Structure of the Divergent Human Astrovirus MLB Capsid Spike

Summary

Despite their worldwide prevalence and association with human disease, the molecular bases of human astrovirus (HAstV) infection and evolution remain poorly characterized. Here, we report the structure of the capsid protein spike of the divergent HAstV MLB-clade (HAstV MLB). While the structure shares a similar folding topology with that of classical-clade HAstV spikes, it is otherwise strikingly different. We find no evidence of a conserved receptor-binding site between the MLB and classical HAstV spikes, suggesting that MLB and classical HAstVs utilize different receptors for host cell attachment. We provide evidence for this hypothesis using a novel HAstV infection competition assay. Comparisons of the HAstV MLB spike structure to structures predicted from its sequence reveal poor matches, but template-based predictions were surprisingly accurate relative to machine-learning-based predictions. Our data provide a foundation for understanding the mechanisms of infection by diverse HAstVs and can support structure determination in similarly unstudied systems.

Graphical Abstract

eTOC Blurb

Delgado-Cunningham et al. report the structure of the capsid spike from human astrovirus strain MLB1. The capsid spike forms protrusions on the surface of the virus and may have a role in attaching to human cells. The MLB1 spike structure has many differences compared to classical human astrovirus spike structures.

Introduction

Astroviruses are a family of nonenveloped, positive-sense RNA viruses which infect a variety of avian and mammalian species. Three astrovirus clades — classical, VA, and MLB — are known to infect humans (Fig. 1A).¹ Human astroviruses (HAstVs) from all three clades are believed to replicate in the intestinal tract and may cause both gastrointestinal and extra-gastrointestinal disease.^2,3 Classical HAstVs infect the intestinal epithelium ⁴ and are a well-established cause of diarrhea, particularly in pediatric patients.^3,5 Infection by HAstV VA1 may cause diarrhea ^1,6–8 and severe inflammation in the central nervous systems of immunocompromised patients.^9,10 HAstV MLBs (including MLB1, MLB2, and MLB3) are found in stool samples worldwide ^11–14 and seroprevalence to HAstV MLB capsid is nearly ubiquitous in humans.¹⁵ However, case-control experiments testing the link between HAstV MLB infection and gastroenteritis have yielded conflicting results. HAstV MLB and/or HAstV MLB1 infection was not associated with gastroenteritis in two studies ^12,14 whereas HAstV MLB1 infection was associated with gastroenteritis in another study.¹³ Several case studies link HAstV MLB to viremia ^16–23 and disease in the central nervous system.^18,21,24 Overall, a conclusive link of HAstV MLB to disease has yet to be established.

Figure 1: Delineation and structure of the HAstV MLB spike domain.

(A) The human astroviruses. Complete capsid protein sequences were aligned using the MUSCLE algorithm. Evolutionary analyses in MEGA-X using the Maximum Likelihood method and JTT matrix-based model yielded the cladogram shown (log likelihood −21535.53). Bootstrap values are shown next to the branches and were computed using 1000 replicates. The tree was rooted using the turkey astrovirus 1-3 capsid sequences as an outgroup. (B) The domain structure of the astrovirus capsid protein. Pairwise sequence identities in the core and spike domains between and within the HAstV clades (classical, MLB, VA) are included. The diagram is drawn so that domains are arranged from N-terminal (basic) to C-terminal (acidic). (C) Size-exclusion chromatography trace of recombinant HAstV MLB1 spike (cyan line) overlaid with the trace of gel filtration standards (black dotted line). (D) Crystal structure of the MLB1 spike dimer.

As HAstV MLBs are found in stool samples and can be cultured in human intestinal enteroids ²⁵, they are believed to initially infect the intestinal epithelium in vivo. Diverse astroviruses have been shown to infect enterocytes, including classical ⁴, VA1 ²⁵, ovine ²⁶, murine ^27,28, bovine ²⁹, and turkey astroviruses.^30–32 Interestingly, all 8 classical HAstV serotypes (HAstV 1-8) and HAstV VA1 can be cultured in the human colonic cancer cell line Caco-2 ^33–35, but the only published effort to grow HAstV MLB1 in Caco-2 cells was unsuccessful.³⁶ Instead, HAstV MLBs could be propagated in two human non-intestinal cell lines Huh-7 and A549, however these cells are only partially permissive, with infection observed in ~10% of cells.³⁶ The molecular drivers of these observed differences in cell tropism are unknown, as are their contributions in human disease.

The HAstV capsid is a T=3 icosahedron made up of 180 copies of the capsid protein.³⁷ The capsid basic and core domains encase the viral genome, while spike domain dimers project outward from the surface ^38–40 and have been shown to mediate cell attachment.^41,42 Classical HAstVs then enter Caco-2 cells (and potentially host cells in vivo) through clathrin-mediated endocytosis.⁴³ Before viral release from host cells, classical HAstV capsid protein undergoes intracellular proteolysis by host caspases to remove a C-terminal acidic domain.^44,45 Upon viral release, further digestion by trypsin in vitro yields an infectious particle containing 30 well-ordered spike dimers on the icosahedral 2-fold axes.^37–39 Extracellular proteolysis using trypsin is required for propagation of classical but not VA or MLB HAstVs in cell culture.^34,36,46 Mature HAstV VA1 is instead produced through intracellular, caspase-independent cleavage by one or more unknown proteases.⁴⁷ The proteolytic maturation of MLB human astroviruses is currently uncharacterized.

