Structural basis for potent antibody-mediated neutralization of human cytomegalovirus

See allHide authors and affiliations

Science Immunology  30 Jun 2017:
Vol. 2, Issue 12, eaan1457
DOI: 10.1126/sciimmunol.aan1457

Function follows form

Congenital human cytomegalovirus (HCMV) is the most common infectious cause of disabilities in newborn infants and the leading cause of deafness in children, highlighting the need for a vaccine that induces neutralizing antibodies to block maternal-fetal transmission of HCMV. Now, Chandramouli et al. report crystal structures of neutralizing antibodies bound to the HCMV pentameric complex (Pentamer), a key determinant of viral entry. These structural and functional studies identify potential entry receptor–binding sites on Pentamer as well as other functional sites that may serve as targets for vaccine development and antibody and small-molecule therapeutics.


Human cytomegalovirus (HCMV) is the leading viral cause of birth defects and organ transplant rejection. The HCMV gH/gL/UL128/UL130/UL131A complex (Pentamer) is the main target of humoral responses and thus a key vaccine candidate. We report two structures of Pentamer bound to human neutralizing antibodies, 8I21 and 9I6, at 3.0 and 5.9 Å resolution, respectively. The HCMV gH/gL architecture is similar to that of Epstein-Barr virus (EBV) except for amino-terminal extensions on both subunits. The extension of gL forms a subdomain composed of a three-helix bundle and a β hairpin that acts as a docking site for UL128/UL130/UL131A. Structural analysis reveals that Pentamer is a flexible molecule, and suggests sites for engineering stabilizing mutations. We also identify immunogenic surfaces important for cellular interactions by epitope mapping and functional assays. These results can guide the development of effective vaccines and immunotherapeutics against HCMV.


Human cytomegalovirus (HCMV), a β-herpesvirus, can cause cross-placental infection of a developing fetus, potentially leading to neurologic deficits such as sensorineural hearing loss and cognitive impairment (1, 2). HCMV is also a substantial threat to transplant recipients (3) and has been associated with numerous inflammatory diseases, cancers, and cardiovascular pathologies (4). Development of an effective HCMV vaccine is thus a public health priority (5).

The envelope glycoprotein complex gH/gL of HCMV is essential for viral entry and may function by activating the fusion protein gB (6). Most gH/gL heterodimers on the virion form either a ternary complex with glycoprotein gO (gH/gL/gO) or a pentameric complex (Pentamer) with UL128, UL130, and UL131A (referred to here as the “ULs”). The gH/gL/gO complex allows HCMV entry into fibroblasts, and recent data suggest that it may contribute to entry into all cell types, whereas Pentamer is required for entry into dendritic, endothelial, and epithelial cells and leukocytes, most likely through a receptor-mediated endocytic pathway (710). Previous studies have revealed that the ULs and gO form mutually exclusive complexes with gH/gL (10, 11), and hence, a balance between their levels may control cell tropism.

Pentamer is the main target of neutralizing antibody responses against HCMV, and Pentamer-based vaccines elicit strong and broadly neutralizing responses in multiple animal models (1216). Pentamer-specific antibodies are 100- to 1000-fold more potent than those targeting gH/gL/gO or gB for neutralization of epithelial and endothelial cell infection (17). Such antibodies may have important protective features, because they inhibit infection of human placental cytotrophoblasts (18), which are the cell types critical for HCMV transmission to fetuses. Early antibody responses to Pentamer in pregnant women with a primary HCMV infection were associated with a reduced risk of fetal transmission (19, 20). Therefore, Pentamer represents a key vaccine target. Determination of its structure and interactions with neutralizing antibodies is critical for the development of effective vaccines and antibody therapeutics against HCMV.

Seven broadly neutralizing epitopes (sites 1 to 7) on Pentamer have been identified previously, and five of them are nonoverlapping (17, 21). Here, we report the structure of Pentamer bound to 9I6 Fab, a site 5 neutralizing antibody, at 5.9 Å resolution, and Pentamer bound to a site 7 Fab, 8I2I, at 3.0 Å resolution. The structures and accompanying biochemical and cell-based functional analyses reveal potential entry receptor–binding sites on Pentamer and suggest mechanisms for antibody-mediated neutralization of HCMV. Together, our data provide new insights into the mechanism of HCMV neutralization and Pentamer function and will guide vaccine antigen design and the development of antibody and small-molecule entry inhibitors.


Structure determination

To gain structural information on Pentamer and its interactions with neutralizing antibodies, we obtained crystals of Pentamer bound to 9I6 Fab, which diffracted to 5.9 Å resolution, and crystals of Pentamer bound to 8I21 Fab, which diffracted up to 3.0 Å resolution (Table 1). We first determined the structure of the Pentamer–8I21 Fab complex by single anomalous dispersion (SAD) and noncrystallographic averaging. We then completed the structure of Pentamer–9I6 Fab by molecular replacement (MR) using the refined Pentamer from the 8I21 Fab complex as a search model (Table 1).

Table 1 X-ray data collection and refinement statistics.

