100 nM, C: 10 nM), CCR6;CCL20 (B: 100 nM, C: 1 nM), CCR7;CCL19 (B: 100 nM, C: 10 nM), CCR8;CCL1 (B: 100 nM, C: 10 nM), CCR9;CCL25 (B: 100 nM, C: 10 nM), CCR10;CCL27 (B: 100 nM, C: 10 nM), CXCR1;CXCL8 (B: 1 μM, C: 10 nM), CXCR2;CXCL8 (B: 1 μM, C: 100 nM), CXCR3;CXCL11 (B: 100 nM, C: 10 nM), CXCR4;CXCL12 (B: 100 nM, C: 10 nM), CXCR5;CXCL13 (B: 100 nM, C: 10 nM), CXCR6;CXCL16 (B: 100 nM, C: 1 nM), XCR1;XCL1 (B: 100 nM, C: 100 nM), CX3CR1;CX3CL1 (B: 100 nM, C: 1 nM), US28Δ300;CX3CL1 (B: 100 nM, C: 1 nM); n = 3–6. All error bars are presented as SEM. (C) Antagonism of 1 μM wild-type vCXCL1GT5, vCXCL1GT5 (4–97), and vCXCL1GT5 (6–97) on chemokine receptor G protein signaling in an IP accumulation assay using submaximal concentrations of the endogenous chemokines described above; n = 3–6. All error bars are presented as SEM.
Fig 5.
Dose/response activity of vCXCL1GT1, vCXCL1GT4, vCXCL1GT6, and vCXCL1GT5 variants on G protein signaling of CXCR1 and CXCR2.
G protein signaling of CXCR1 and CXCR2 by vCXCL1GT1 and vCXCL1GT4 (A+B), wild-type vCXCL1GT5 and vCXCL1GT6 (C+D), and vCXCL1GT5 (4–97), and vCXCL1GT5 (6–97) (E+F) in an IP accumulation assay. Error bars presented as SEM. For CXCR1, the number of repeated observations for each sum curve were: CXCL8 n = 10; vCXCL1GT1 n = 8; vCXCL1GT4 n = 3; wild-type vCXCL1GT5 n = 10; vCXCL1GT6 n = 3; vCXCL1GT5 (4–97) n = 4; vCXCL1GT5 (6–97) n = 4. For CXCR2, the number of repeated observations for each sum curve were: CXCL8 n = 10; vCXCL1GT1 n = 8; vCXCL1GT4 n = 3; wild-type vCXCL1GT5 n = 10; vCXCL1GT6 n = 3; vCXCL1GT5 (4–97) n = 3; vCXCL1GT5 (6–97) n = 3.
Fig 6.
The effect of vCXCL1GT1 and vCXCL1GT5 on β-arrestin recruitment through CXCR1 and CXCR2 with/without GRKs.
Agonism of CXCL8 (A+D), vCXCL1GT1 (B+E), and wild-type vCXCL1GT5 (C+F) on the β-arrestin recruitment pathway of CXCR1 and CXCR2 in a BRET-based assay without GRKs (circle, ●), with GRK2 (square, ■), and with GRK6 (triangle, ▲); n = 3–4. Error bars presented as SEM.
Fig 7.
vCXCL1GT1, vCXCLGT5 and CXCL8-induced migration of L1.2 cells stably expressing either CXCR1 or CXCR2, and neutrophils.
Chemotactic activities of L1.2 cells stably transfected with (A) CXCR1 (n = 4) and (B) CXCR2 (n = 3) toward vCXCL1GT1, wild-type vCXCLGT5 and CXCL8 in a transwell migration assay. Representative assays with error bars presented as SEM. (C) Chemotactic activity of human neutrophils toward vCXCL1GT1, wild-type vCXCL1GT5 and CXCL8 in a transwell migration assay, shown as migration index (MI) determined from point measurement divided by the background measurement (n = 4). Representative assay with error bars presented as SEM.
Fig 8.
Functional activity and thermal melting curves for tailless and wild-type vCXCL1GT1.