While no astrovirus receptors have been identified to date, structural studies of classical HAstV spikes have yielded sites on the spike surface that may function in receptor binding.^40–42 Of these, the P site is of particular interest. This patch of conserved residues is located at the top of the spike, between the epitopes of antibodies known to block viral attachment.^40,41 The hydrophilicity of the P site, including several hydrogen-bond donors and acceptors, supported the hypothesis that HAstV may attach to host cells by binding a carbohydrate.⁴⁰ While heparin, dextran sulfate, heparan sulfate and chondroitinase treatment of host cells partially inhibit HAstV infection ^40,48, whether HAstVs use or require carbohydrates for infection in vivo is unknown. Crucially, our current understanding of astrovirus cell entry has been built using primarily classical HAstVs. The molecular basis for entry by HAstV VA and MLB remains wholly unstudied.

Here, we delineated the HAstV MLB capsid spike domain and solved its structure as a first step towards understanding the molecular basis for cell entry by HAstV MLB. While the HAstV MLB spike shares a folding topology with the spike domain of classical HAstV, it is otherwise strikingly different. None of the putative receptor-binding sites from classical HAstV spikes were observed on the surface of the MLB spike, supporting the hypothesis that the MLB and classical HAstVs require different host factors for cell entry. To test this, we developed a novel in vitro HAstV infection competition assay and provide evidence that HAstV MLB utilizes a different receptor for cell infection than classical and VA HAstVs. Our data lay a foundation for elucidating the cell entry mechanisms used by divergent HAstVs.

Results

Delineation of the MLB spike domain

To delineate the boundaries of the HAstV MLB capsid spike domain, we evaluated the capsid protein sequences of classical, MLB, and VA1 HAstVs (Fig. 1A). Sufficient identity between the HAstV capsid core domain sequences (37-43%), along with the classical HAstV core domain structures, allowed the delineation of the core domain boundaries (Fig. 1B).^38,39 However, the HAstV MLB spike domain boundaries could not be precisely mapped by comparison to classical HAstV spike sequences and structures due to low sequence identity (22-26%) (Fig. 1B)^38,40,42,49 Thus, we tested several recombinant HAstV MLB1 capsid spike domain constructs and identified a construct containing residues 399-653 of the HAstV MLB1 capsid protein (30 kD) that could be expressed in Sf9 insect cells and E. coli and remains soluble after affinity purification. Analysis of this construct by size exclusion chromatography revealed an elution volume that is consistent with a homodimer (~ 60 kD), as observed for other astrovirus capsid spike domains (Fig. 1C).^40,48

Structure of MLB spike and comparison to the classical HAstV spike

To gain insights into the function of the HAstV MLB1 spike, we used X-ray crystallography to determine its high-resolution structure. The HAstV MLB1 spike crystallized readily and yielded X-ray diffraction data to a resolution of 1.86 Å (Table 1), but we were unable to solve its structure by molecular replacement with the classical HAstV spike structure nor predicted models (discussed further below). Instead, we generated selenomethionine-substituted HAstV MLB1 spike and used single-wavelength anomalous diffraction to solve the structure (Table 1). Structure determination revealed three HAstV MLB1 spike chains (A-C) in the asymmetric unit, with pairwise root-mean-square-deviations (RMSDs) between 0.18 and 0.34 Å. Each HAstV MLB1 spike chain is comprised of ordered residues 420-645, except for two or three disordered loops (discussed below). Each HAstV MLB1 spike chain forms a homodimer through crystallographic symmetry, consistent with size-exclusion chromatography data (Fig. 1D).

Table 1:

Crystallographic statistics. ^{a, b, c}

Parameter	MLB1 spike	Se-Met MLB1 spike
PDB Code	7UZT
Data Collection
Wavelength (Å)	1.00004	0.97942
Space group	P 1 2 1	P 1 2 1
Cell dimensions
a, b, c (Å)	72.10, 77.11, 75.44	72.15, 77.07, 75.44
α, β, γ (°)	90.00, 110.71, 90.00	90.00, 110.75, 90.00
Resolution (Å)	47.66-1.86 (1.90-1.86)	47.67-2.53 (2.63-2.53)
Rmerge	0.082 (0.433)	0.425 (4.975)
I/σI	8.7 (2.7)	14.7 (5.2)
Completeness (%)	98.1 (94.2)	99.8 (98.2)
Redundancy	3.4 (3.3)	27.1 (26.2)
CC1/2	0.995 (0.884)	0.995 (0.612)
No. of Selenium sites		27
Overall figure of merit		0.382
Refinement
No. of reflections	63794
Resolution (Å)	47.66-1.86
Rwork/Rfree	0.242/0.260
No. of atoms	5271
Protein	5040
Ligands	0
Water	231
B factors (Å2)	22.50
Protein (Å2)	22.34
Ligands (Å2)	N/A
Water (Å2)	26.08
RMSD
Bond lengths (Å)	0.011
Bond angles (°)	1.449
Ramachandran statistics
Favored (%)	96.28
Allowed (%)	3.23
Outliers (%)	0.48

^aData from one crystal were used to solve each structure. Statistics in parentheses correspond to the highest-resolution shell in each dataset.