Values in parentheses are for the outer shell. Rwork = ∑||F(obs)| − |F(calc)||/∑|F(obs)|. Rfree = Rwork but calculated for 5% of the total reflections, chosen at random, and omitted from refinement. APS, Advanced Photon Source; ESRF, European Synchrotron Radiation Facility; Rmsd, root mean square deviation.

View this table:

Overall structure of the HCMV Pentamer

HCMV Pentamer adopts a helicoid shape 180 Å in length and 30 to 80 Å in cross section (Fig. 1 and movie S1). The gH/gL part of the complex has a similar domain organization to that of other herpesviruses, with gH domains (D-I to D-IV) stacking on top of each other and the N-terminal D-I cofolding with gL (Fig. 1 and fig. S1A). The polypeptide chains of the ULs form a gently curved subcomplex that binds to an extension at the N terminus of gL. Ten glycans—6 in gH, 1 in gL, and the remaining 3 in the ULs—decorate one face of Pentamer. The other face features a single glycan in UL130 and positively charged areas with clusters of exposed arginine and lysine residues (Fig. 1 and fig. S2). Similar to other herpesvirus gH/gLs, a large negatively charged surface is located at the bottom of the gH C-terminal domain, suggesting a conserved role (fig. S2) (2225).

Fig. 1 Crystal structure of HCMV Pentamer.

Two 180°-rotated views of a cartoon (A) and surface (B) representation of Pentamer. gH, gL, UL128, UL130, and UL131A are colored and labeled in gray, green, magenta, yellow, and blue, respectively. Disulfide bonds are shown in (A) as black sticks, and Asn-linked oligomannose residues modeled using the GlyProt server (41) are shown as spheres colored in black, red, blue, and white for carbon, oxygen, nitrogen, and hydrogen atoms, respectively.

HCMV gH/gL structure reveals a close similarity to Epstein-Barr virus gH/gL

Structural comparison revealed a close similarity between the gH domains of the γ-herpesvirus Epstein-Barr virus (EBV) and HCMV (fig. S1B) including the presence of a D-I/D-II linker helix proposed to act as an internal hinge during activation of EBV membrane fusion, the number and length of helices in D-III, and several conserved disulfide bonds (24). Nonetheless, as a result of interdomain rotations, HCMV gH/gL adopts a structure that is intermediate between the boot-shaped and the rod-like conformations of herpes simplex virus-2 (HSV-2) and EBV gH/gLs, respectively (fig. S1A).

The structure analysis also reveals some notable differences between EBV and HCMV gH, particularly in the relative orientation of D-I and D-II. Consequently, the groove between D-I and D-II, implicated in EBV epithelial cell fusion (26), is less pronounced in HCMV gH/gL (fig. S1A). In addition, HCMV gH has an N-terminal extension, including three β strands (gH-β1/β2/β3) interacting with residues from both gH and gL (Fig. 2 and fig. S1), which is missing in EBV gH. The first 20 N-terminal residues of this region, disordered in the present structure, have been shown to be a target for strain-specific neutralizing antibodies (27).

Fig. 2 Analysis of Pentamer interfaces.

Interfaces of Pentamer with gH, gL, UL128, UL130, and UL131A are colored in gray, green, magenta, yellow, and blue, respectively. The N-terminal extension in gL is colored bright green, whereas only D-I of gH is shown for clarity. Starting with the top right and going clockwise, boxed subpanels show details of the interactions at the subunit interfaces of UL130-UL131A, gL-UL130, gH-gL, gL-UL128, and UL128-UL130-UL131A, with key residues depicted as sticks. Disulfides are shown as black sticks, and polar contacts are shown as dashed black lines. For gH/gL, only the interface involving the double disulfide between the respective N termini is shown.

The structural similarity with EBV is retained for the HCMV gL C-terminal domain that adopts a chemokine fold followed by a long α helix (fig. S1B). The similarity between HSV-2, VZV, and EBV gLs and chemokines has been previously highlighted (28, 29). However, the HCMV gL C-terminal domain lacks the conserved disulfide bonds of chemokines, suggesting that it may have further diverged from the common ancestor gene. HCMV gL also has an N-terminal extension of 120 residues missing in EBV gL (figs. S1 and S3). Within this extension, gL-Cys47 and gL-Cys54, at the opposite ends of helix α1, form disulfide bonds with gH-Cys95 and gH-Cys59, respectively, cross-linking gL and gH (Fig. 2). The rest of the gL N-terminal extension forms a separate domain, composed of a three-helix bundle (gL-α2α3α4) followed by an elongated β hairpin (gL-β1β2) (Fig. 2 and fig. S1) that mediates most of the interactions with the ULs. The N-terminal extension of gL and all cysteines forming disulfide bonds between gH and gL are strictly conserved in β-herpesviruses (fig. S4), suggesting that this region of gL has evolved to be a multipurpose platform that allows docking by different proteins such as ULs, gO, or gQ1/gQ2 in human herpesvirus-6.