(A) C-terminal sequences and secondary structure of wild-type and tailless vCXCL1GT1. (B+C) G protein signaling in an IP accumulation assay of wild-type and tailless vCXCL1GT1 on CXCR1 and CXCR2; n = 3. Error bars shown as SEM. (D+E) Thermal melting curves and their first derivative for wild-type and tailless vCXCL1GT1; n = 3.
Fig 9.
The predicted third disulfide bond of vCXCL1GT4 allows the C-terminal β-hairpin to mimic CXC chemokine dimerization.
(A) The top model of vCXCL1GT4 is shown in cartoon form with disulfide bridges shown as spheres. The canonical disulfide bonds (DS1 and DS2) stabilize chemokine tertiary structure. A unique disulfide (DS3) links the β1-strand to the C-terminal α-helix. (B) The top model of vCXCL1GT4 is rotated 180° to show the front view. vCXCL1GT4 models have a consensus location for the C-terminal β-hairpin, which continues the 3-stranded β-sheet core. This mimics the CXC dimer interface shown in panel C. The constraint from the third disulfide bond may help stabilize this conformation, as there was no consensus β-hairpin location in other genotype models. (C) The top vCXCL1GT4 model was aligned to one of the CXCL8 chains in a structure of a CXCL8 dimer (PDB ID: 6LFM). The other dimer subunit is shown in cartoon form (gray), and the extended C-terminus of vCXCL1GT4 is hidden for clarity as it overlaps with the CXC dimer interface. The third disulfide bond in vCXCL1GT4 locks the α-helix into an orientation that would clash with a classic CXC dimer. The vCXCL1GT4 α-helix angle varies 20° from the α-helix of CXCL8, shown as a dotted orange line.
Fig 10.
Alignment of vCXCL1GT1 and vCXCL1GT5 models with CXCL8 in the experimental CXCL8:CXCR2 complex.
(A) The top model (lowest Rosetta total score) for vCXCL1GT1 is shown in cartoon form (red) aligned to the structure of the CXCL8:CXCR2 complex (PDB: 6LFO). The extended C-terminus is unlikely to interfere with binding to CXCR2, as the N-terminus of CXCR2 (chemokine-recognition site 1) binds the opposite side of the chemokine. (B) Expanded view of the boxed region in panel A showing important residues as sticks and numbered by residues present in the structure. The first conserved cysteine residue is shown as a sphere. The glutamate in the ELR motif is in a very similar position for both vCXCL1GT1 and CXCL8. This residue has been shown to be critical for the activation of CXCR2 and interacts with receptor residues R208, R212, and R278. (C) The top scoring model for vCXCL1GT5 is shown in cartoon form (blue) aligned to the structure of the CXCL8:CXCR2 complex. The β-hairpin is not predicted to interrupt site-1 binding in this model. (D) Expanded view of the boxed region in panel C showing important residues as sticks and numbered by residues present in the structure. The first conserved cysteine residue is shown as a sphere. vCXCL1GT5 does not contain an ELR motif, instead the five residues proceeding the cysteine are 5-EGNGR-9 (Glu5 shown in stick form). The longer N-terminus in vCXCL1GT5 is an unstructured loop in this model, which was generated without the receptor present. In this position, it is unlikely that Glu5 will make sufficient contact with the displayed receptor residues. During a binding event, the chemokine N-terminus may adopt a stable conformation, facilitated by Arg9, whereby Glu5 could participate in receptor activation. Alternatively, Asn7 may form ELR-like contacts sufficient to activate the receptor.
Fig 11.
Hypothetical model explaining the existence of 14 stable UL146 genotypes by bottlenecks in the genetic makeup of the host.
Unidentified bottlenecks in the host allowing only one or some of several possible UL146 genotypes among circulating HCMV strains to establish infection and disseminate in a given host. Small circles show which UL146 strains are able to pass the bottleneck. Note that i) different bottlenecks can allow infection from between one and 14 genotypes, and ii) a given host/HCMV strain pair can be the result of different bottlenecks (yellow virus on human that could potentially be infected by multiple different strains). Bottom left: examples of human proteins representing potential bottlenecks. Bottom right: examples of possible research strategies for identifying host bottlenecks.