^bSe-Met MLB1 spike: selenomethionine-substituted MLB1 spike

^cRMSD: root-mean-square deviation.

The HAstV MLB1 spike homodimer has ~3100 Ų buried at the dimer interface (Fig. 2A) similar to the classical spike, which has a 3500-3800 Ų dimer interface.^38,40,42,49 The structural scaffold of each protomer is anchored by an antiparallel β-barrel formed by strands 1, 5-8, 11, and 15 (Fig. 2B). This core structural motif is stabilized by interactions with β-hairpin motifs (β3-4 and β12-13), an antiparallel β-sheet running across the top of the spike (β2, β9-10, and β14), a short α-helix and several loops that connect secondary structure elements (Fig. 2B). Interestingly, electron density for residues 441-444 and 555-562 of all three chains and 609-612 of chain A was missing in our maps, indicating that these sections of the loops connecting strand 1 to strand 2, strand 9 to strand 10, and strand 13 to strand 14 may be flexible in solution (Fig. 2B). The corresponding loop residues are ordered in the classical HAstV spike structures published to date, and in particular the large loops crossing the top of the classical HAstV spike have been implicated in spike dimerization and reactivity with neutralizing antibodies.^40,50 The biological effects of these differences in HAstV MLB1 spike are currently unknown.

Figure 2: The classical and MLB spikes are structurally distinct but share a conserved folding topology.

(A) Structures of the MLB and classical HAstV spike dimers and the results of their structural alignment. The individual spike protomers were aligned structurally, in a sequence-independent manner, using the PDBeFold server.⁶⁵ The HAstV 2 strain Oxford spike was used for the alignment and analyses in panels 2A and 2B (PDB: 5W1N).⁴⁹ The HAstV 8 spike was used for alignment in panel 2A (PDB: 3QSQ).⁴⁰ (B) Topology representations of the MLB and classical HAstV spike domains. The core β-barrel conserved in hepatitis E capsid P2 is colored as in ⁴⁰, except that the most C-terminal β-strand is colored magenta instead of red. Other β-strands conserved between MLB and other known astrovirus spike domains are colored grey. (C) Structures of the MLB and classical HAstV spike domains, the turkey astrovirus (TAstV 2) spike domain, and the hepatitis E (Hep E) P2 domain, with the core β-barrel strands colored as in (B). Flexible residues that were not included in each model are drawn approximately to scale as dashed lines.

Despite low sequence identities, the folding topologies of the MLB and classical HAstV spikes are remarkably similar (Figs. 1B, 2B). The seven-stranded β-barrel at the core of the HAstV MLB1 spike is conserved with the classical HAstV spike, with the turkey astrovirus spike⁴⁸, and with the P2 domain of the hepatitis E virus capsid protein (Fig. 2C).⁴⁰ Additionally, the MLB and classical spikes share a structural element, the β-hairpin motif (β3-4) (Fig. 2B), which is also found in the turkey astrovirus spike ⁴⁸, suggesting that this structural motif may be universal in astrovirus spike domains. Most secondary structure elements, including loops, on the MLB1 spike align to an equivalent secondary structure element on the classical spike domain. However, there are notable differences in the lengths, orientations, and backbone conformations of structural elements between the two spikes. The result is a striking difference in global conformation between the MLB and classical spikes, resulting in a 3.16 Å RMSD upon alignment (Fig 2A). These observations provide a structural basis for the hypothesis that MLB human astroviruses are related to — and phylogenetically distinct from — the classical human astroviruses.

MLB, classical, and VA astroviruses may use different receptors for cell entry

As the classical HAstV capsid spikes are implicated in viral attachment to host cells, we were interested in determining whether HAstV MLBs may use the same host cell receptor as classical and/or VA HAstVs. Receptor-binding residues have been found to be conserved in other diverse virus families.^51,52 Thus, we searched for surface-exposed patches of conserved amino acids between the classical and MLB spikes (HAstV MLB and VA spike structures cannot currently be compared as no structures of VA spikes have been published). Structural alignment of the MLB1 spike to structures of the classical human astrovirus serotypes 1, 2, and 8 spike domains revealed a striking lack of conserved residues which might form a conserved receptor-binding site (Fig. 3A). Importantly, we see no conservation of the putative receptor-binding residues within the P site identified previously ⁴⁰ in a structural alignment of the HAstV MLB1 and HAstV 8 spikes (Fig. 3B). Together with published data indicating that classical HAstV and HAstV MLB may differ both in their association with diarrhea ^12,14 and their ability to propagate in Caco-2 cells ³⁶ our observations support a hypothesis that MLB and classical HAstV bind different receptors.

Figure 3: HAstV MLB may use a different receptor than the classical and VA1 HAstVs.