ULs fold into an α/β core flanked by exposed chemokine domains

The ULs assemble around a central core domain formed by UL131A and the UL130 C-terminal end that together form a flat β sheet covered on one face by helices (UL130-UL131A α/β core) (Fig. 2). The UL130 and UL128 N-terminal regions form two globular domains with a chemokine fold of C- and CC-type (28, 30), respectively, both docking at opposite ends of the UL130-UL131A α/β core (Fig. 2 and fig. S5). The UL130 chemokine domain contacts the gL N-terminal domain (Fig. 2), whereas the UL128 chemokine domain packs against a 30-residue small domain (UL128-α2β4β5β6). The latter is held by a claw-like structure formed by two UL130 β strands (UL130-β4β5) and a helix-loop-helix from the N terminus of UL131A (UL131A-α1α2) (Fig. 2).

Remarkably, UL128 is connected to gH/gL by a 50 Å–long flexible linker that projects from UL128-α2β4β5β6 to end in a three-turn α helix (UL128-α3) docking on a hydrophobic groove formed by gL-α2α3α4 and gL-β1β2 (Fig. 2). Consistent with previous mass spectrometry (MS) data (11), Cys162, immediately after UL128-α3, forms a disulfide bond with Cys144 on gL-α4, securing UL128-α3 to the gL helical bundle. Thus, the gL-α2α3α4 bundle, UL128-α3, and gL-β1β2 are constrained to each other and are, in turn, connected to the tip of Pentamer by the flexible linker. Cys144 on gL is also strictly conserved among β-herpesviruses.

Small and flexible interdomain interfaces

The elongated structure of Pentamer has relatively small interfaces among the subunits and several large cavities, suggesting intrinsic flexibility of the complex (Fig. 2, fig. S6, and table S1). The UL128 chemokine domain packs loosely against the UL128-α2β4β5β6 domain with only a few hydrophobic contacts and one salt bridge. Likewise, the UL130 chemokine domain uses only a few residues, mainly from its C-terminal helix (UL130-α2) and the N-terminal loop [“N-loop” in chemokines (28)], to interact with the UL130-UL131A α/β core. Similarly, the gL/UL130 interface is centered on the long gL-β1β2 hairpin that extends the β sheet of the UL130 chemokine domain. Finally, the side chains of the hydrophobic residues from the UL128 C-terminal α3 occupy only part of the hydrophobic groove formed by the gL three-helix bundle, suggesting that UL128-α3 might be mobile (Fig. 2 and fig. S6).

Structural analysis also reveals buried histidine pairs facing each other that could destabilize the complex if protonated at acidic pH. Specifically, UL130-His209 and UL131A-His35 located at the interface between UL130 and UL131A N terminus may affect the interaction of UL128 with the UL130-UL131A α/β core (Fig. 2), whereas UL130-His150 and UL131A-His69, underneath UL130 α2, might destabilize the interaction between the UL130 chemokine domain and the UL130/UL131A α/β core (Fig. 2).

Molecular basis for 8I21 binding to Pentamer and virus neutralization

The Pentamer–8I21 Fab structure reveals that the Fab binds to a concave and positively charged surface contacting almost exclusively residues from the UL130 chemokine domain (Fig. 3, A and B, and table S2). Two glycans, attached to UL130-Asn85 and UL130-Asn118, flank the 8I21 epitope, potentially limiting accessibility. The interface buries 2690 Å2 of total surface area and has a surface shape complementarity of 0.71. Both values are above the range observed for most antigen-antibody complexes (31), consistent with the high binding affinity (KD = 0.15 nM; table S3).

Fig. 3 Pentamer–8I21 Fab and Pentamer–9I6 Fab interactions and structural superpositions.

(A) Structure of the Pentamer–8I21 Fab complex, with 8I21 HC (H) and LC (L) colored in red and teal, respectively. The boxed region on the complex is shown on the right as an “open-book” view, revealing details of the Pentamer–8I21 Fab interface. Oligomannose residues linked to Asn85 and Asn118 of UL130 are shown as sticks. (B) Details of Pentamer–8I21 Fab interactions. Residues involved in direct H-bond interactions are shown with sticks, and direct bonds are depicted with blue dashes. (C) Structure of the Pentamer–9I6 Fab complex, with Pentamer depicted as cartoon and colored as in Fig. 1 and (A), and 9I6 HC and LC depicted as brown and light blue surfaces, respectively. A top view of the 9I6 epitope on Pentamer (boxed region) is shown as surface on the right and colored according to subunits (UL128, magenta; UL131A, blue), with residues contacting the HC and LC of 9I6 colored in brown and light blue, respectively. (D) gH-based superposition of the 4.0 Å–resolution (violet) and 3.0 Å–resolution (cyan) Pentamer–8I21 Fab and of the 9I6 (gold) and the 3.0 Å–resolution 8I21 (cyan) complexes viewed from the side and from the top. A UL130/ UL131A-based superposition of the 9I6 (gold) and the 3.0 Å–resolution 8I21 (cyan) Fab complexes is also shown in the upper right panel. Blue semitransparent arrows in (D) show the movements of secondary structure elements.