(A) Sequence conservation mapped onto the MLB1 spike following structural alignment with the HAstV 8 (PDB: 3QSQ) ⁴⁰, HAstV 1 (PDB: 5EWO) ³⁸, and HAstV 2 spikes (PDB: 5W1N).⁴⁹ Identical residues that aligned structurally were colored red (for 3/3 alignments), magenta (for 2/3 alignments), or purple (for 1/3 alignments). The MLB1 spike is shown as a solvent-accessible surface. Spike structures were aligned using the PDBeFold server.⁶⁵ (B) Comparison of the MLB1 and HAstV 8 spike amino acids in the P site. The putative receptor-binding residues in the P site ⁴⁰ are shown as sticks, as are structurally-aligned HAstV MLB1 residues. (C) In vitro HAstV infection competition assay. Caco-2 cells were infected with HAstV in the presence of the indicated concentration of recombinant spike proteins. The MLB1 spike does not block infectivity of either HAstV 8 or HAstV VA1. The data represent the infectivity of cells infected in the presence of each recombinant spike as compared to control cells infected in its absence. The arithmetic means ± SEM from three independent experiments performed in duplicate are shown. ** P < 0.01. (D) Conservation of MLB spike residues. All complete MLB spike sequences available in the NCBI nr database were aligned and each position was assigned to a Clustal conservation group using UCSF Chimera.⁶⁶ Each residue in the MLB1 spike structure was then colored according to its conservation group: complete identity (red), strong conservation (white), weak conservation (light blue), and minimal conservation (dark blue).⁶⁷

We tested this hypothesis by developing an in vitro HAstV infection competition assay and evaluating if recombinant MLB1 spike could inhibit infection by classical and VA HAstVs, with the rationale that if all HAstVs used the same receptor, then MLB1 spike would compete for receptors and inhibit infection. Caco-2 cells were incubated with virus on ice to allow for attachment but not endocytosis. Cells were washed, incubated with recombinant spike protein on ice, and then incubated at 37 °C to allow the viral life cycle to proceed. Infected cells were quantified using immunoperoxidase detection of virus capsid proteins. We find that classical HAstV 1 spike inhibits classical HAstV 8 infection in a dose-dependent manner, likely by competing for receptors at the cell surface (Fig. 3C). In contrast, classical HAstV 1 spike had no discernable effect on HAstV VA1 infection at the concentrations tested (Fig. 3C). Importantly, HAstV MLB1 spike did not inhibit infection by HAstV 8 or HAstV VA1 (Fig. 3C). Unfortunately, not all potential experiments testing matching or unmatching spikes with viruses could be performed because HAstV MLB does not infect Caco-2 cells and because recombinant HAstV VA1 spike has not yet been produced. Nevertheless, these data provide evidence that two classical HAstV serotypes (HAstV 8 and HAstV 1) bind a common receptor, since a spike from one classical HAstV serotype inhibits infection by another. These data also support the possibility that HAstV MLB utilizes a different receptor than both classical and VA HAstVs, since MLB1 spike did not inhibit infection by either virus.

To identify putative receptor-binding residues on the MLB1 spike, we aligned MLB1-3 spike sequences and mapped conservation onto the MLB1 spike structure (Fig. 3D). Several regions of high identity or similarity were observed, making it difficult to identify candidate receptor binding site residues. The high degree of sequence conservation between MLB1-3 spikes, as well as the limited number of sequences currently available, contribute to this challenge (Fig. 1B).

Retrospective structural predictions of MLB1 spike

Finally, given the recent advancements in protein structure prediction ^53,54, we sought to retrospectively ascertain whether the MLB1 spike structure could have been predicted. We first used HHpred with Modeller ⁵⁵, as their server turnaround time is much faster than other servers. ⁵⁶ While HHpred does not include the deep-learning methods used by other structure-prediction programs, in our specific case, HHpred with Modeller outperformed some of the methods that use deep-learning, such as C-Quark ⁵⁷ and both of the AlphaFold2 methods tested ^53,58 (Table 2). The I-TASSER server ⁵⁹ – which is from the same lab as C-Quark but does not implement deep-learning methods – was the closest to the MLB1 spike structure by RMSD (4.19 Å) and TM-score (0.68). RoseTTAFold ⁵⁴ — the deep-learning version of Rosetta — produced the closest predictions by all-atom RMSD and LDDT ⁶⁰ when the MLB spike was predicted as a monomer (0.7516 Å RMSD, 0.4211 LDDT) or a dimer (7.069 Å RMSD, 0.4278 LDDT) (Table 2). While AlphaFold2 is now able to run monomers and protein complexes, the closest it was able to get to the MLB1 spike structure using oligomers was over 11 Angstroms RMSD — closer than the monomeric AlphaFold2 predictions (Table 2). Unsurprisingly, our efforts to solve the MLB1 spike structure by molecular replacement with these predicted models were unsuccessful.

Table 2:

Comparison of predicted and experimentally-determined MLB1 spike structures ^{a, b, c, d, e}

Model comes from:	RMSD	all-atom RMSD	TM-score	LDDT
I-TASSER	4.913	10.399	0.67784	0.357
RoseTTAFold Multimer	5.006	7.467	0.64681	0.4278	model 1
	5.725	7.069	0.6161	0.404	model 3
RoseTTAFold	5.271	7.516	0.63261	0.4211
HHpred/Modeller	7.098	9.898	0.65687	0.3275
AlphaFold2 Multimer	11.187	12.886	0.43945	0.1159
AlphaFold2 48 cycles	12.496	18.725	0.42833	0.2703
AlphaFold2 12 cycles	15.93	21.208	0.41335	0.236
C-Quark	16.388	19.328	0.29973	0.2215
RoseTTAFold CASP15	1.451	3.937	0.88319	0.6689
AlphaFold2 CASP15	1.227	7.377	0.82319	0.6011

^aThe best values in each category are shown in bold font, excluding CASP15 models.