The main feature of the Pentamer–8I21 Fab complex is the interaction between the long heavy chain complementarity-determining region 3 (HCDR3) of the Fab and the UL130 chemokine domain (Fig. 3, A and B, and table S2). Arg104 and Trp105, at the tip of HCDR3, protrude into a crevice composed of hydrophobic and polar residues from the N-terminal UL130-α1 helix and UL130/gL β sheet, establishing H-bonds with UL130-Ser47 and gL-Asp156, respectively. Mutation of HCDR3 Trp105 to alanine resulted in a more than 150-fold decrease in binding affinity, consistent with its prominent role in the interaction (table S3). Several other interactions stabilize the complex, including hydrogen bonds between the guanidinium group of HCDR3 Arg107 and the hydroxyl group of UL130-Tyr46, and H-bonds between main-chain carbonyl oxygens of HCDR3 and the side chains of light chain CDR3 (LCDR3) Trp94 and Trp97 (Fig. 3B).

Neutralizing antibody binding site 5 revealed by Pentamer–9I6 Fab structure

The structure of the Pentamer–9I6 Fab complex, determined by MR and refined at 5.9 Å resolution, reveals a very elongated structure more than 220 Å in length, with the 9I6 Fab binding on the tip of the ULs. Although the low resolution of the data does not allow us to dissect specific interactions, the structure suggests that 9I6 CDRs contact both UL128 (residues 47 to 52 on the chemokine domain and residues 92 to 93 and 106 to 109 on α2β4β5β6) and UL131A (residues 23 to 24 and 27 to 31) (Fig. 3C). The epitope is consistent with published negative stain electron microscopy (NS-EM) data (21) and mapping studies suggesting that site 5 antibodies require all three ULs for binding (17), likely due to the cofolding of UL130 and UL131A.

Pentamer is flexible and can accommodate large rotations of gH/gL D-I and ULs

Superpositions among two Pentamer–8I21 Fab structures, obtained from two non-isomorphous crystals (Table 1), and the Pentamer–9I6 Fab structure reveal differences in the gH/gL D-I domain, the gL N-terminal extension, and the ULs (Fig. 3D and movie S2). These regions of Pentamer undergo a rigid-body rotation around the gH D-I/D-II linker helix, which acts as a hinge, with the ULs moving as a rigid arm, and resulting in differences among equivalent carbon α atoms of the UL130/UL131A α/β core in the range of 10 to 30 Å. When the UL130/UL131A α/β cores of the Pentamer–8I21 Fab and Pentamer–9I6 Fab complexes are superposed, a rotation of the UL128 chemokine domain around the loop connecting the UL128-α1 and -α2 helices also becomes evident (Fig. 3D). This movement is likely induced by 9I6 Fab binding, further suggesting that the UL128 chemokine domain loosely attaches to the rest of Pentamer.

Combination of HDX-MS and EM reveals binding sites for Pentamer-specific neutralizing antibodies

To locate other neutralizing epitopes, we fitted the crystal structure of Pentamer and homology-modeled Fabs into NS-EM reconstructions [reported previously in (21)] of Pentamer bound to 3G16, a gH-specific Fab, and combinations of representative Fabs binding to five nonoverlapping sites (1, 2, 3/7, 4/6, and 5). The Pentamer model from the 8I21 or 9I6 Fab complexes fitted well in the NS-EM density maps of Pentamer-3G16-10P3-8I21 complex (sites 4 and 7) and Pentamer-3G16-15D8-2C12 complex (sites 1 and 5) (Fig. 4). However, none of the three x-ray structures of Pentamer fitted well into the density maps of Pentamer-3G16-10F7 complex (site 2), requiring a large rotation of the ULs (Fig. 4B). Together, these data confirm that the ULs can undergo large rigid-body rotations, and suggest that antibodies can stabilize the ULs in alternative conformations. In addition, our structural analyses reveal that 10P3 (site 4) and 15D8 (site 1) bind into the concave surface of the ULs, and 10F7 (site 2) binds on the other face of Pentamer along the UL130/UL131A β sheet.

Fig. 4 Neutralizing epitopes and cell binding of Pentamer.

EM maps of Pentamer-Fab complexes from (21) are shown as gray semitransparent isosurfaces, and crystal structures of Pentamer-Fab complexes described in this work are shown as cartoons and colored as in previous figures. Red surfaces mapped on the crystal structures show Pentamer peptides identified by HDX-MS that are closest to the relative Fab for 15D8 (A), 10F7 (B), 10P3 (C), and 2C12 (D) in the EM-fitted models. Red labels show the specific Fab for which the epitopes are shown as red surfaces. Peptides that are part of the epitopes are labeled below each fitting.

Hydrogen deuterium exchange MS (HDX-MS) analysis of Pentamer-Fab complexes was performed to expand the EM analysis (Fig. 4, figs. S7 to S9, and table S4). Consistent with our EM fitting, some of the peptides with reduced deuterium uptake are located within the epitopes. However, additional peptides are far from the antibody binding sites [for example, all antibodies affect common residues in gH, gL, and UL130 (fig. S9)]. Likely, the interaction with Fabs stabilizes Pentamer, resulting in differential deuterium incorporation in regions not directly involved in the binding.