^bIDDT: The local Distance Difference Test is a superposition-free score that evaluates local distance differences of all atoms in a model, including validation of stereochemical plausibility.⁶⁰

^cTM-score: Template modeling score weights smaller distance errors stronger than larger distance errors and makes the score value more sensitive to the global fold similarity than to the local structural variations.⁶⁸

^dRMSD: root-mean-square deviation between Cα atoms, in angstroms.

^eall-atom RMSD: root-mean-square deviation between all atoms, in angstroms.

However, during the course of peer review of this article, we submitted the MLB1 spike sequence as a target for structure prediction in the 15th Community Wide Experiment on the Critical Assessment of Techniques for Protein Structure Prediction (CASP15). Notably, the structures predicted by the newest versions of AlphaFold2 and RoseTTAFold and posted online during CASP15 were significantly improved (Table 2). In fact, we found that either of these structures could have been used successfully for molecular replacement to solve the MLB1 spike structure.

Discussion

Here, we report the first structure of a divergent HAstV spike domain, the HAstV MLB1 spike. We find that the classical and MLB HAstV spikes share a common folding topology despite low sequence identities. This raises the possibility that all HAstV spikes share a conserved fold, although structures of VA-clade spikes will be needed to determine if this is the case. Outside of its folding topology, the HAstV MLB spike is strikingly different from the classical spike structures published to date. Crucially, there are no residues on the surface of the MLB spike which might clearly form a receptor-binding site that is conserved with the classical HAstVs. This suggests that the two viruses use different receptors for host-cell infection but could also be explained if MLB and classical HAstVs bind a shared receptor through different residues.

To address this ambiguity, we developed a novel in vitro HAstV infection competition assay and tested cross-clade inhibition of infection by recombinant HAstV spikes. We provide the first direct evidence that HAstV MLB uses a different receptor than both HAstV VA1 and the classical HAstVs. This is consistent with published data showing that HAstV VA1 and all 8 classical HAstV serotypes — but not HAstV MLB1 — grow in Caco-2 cells.^33–36 Our data raise the possibility that HAstV MLB has a very different cell tropism in vivo than the classical or VA1 HAstVs. This would help to explain why MLB infection has not been clearly associated with diarrhea to date.^12,14 We unfortunately could not test for inhibition of HAstV MLB infection due to the low efficiency of MLB infection in the cell lines used to date.³⁶ This leaves open the possibility that our MLB1 spike construct was not sufficient for receptor binding. We believe this to be unlikely as recombinant classical HAstV spike, also produced in E. coli, is sufficient for cell attachment ^41,42 and for inhibition of infection, as demonstrated here. Further studies are needed to confirm our findings definitively and to test their implications for cell tropism.

Having delineated the HAstV MLB spike and solved its structure also opens the door to addressing central questions in the field through experiments that were previously infeasible. The seropositivity of HAstV MLB2 and MLB3 has not been quantified to date and could not be evaluated in a serosurveillance study of HAstV MLB due to antibody cross-reactivity.¹⁵ Using recombinant HAstV MLB1-3 spikes as antigens in seroepidemiology studies, as was performed with classical HAstV spikes ⁶¹, should minimize cross-reactivity and may allow for the extent of HAstV MLB2 and MLB3 infection to be better understood. In a similar vein, recombinant HAstV MLB spike could be tested as an immunogen for its ability to elicit HAstV MLB-neutralizing antibodies, as has been shown for classical HAstV spikes. ⁵⁰ Recombinant HAstV MLB spike could be used as a bait to identify HAstV MLB receptors in co-purification experiments.⁶² Finally, reverse genetics could be used to generate chimeric HAstVs with MLB spike domains for infection studies in intestinal enteroids ²⁵ to investigate the extent to which the spike drives cell tropism.

Given the remarkable advancements in protein structure prediction including deep-learning-based approaches ^53,54, we were surprised to find that predictions of the HAstV MLB spike structure using publicly available prediction programs were poor matches to the experimentally-determined structure. One reason, particularly for methods which predicted the MLB1 spike as a monomer, could be that dimerization influences the structure of the individual protein protomers. Methods using evolutionary reasoning and contact prediction may have been especially affected by a lack of sequence information; MLB spike sequences have pairwise identities from 64-84 % (Fig. 1B) and only 27 sequences align to our MLB1 spike sequence in BLASTP searches of the NCBI nr database. Alignment of the MLB spike sequence to classical HAstV spike templates during AlphaFold2 and RoseTTAfold predictions may also have been hindered by the low sequence identities between the MLB and classical spikes. The relatively accurate models we obtained using I-TASSER and HHpred/Modeller may have come in part because these approaches use predicted secondary structure, in addition to sequence information, for template alignment.^63,64 We believe that our findings may be useful both to our colleagues developing structure prediction methods and to other structural biologists working with uncharacterized, oligomeric proteins for which limited sequence information is available. Remarkably, AlphaFold2 and RoseTTAfold predictions of the MLB1 spike generated for the CASP15 competition were very close to the experimental structure, so much so that the models could have been used for molecular replacement. While the latest versions of these programs are not yet publicly available, these results preview the advancements that protein structure prediction programs have made in the past year.