Pentamer binds to endothelial/epithelial cells, and binding is inhibited by a subset of neutralizing antibodies

To further dissect the molecular mechanisms of antibody-mediated HCMV neutralization, we first tested whether recombinant Pentamer could bind epithelial and endothelial cells. As shown in Fig. 5A, we detected strong, specific binding of Pentamer to human umbilical vein endothelial cells (HUVECs), and moderate binding to adult retinal pigment epithelial cells (ARPE-19), but barely detectable binding to MRC5 fibroblasts, consistent with previous results (9). We next assessed whether neutralizing monoclonal antibodies (mAbs) could interfere with Pentamer binding to the cells. Preincubating Pentamer with mAb 15D8 (site 1), 10P3 (site 4), 9I6 (site 5), or 7I13 (site 6) inhibited Pentamer binding to endothelial cells (Fig. 5, B and C). In contrast, mAbs 10F7 (site 2), 4N10 (site 3), and 8I21 (site 7) did not affect Pentamer binding to cells. Thus, our data suggest that Pentamer-specific antibodies likely neutralize HCMV through interference of multiple Pentamer functions during viral infection.

Fig. 5 Binding of Pentamer to cells.

(A) Binding of Pentamer, gH/gL, Shiga toxin B (positive control), or TEV protease (negative control) to ARPE-19, HUVEC, and MRC5 cell surface. Abs, absorbance at 450 nm. (B) Effect of neutralizing antibodies on binding of Pentamer to HUVEC cell surface. The graph represents relative binding compared with the control sample with no antibodies (100%). (C) Two 180°-rotated surface views of Pentamer colored in white (gH), and four shades of gray from dark to light for gL, UL128, UL130, and UL131A. The location of the epitopes for neutralizing antibodies, as determined from the combination of x-ray crystal structures, NS-EM, and HDX-MS (this study), is depicted with colored surface patches (blue for site 1, red for site 2, green for site 3/7, pink for site 4/6, and orange for site 5). Additional epitopes located on gH and previously described (21) are mapped in cyan (site A; mAb 3G16), yellow (site B; mAb 13H11), and dark red (site C; mAb MSL-109). Glycosylation sites on the surface of Pentamer, modeled as oligomannose using the GlyProt server, are depicted with spheres, colored in black, red, blue, and white for carbon, oxygen, nitrogen, and hydrogen atoms, respectively. Notably, several of the neutralizing sites on the ULs are in the vicinity of, and potentially partially protected by, N-linked glycans.


HCMV Pentamer is a key determinant of host cell entry and an important target for the development of vaccines and neutralizing antibodies (14, 32). The Pentamer structure reveals that the gH and gL components of the complex have a close structural similarity with EBV gH/gL, whereas the ULs are characterized by an α/β core domain flanked at opposite ends by UL128 and UL130 chemokine domains. Notably, HCMV gL has a unique N-terminal extension, missing in EBV and HSV gLs, which forms a docking site for the UL128 C-terminal α3 helix and the UL130 chemokine domain.

Characteristic features of the Pentamer structure include relatively small interfaces and cavities between some domains, suggesting intrinsic flexibility of the complex. Structure comparisons revealed large rigid-body rotations of the gH/gL D-I domain and the ULs arm around the gH D-I/D-II linker helix, resulting in a large displacement of the ULs. Although the 8I21 epitope remains the same in the Pentamer–9I6 Fab complex, it is likely that antibody binding stabilizes Pentamer in discrete, yet different, conformational states. Single-particle reconstructions reveal a large rigid-body rotation of the ULs in Pentamer bound to 10F7 Fab compared with the 8I21 and 9I6 Fab complexes. Consistent with this hypothesis, HDX-MS analysis of Pentamer-Fab complexes shows that antibodies stabilize regions of Pentamer distant from their corresponding epitopes. Therefore, the combination of HDX-MS and crystallographic data reveals areas of Pentamer where mutations could be introduced to generate a more stable complex. Previously, greater stability of a prefusion conformation of the respiratory syncytial virus fusion protein was linked with improved immunogenicity (33).

The structural analysis of the Pentamer-Fab complexes, together with cell binding analysis, provides new insights into the mechanism of antibody-mediated HCMV neutralization. We show that Pentamer binds to ARPE-19 and HUVEC cells but not to MRC-5 cells. We also show that antibodies binding to UL128 and UL131A residues located at the tip of Pentamer (9I6; site 5), UL128-α3 helix (15D8; site 1), and close to the linker connecting UL128-α2β4β5β6 to UL128-α3 (10P3 and 7I13; sites 4 and 6, respectively) prevent Pentamer binding to cells. These antibodies may inhibit the interaction of Pentamer with a cell surface receptor, by either direct competition or steric hindrance, and their epitopes identify a surface of Pentamer as the potential binding site for a cell surface receptor.

In contrast, antibody binding to the elbow of the ULs arm (4N10 and 8I21; sites 3 and 7, respectively) and the solvent-exposed side of the UL130/UL131A β sheet (10F7; site 2) did not affect Pentamer binding to cells, suggesting a different mechanism of neutralization. 8I21 binds to a positively charged surface on UL130, with a long HCDR3 simultaneously protruding into a groove in the UL130 chemokine domain and contacting UL130 N-terminal residues, both of which are implicated in receptor binding in chemokines. It is plausible that this site in UL130 might bind to a co-receptor at the cell surface or in a post-entry step. Therefore, site 2 and site 3/7 antibodies may interfere with these interactions and processes without affecting Pentamer binding to cells.