Overall, this study yields new insights into host-cell entry by HAstV MLB and protein structure prediction in understudied systems. Our data provide a starting point for elucidating the entry mechanisms used by nonclassical human astroviruses and indicate that doing so may be important for understanding human astrovirus-associated diseases.

STAR Methods

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Rebecca DuBois (rmdubois@ucsc.edu).

Materials availability

Plasmids produced in this study are available from the lead contact.

Data and code availability

Coordinates and structure factors for the MLB1 spike structure have been deposited in the Protein Data Bank and are publicly available as of the date of publication with the accession code 7UZT. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell lines

Recombinant MLB1 spike protein was purified from the expression in insect cells, Spodoptera frugiperda Sf9 cell line IPLB-Sf-21-AE (Expression Systems), or in bacterial cells, E. coli strain T7 Express (New England Biolabs) or E. coli strain B843 (DE3) pLysS (Novagen). C2Bbe1 cells, derived from the colon adenocarcinoma Caco-2 cell line, were obtained from the American Type Culture Collection, and propagated in a 10% CO₂ atmosphere at 37°C in Dulbecco’s modified Eagle’s medium (DMEM)-reduced serum (Thermo Scientific HyClone, Logan, UT) supplemented with 5% fetal bovine serum (FBS) (Cansera, Ontario, Canada).

Viruses

HAstV 8 strain Yuc8 was isolated in our laboratory, and it was grown as described previously ⁴⁴. HAstV VA1 was obtained from David Wang’s laboratory ; a virus stock was produced in C2Bbe1 cells infecting at a MOI of 0.1, and collecting the virus at 4 days post-infection by lysing the cells with three of freeze-thaw cycles.

METHOD DETAILS

Sequences used for phylogenetic analysis

The cladogram in figure 1A was constructed using complete capsid protein sequences with the following accession numbers: Human astrovirus 1: AAC34717.1 (NCBI), Human astrovirus 2: AAA62427.1 (NCBI), Human astrovirus 3: Q9WFZ0 (uniprot), Human astrovirus 4: Q3ZN05 (uniprot), Human astrovirus 5: Q4TWH7 (uniprot), Human astrovirus 6: AZB52207.1 (NCBI), Human astrovirus 7: Q96818 (uniprot), Human astrovirus 8: AAF85964.1 (uniprot), Human astrovirus MLB1: YP_002290968.1 (NCBI), Human astrovirus MLB2: YP_004934010.1 (NCBI), Human astrovirus MLB3: YP_006905854.1 (NCBI), Human astrovirus VA1: ADH93577.1 (NCBI), Human astrovirus VA2: ACX83591.2 (NCBI), Human astrovirus VA3: YP_006905860.1 (NCBI), Human astrovirus VA4: YP_006905857.1 (NCBI), Human astrovirus VA5: AJI44022.1 (NCBI), Turkey astrovirus 1: NP_059949.1 (NCBI), Turkey astrovirus 2: NP_987088.1 (NCBI), Turkey astrovirus 3: AAV37187.1 (NCBI).

Sequences used for pairwise identity analysis

Pairwise identities of the following astrovirus spike and core domain sequences were used in figure 1B: Human astrovirus 1: residues 431-644 (spike) and 80-412 (core), Human astrovirus 2: residues 429-644 (spike) and 77-411 (core), Human astrovirus 3: residues 432-645 (spike) and 79-414 (core), Human astrovirus 4: residues 430-644 (spike) and 77-412 (core), Human astrovirus 5: residues 429-641 (spike) and 76-411 (core), Human astrovirus 6: residues 430-642 (spike) and 77-412 (core), Human astrovirus 7: 431-644 (spike) and 78-413 (core), Human astrovirus 8: residues 490-705 (spike) and 137-472 (core), Human astrovirus MLB1: residues 420-646 (spike) and 62-397 (core), Human astrovirus MLB2: residues 417-643 (spike) and 62-395 (core), Human astrovirus MLB3: residues 417-643 (spike) and 62-395 (core), Human astrovirus VA1: residues 394-697 (spike) and 69-390 (core), Human astrovirus VA2: residues 404-688 (spike) and 69-392 (core), Human astrovirus VA3: residues 388-691 (spike) and 63-384 (core), Human astrovirus VA4: residues 408-685 (spike) and 69-392 (core), Human astrovirus VA5: residues 406-678 (spike) and 68-391 (core). Accession numbers match those used for phylogenetic analysis.

Production of MLB1 spike in insect cells

A codon-optimized synthetic gene encoding residues 399-653 of the MLB1 capsid protein (NCBI YP_002290968.1) was cloned into the plasmid pBacPAK8 in frame with an N-terminal 10-histidine tag. A recombinant baculovirus stock was generated using the flashBAC system (Mirus bio). Sf9 insect cells (Expression Systems) in ESF-921 media were infected at a density of 2 million viable cells/mL with 0.025 mL of baculovirus stock/mL and cultured at 180 rpm, 27°C. Cells were harvested by centrifugation at 4 days post-infection, re-suspended in buffer A (20 mM Tris-HCl pH 8.0, 300 mM NaCl, 20 mM imidazole, 2 mM MgCl₂, protease inhibitor cocktail (Millipore), and benzonase), and lysed via microfluidization. The lysate was clarified by centrifugation (40,000 x g), 0.22-um-filtered, and MLB1 spike was purified from the supernatant using TALON metal affinity chromatography. Fractions containing the MLB1 spike were dialyzed into 20 mM Tris-HCl pH 8.0, 20 mM NaCl and further purified using anion exchange chromatography. Fractions containing the MLB1 spike were dialyzed into 10 mM Tris-HCl pH 8.0, 150 mM NaCl, concentrated to ~ 5 mg/mL, flash frozen in liquid nitrogen, and stored indefinitely at −80°C.

Production of MLB1 spike in E. coli

The MLB1 spike gene was cloned into the pRSFDuet plasmid in frame with an N-terminal 10-histidine tag, an alanine-serine linker, and a thrombin cleavage site. The plasmid was transformed into E. coli strain T7 Express (New England Biolabs) and grown in LB/kanamycin, and expression was induced with 1 mM IPTG at 18°C for 16 hours. For selenomethionine-substituted protein, the plasmid was transformed into E. coli strain B843 (DE3) pLysS (Novagen) and grown in M9 minimal medium supplemented with kanamycin, chloramphenicol, and 0.04 mg/mL selenomethionine. Cells were harvested by centrifugation, re-suspended in buffer A, and lysed using ultrasonication. The lysate was clarified by centrifugation (40,000 x g), 0.22-um-filtered, and MLB1 spike was purified from the supernatant using TALON metal affinity chromatography. For crystallography, fractions containing selenomethionine-substituted MLB1 spike were dialyzed into 10 mM Tris-HCl pH 8.0, 150 mM NaCl, and concentrated and frozen as described above. For HAstV-inhibition experiments, fractions containing the MLB1 spike were dialyzed into phosphate-buffered saline pH 7.65, concentrated to ~ 1 mg/mL, and frozen as described above.

Size exclusion chromatography analysis of MLB1 spike

Purified MLB1 spike was analyzed by size exclusion chromatography with a Superdex 200 increase 10/300 column. The oligomeric state of MLB1 spike was estimated by comparing its retention volume to those of the proteins in Gel Filtration Standards (Bio-Rad).

Crystallization, data collection, and structure determination

Native MLB1 spike at 5 mg/mL was crystallized in 100 mM Bis-Tris HCl pH 5.5, 350 mM MgCl₂, 22.5% w/v PEG 3350 using the hanging drop method. Selenomethionine-substituted MLB1 spike at 5 mg/mL was crystallized in 100 mM Bis-Tris HCl pH 5.5, 300 mM MgCl₂, 21% w/v PEG 3350 using the hanging drop method. Native spike crystals were cryoprotected using 100 mM Bis-Tris HCl pH 5.5, 300 mM MgCl₂, 25% w/v PEG 3350, and 18% v/v EDG (1:1:1 ethylene glycol, DMSO, and glycerol). Selenomethionine-substituted spike crystals were cryoprotected using a solution which instead had 200 mM MgCl₂ but was otherwise identical. All crystals were flash-frozen in liquid nitrogen. Data were collected at cryogenic temperatures at ALS beamline 5.0.2 and were processed using XDS.⁷⁰ The MLB1 spike structure was solved using SAD and an initial model was autobuilt using PHENIX AutoSol. ^71,72 This model was then used to phase a higher-resolution native dataset using molecular replacement with PHASER.⁷³ The final model was refined and manually built using PHENIX and COOT ⁷⁴, respectively. Coordinates and structure factors for the MLB1 spike structure have been deposited in the Protein Data Bank with the accession code 7UZT.

Purification of HAstV 1 spike

The spike domain of classical human astrovirus serotype 1 was produced in E. coli and purified as described previously.³⁸ Briefly, cDNA encoding HAstV-1 capsid protein residues 429 to 645 (Accession# AAC34717.1) was cloned into pET52b in frame with a C-terminal thrombin cleavage site and 10-histidine purification tag. The plasmid was transformed into E. Coli strain BL21 (DE3) and expression was induced with 1mM IPTG at 18°C for 16 h. Cell pellets were resuspended in Buffer A and lysed by ultrasonication. The lysate was clarified by centrifugation (40,000 x g), 0.22-um-filtered, and HAstV-1 spike was purified HisTrap metal-affinity chromatography. Fractions containing HAstV-1 spike were dialyzed into phosphate-buffered saline pH 7.65, concentrated to ~ 1 mg/mL, and flash frozen as described above.