Our structural and functional characterizations of neutralizing mAbs allow us to propose a potential mechanism for Pentamer-mediated activation of HCMV entry. We suggest that a cell receptor would bind to a surface in proximity of UL128 and the epitopes for site 1 and 4/6 neutralizing antibodies. Receptor binding may, in turn, result in a rotation of gH/gL D-I mediated by the UL128 linker and UL128-α3 interaction with the gL three-helix bundle. Repositioning of D-I may affect the width of the D-I/D-II groove, implicated in gB binding in EBV gH/gL, ultimately triggering membrane fusion.

In conclusion, the structure of Pentamer reveals binding sites for potent and broadly neutralizing mAbs, suggesting the location of important functional sites and targets for antibody therapeutics. The structures also reveal a dynamic repositioning of the ULs upon antibody binding, suggesting a mechanism of ligand-induced conformational changes during cell entry. Finally, the structural, biochemical, and cell-based functional analyses of HCMV Pentamer reported in this study provide an atomic-level framework for investigating the mechanism of Pentamer activity and for guiding antigen design efforts and will facilitate the interpretation of ongoing and future vaccine clinical trial data.


Wild-type Pentamer expression and purification

HCMV Pentamer (Merlin strain), with a tobacco etch virus (TEV) protease–cleavable double strep tag on UL130 (34), was stably expressed in human embryonic kidney (HEK) 293S GnTI cells. Expression medium was loaded directly onto a StrepTrap HP column (GE Healthcare), and the proteins were eluted with 2.5 mM desthiobiotin (IBA Lifesciences). The Strep tag was cleaved off by incubating with TEV protease (AcTEV, Thermo Fisher Scientific) overnight at 4°C. The sample was diluted with 20 mM Hepes buffer (pH 7.0) to lower total salt concentration to 50 mM and loaded onto Mono S 10/30 column (GE Healthcare). Pentamer was eluted on a salt gradient from 50 mM to 1.0 M NaCl.

Pentamer-Fab complex purification

Fab expression and purification have been previously reported (21). For complex formation, an excess of Fab was incubated with Pentamer (2:1 molar ratio) for at least 15 min at room temperature (RT). The resulting sample was concentrated and loaded onto Superose 6 10/300 column (GE Healthcare) to remove excess Fab. Peak fractions were pooled and concentrated to 6.0 to 7.0 mg/ml and were used in crystallization experiments.

Mutants and selenomethionine-substituted Pentamer and IgG expression and purification

Expression of Pentamer mutants was performed by transient transfection in Expi293 cells (Invitrogen) in shaker flasks following the manufacturer’s recommendations (Invitrogen). For selenomethionine (SeMet)–substituted Pentamer expression, 5 hours after adding enhancers, intracellular methionine was depleted by switching out the Expi medium with methionine-free Dulbecco’s modified Eagle’s medium (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS), 0.2 mM l-cysteine, 8 mM l-glutamine, 3 mM glucose, and kifunensine (1 mg/liter; starvation medium). After 16 hours, the medium was replaced with starvation medium supplemented with l-selenomethionine (60 mg/liter; Sigma-Aldrich). Cells were cultured for an additional 48 hours, and then the supernatant was harvested. SeMet-Pentamer and mutants were purified similarly to wild type (WT). WT and mutant immunoglobulin G (IgG) 8I21 were transiently expressed in 293Expi cells in a 12-well format.

Crystallization of Pentamer–8I21 Fab complex

Initial crystal hits for Pentamer–8I21 Fab complex appeared as small microcrystals in a drop containing 0.1 μl of protein and 0.1 μl of 20% ethanol and 0.1 M tris (pH 8.5) at 20°C. Ethanol was replaced with the less volatile isopropanol in subsequent experiments. Further additive screening (Hampton Research) resulted in benzamidine addition to crystallization conditions. The best crystals for WT or SeMet–Pentamer–8I21 Fab crystals were obtained from a reservoir solution containing 10% (w/v) PEG 400 (polyethylene glycol, molecular weight 400), 10% isopropanol, 2% (w/v) benzamidine, and 0.1 M tris (pH 8.2).

Crystallization of Pentamer–9I6 Fab complex

Crystal hits for Pentamer–9I6 Fab initially appeared in 0.1 M MES (pH 6.5) and 15% (w/v) PEG methyl ether 500 and optimized for growth with different additives. The best diffracting crystal was obtained from a solution containing 10% (w/v) PEG methyl ether 500, 0.1 M MES (pH 6.2), and 10 μM phenol.

Data collection

For Pentamer–8I21 Fab data collection, crystals were briefly soaked in their reservoir buffer with the addition of 10 to 30% ethylene glycol as cryoprotectant and then flash-cooled in liquid nitrogen. For Pentamer–9I6 Fab, crystals were soaked in reservoir buffer with sequential 15-25-50% (w/v) increase in PEG 400 before cryocooling. Diffraction data were collected on LS-CAT beamlines 21D and 21F at the APS (Argonne) and beamlines ID23-1, ID29, and ID30A-1 (MASSIF-1) at the ESRF and were processed with XDS (35). Crystals of the Pentamer–8I21 Fab complex belong to space group P212121, with a Matthews coefficient of 2.77 that corresponds to a solvent content of 55.7%, whereas crystals of the 9I6 Fab complex belong to space group I212121, with a Matthews coefficient of 2.54 and a solvent content of 51%. Both crystal forms accommodate one copy of the respective complexes in the asymmetric unit.

Pentamer–8I21 Fab structure determination and refinement

Initial phases for the Pentamer–8I21 Fab complex were obtained by the SAD method from a highly redundant 4.0 Å–resolution data set collected at the SeMet absorption-edge peak wavelength from a crystal of SeMet-Pentamer bound to the 8I21 Fab. With these data, Phenix.autosol (36) was able to locate 20 of 23 total SeMet sites, resulting in a figure of merit of 0.443 after phasing. Density modification and automatic model building with Phenix resulted in a model consisting of 768 residues (77 continuous fragments mostly belonging to gH and the 8I21 Fab) and a Rwork/Rfree of 42.8/45.6%. Refinement of this partial model was performed with Phenix.refine and was followed by extensive rebuilding with Coot (37) using the EBV gH/gL model and SeMet positions as guide. At this point, a 3.0 Å–resolution data set from non-isomorphous Pentamer–8I21 Fab crystals was collected, and the partial Pentamer model was placed in the new cell by MR using Phaser (38). Multicrystal averaging and phase extension with Phenix resulted in electron density maps of excellent quality that allowed building of the entire complex. Model refinement of the 3.0 Å–and 4.0 Å–resolution structures was performed with Buster (39). The target-structure restraints option (LSSR), with the coordinates of the 3.0 Å–resolution structure as target, was used to refine the 4.0 Å–resolution structure. Rwork/Rfree values of the 3.0 Å–and 4.0 Å–resolution Pentamer–8I21 Fab complexes converged to 19.0/23.7% and 25.9/27.6%, respectively. The final electron density map with the fitted model for the 3.0 Å structure is shown in fig. S10A.

Pentamer–9I6 Fab structure determination

The structure of Pentamer–9I6 Fab was determined at 5.9 Å resolution by MR with Phaser using gH/gL and ULs from the 3.0 Å–resolution Pentamer–8I21 Fab complex and the heavy chain and light chain from Protein Data Bank (PDB) entries 4zd3 and 3u0t, respectively, as search models. Refinement was performed with Buster using the LSSR option and Pentamer coordinates from the high-resolution Pentamer–8I21 Fab complex as fixed target, resulting in final Rwork/Rfree values of 25.6/30.0%. The final electron density map with the fitted model for this structure is shown in fig. S10B.

Surface plasmon resonance studies of Pentamer binding to WT and mutant 8I21 antibodies

Surface plasmon resonance (SPR) single-cycle kinetic experiments were carried out using a Biacore T200 instrument. Two adjacent channels on a CM5 sensor chip were immobilized with IgG binder using the Human Antibody Capture kit (GE Healthcare) according to the manufacturer’s recommendations. HBS-EP buffer (GE Healthcare) was used as diluent for both ligand and analyte samples. Ligand 8I21 IgG was captured on one channel (leaving the other channel as reference). For the 8I21 mutants, cell culture supernatant expressing the various mutants was flowed over the human IgG binder for the capture step. Ligand capture levels were around 200 resonance units (RUs) for the analysis of Pentamer UL130 mutants and roughly 40 RUs for the analysis of the 8I21 mutants. For both experiments, WT and mutant Pentamer at 0, 1.56, 3.125, 6.25, 12.5, and 25 nM were injected over the two channels at 50 μl/min for 120 s, followed by 600 s of dissociation time. The single-cycle kinetic curves were fitted with 1:1 binding stoichiometry for ka, kd, and KD using the Biacore T200 evaluation software.

HDX-MS analysis

For labeling, 315 pmol of Pentamer alone or in complex with a Fab was diluted at 0°C with ice-cold deuterated buffer (25 mM tris-HCl, 150 mM NaCl, pD 7.1), reaching a deuterium excess of 85.7%. At different time points, ranging from 30 s to 30 min, 30 μl of the sample was removed and quenched with the same volume of an ice-cold 7 M urea, 400 mM guanidinium chloride, 800 mM tris(2-carboxyethyl)phosphine, 0.1% formic acid, and pH 2.1 buffer to quench the deuterium exchange reaction and promote Fab dissociation. Quenched aliquots were flash-frozen in liquid nitrogen and stored at −80°C before analysis. Samples were analyzed with a Waters nanoACQUITY UPLC with HDX Technology coupled to Waters SYNAPT G2 mass spectrometer equipped with a standard electrospray ionization source. The generated data were interpreted using the DynamX software (Waters). Only the peptides present in at least three repeated digestions with pepsin of the unlabeled proteins were considered for the analysis.