HAstV infection competition assay with recombinant HAstV spike proteins

C2Bbe1 cells were grown in 96-well plates until confluence, the growth medium was removed and replaced by MEM pre-cooled to 4°C, and the cells were incubated for 20 min on ice. After this time the MEM was replaced by virus diluted in MEM at an MOI of 0.02 and incubated on ice for 1 h; the unbound virus was washed away with MEM and the indicated concentration of the purified HAstV spike proteins, diluted in MEM, was added and incubated for 1 h on ice. The plates were then transferred to 37°C for 1 h, washed with MEM to remove the proteins, and then were incubated in DMEM supplemented with non-essential amino acids (NEAA) for 18 h for HAstV 8 (Yuc8), or for 48 h for HAstV VA1. After this time, the cells were processed by an immunoperoxidase assay to detect the infected cells, as previously described ⁷⁵, with some modifications. Briefly, the cells were fixed at room temperature for 20 minutes with 2% formaldehyde diluted in PBS and permeabilized with 0.2% Triton X-100-PBS for 15 min. To stain the cells infected with HAstV 8, a rabbit polyclonal serum to HAstV 8 (Yuc8) was used ⁴⁴, and for cells infected with HAstV VA1 we employed the monoclonal antibody MAb 2A2.⁶⁹ Three independent experiments were performed in duplicate.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis

Statistical significance of the HAstV infection competition assay data was evaluated by comparing the infectivity in each protein dilution, compared to control cells without protein, using the Mann–Whitney test with GraphPad Prism 5 (GraphPad Software Inc., San Diego, CA, USA).

MLB spike sequences used for conservation analysis

The following sequences were used for the analysis of MLB-spike conservation in figure 3D: NCBI accession numbers ASU91602.1 residues 345-571, QCC21357.1 residues 420-646, QAA77562.1 residues 467-693, BAN57337.1 residues 420-646, ADJ38388.1 residues 420-646, AFD61563.1 residues 420-646, BAU68081.1 residues 420-646, BAN57334.1 residues 420-646, ACN44171.1 residues 420-646, UAW96012.1 residues 420-646, AGZ15312.1 residues 420-646, YP_006905854.1 residues 417-643, AJC68713.1 residues 417-643, BAN62843.1 residues 417-643, AZU90755.1 residues 417-643, YP_004934010.1 residues 417-643, QCC21360.1 residues 417-643, APB03099.1 residues 417-643, AER41415.1 residues 417-643, AMR45107.1 residues 417-643, and YP_002290968.1 residues 399-653.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal serum to HAstV 8 (Yuc8)	Méndez et al., 2004	N/A
Monoclonal antibody MAb 2A2	Janowski et al., 2021	N/A
Bacterial and virus strains
Bacterial cells: E. coli strain T7 Express	New England Biolabs	Cat#C2566H
Bacterial cells: E. coli strain B843 (DE3) pLysS	Novagen	Cat#69042
HAstV 8 strain Yuc8	Méndez et al., 2004	N/A
HAstV VA1	Janowski et al., 2017	N/A
Critical commercial assays
flashBAC Baculovirus Expression System	Mirus Bio	Cat#MIR6115
Deposited data
MLB1 Spike structure	This paper	PDB: 7UZT
Experimental models: Cell lines
Insect cells: Spodoptera frugiperda Sf9 cell line IPLB-Sf-21-AE	Expression Systems	Cat#94-001S
Caco-2 cells: C2Bbe1 cells, derived from the colon adenocarcinoma Caco-2 cell line	American Type Culture Collection	Cat# CRL-2102
Oligonucleotides
pBacPAK8-MLB1-spike_Forward1: ccaccatcaTCATTCAGAGACAACCTACACTGGCC	This paper	N/A
pBacPAK8-MLB1-spike_Forward2: gcagatgcaccaccaccatcatcaccaccatcaTCATTCAGAGACAACC	This paper	N/A
pBacPAK8-MLB1-spike_Reverse: gcaccagagcgagctctTTAGCGGTGCGGTATTCCCTC	This paper	N/A
pRSFDuet-MLB1-spike_Forward: ctggttccgcgtggtTCATCAGAGACAACCTACACTGGCC	This paper	N/A
pRSFDuet-MLB1-spike Reverse: GCGGTGCGGTATTCCCTC	This paper	N/A
Recombinant DNA
Codon-optimized synthetic gene encoding residues 399-653 of the MLB1 capsid protein (NCBI YP_002290968.1)	Genewiz	N/A
pBacPAK8	Takara Bio	631402
pRSFDuet	Novagen	71341
Software and algorithms
PyMol Molecular Graphics	System Schrodinger, LLC	https://pymol.org/2/
TM-align	Zhang Lab, University of Michigan	https://zhanggroup.org/TM-align/
IDDT (Local Distance Difference Test)	University of Basel	https://swissmodel.expasy.org/lddt

Highlights.

• Crystal structure of the human astrovirus MLB capsid spike domain is reported

• Classical and MLB human astroviruses have structurally-divergent capsid spikes

• Competition data support the existence of multiple human astrovirus receptors

• The MLB spike was a challenging target for structure prediction

Inclusion and Diversity

One or more of the authors of this paper self-identifies as an underrepresented ethnic minority in science. One or more of the authors of this paper received support from a program designed to increase minority representation in science. While citing references scientifically relevant for this work, we also actively worked to promote gender balance in our reference list.