Fitting of Pentamer x-ray structures into NS-EM maps

The x-ray crystal structures of Pentamer from the complexes with Fabs 8I21 and 9I6 were rigid body–fitted onto random conical tilt NS-EM reconstructions of Pentamer bound to neutralizing antibodies (21). Portions of the EM maps that remained empty after fitting of the Pentamer model were used for successive cycles of rigid-body fitting of single Fab molecules (pale pink and red cartoon in Fig. 4A). To fit the crystallographic model of Pentamer in the EM map of the Pentamer-3G16-10F7 complex, an arbitrary and manual rotation of the ULs was performed (red arrow, Fig. 4A, site 2). All rigid-body fittings were performed using the “fit in map” function in Chimera (40).

Pentamer cell binding

ARPE-19 and MRC5 cells were grown according to the American Type Culture Collection guidelines. HUVECs were grown in Vasculife EnGS medium (Lifeline Cell Technology). About 40,000 cells were seeded per well of a 96-well plate (Costar) and incubated at 37°C with 5% CO2 overnight. Titrations of the various protein complexes were diluted in binding medium (BM) (for ARPE-19 and MRC5 cells, this was the normal medium supplemented with 5% fetal bovine serum and 30 mM Hepes; for HUVECs, this was the normal medium supplemented with 30 mM Hepes) starting at a concentration of 0.1 μM. The medium was removed from the cells, 100 μl of BM was added to each well, and the cells were chilled to 4°C for 10 min. The BM was removed, and 50 μl of diluted protein was added before the cells were returned to 4°C for 1 hour. The cells were washed three times in 1× phosphate-buffered saline (PBS) (Hyclone) and fixed in 4% paraformaldehyde (USB). Protein binding was detected using an anti-His tag mouse antibody (Invitrogen) at a dilution of 1:500 in 1× PBS containing 2% FBS. The cells were washed three times in 1× PBS and then incubated with a peroxidase-conjugated goat anti-mouse antibody from Jackson ImmunoResearch Laboratories (1:1000). The cells were washed three times in 1× PBS, and the peroxidase activity was measured by the addition of 50 μl of 1-Step Ultra TMB ELISA reagent (Thermo Fisher Scientific). The reaction was stopped after 15 min by adding 50 μl of 2 N sulfuric acid (BDH). The absorbance at 450 nm was read on a plate reader using Softmax Pro. To test antibody inhibition of binding, the cells were set up as described above, but the protein complexes were diluted to a concentration of 0.025 μM in BM and preincubated with the antibodies (10 μg/ml) at RT for 1 hour before being added to the cells. Detection was carried out as described above. Titration curves for cell binding and antibody concentration for achieving 50% inhibition (IC50) values are shown in fig. S11.


Fig. S1. Comparison of HCMV gH/gL with related herpesvirus gH/gL.

Fig. S2. Surface charge distribution on Pentamer.

Fig. S3. Sequence conservation in gH and gL of HCMV and EBV.

Fig. S4. Sequence conservation in gH and gL among β-herpesviruses.

Fig. S5. Superposition of UL128 and ULl30 N-terminal domains with the most structurally similar chemokines, as identified by DALI.

Fig. S6. Cavities in the Pentamer structure.

Fig. S7. Pepsin digestion coverage of gH, gL, UL128, UL130, and UL131A.

Fig. S8. Mapping of neutralizing epitopes by HDX-MS.

Fig. S9. Regions of Pentamer stabilized by Fab binding.

Fig. S10. Electron density maps of Pentamer complexes with Fabs 8I21 and 9I6.

Fig. S11. Binding of soluble Pentamer to cell surfaces.

Table S1. Analysis of Pentamer subunit interfaces.

Table S2. Analysis of the Pentamer–8I21 Fab interface.

Table S3. Molecular determinants of Pentamer–8I21 Fab binding.

Table S4. Pentamer epitope mapping by HDX-MS.

Movie S1. Pentamer structure.

Movie S2. Pentamer is flexible and adopts different conformations.


Acknowledgments: We would like to thank R. Nolte and J. Brunzelle for access to and help with data collection at the APS and M. A. Luftig, B. Chen, M. J. Bottomley, and A.-M. Steff for critical review of the manuscript. This work was sponsored by GlaxoSmithKline Biologicals S.A. The sponsor was involved in all stages of the study conduct and analysis. Author contributions: Conceptualization: A.C.; methodology: S.C., E.M., T.V.N., K.L., D.D., and N.N.; investigation: S.C., E.M., T.V.N., K.L., D.D., Y.X., N.N., D.Y., and A.C.; writing (original draft): S.C., E.M., and A.C.; writing (review and editing): S.C., E.M., and A.C.; supervision: A.C., D.Y., and N.N. Competing interests: All authors are/were employees of the GSK group of companies at the time of the study. S.C., E.M., K.L., N.N., D.Y., and A.C. own shares in the GSK group of companies. S.C., E.M., and A.C. are inventors on U.S. patent applications for modified cytomegalovirus proteins and stabilized complexes (62/487,065 and 62/504,059). Data and materials availability: Coordinates and structure factors for the two Pentamer–8I21 Fab complexes (3.0 and 4.0 Å) and Pentamer–9I6 Fab complex (5.9 Å) have been deposited in the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB) under accession numbers 5VOB, 5VOC, and 5VOD, respectively.

Stay Connected to Science Immunology

Navigate This Article