The Regulation of Nucleic Acid Vaccine Responses by the Microbiome

This article is featured in Top Reads, p. 1607

Vaccines are among the most effective tools for preventing human morbidity and mortality from infectious disease. Recently, nucleic acid vaccines encompassing both RNA and DNA platforms have emerged as a key technology offering potential advances over traditional protein-adjuvant approaches (1–3). In particular, nucleic acid vaccines promote enhanced cellular as well as humoral immunity, which may be particularly effective in combating intracellular pathogens and cancers (4, 5). Furthermore, advantages in the speed and flexibility of design and production make nucleic acid vaccines particularly valuable in combating emerging pathogens (1, 6). RNA-based COVID-19 vaccines combining modified mRNA molecules with lipid nanoparticle (LNP) technology were brought to the clinic within 1 y of SARS-CoV-2 emerging in humans, and their remarkable clinical efficacy has played a vital role in controlling the COVID-19 pandemic (7, 8). DNA vaccines demonstrated impressive preclinical results; however, this has so far not translated easily to the human setting, where DNA vaccine immunogenicity has been lower than was anticipated (9). Modified DNA vaccine delivery regimens have been developed to overcome low immunogenicity, including combining DNA and protein vaccines in prime-boost regimens that enhance the quality and quantity of the elicited responses relative to either modality alone (10–14).

Despite considerable past success, vaccine effectiveness remains hard to predict, and considerable interindividual heterogeneity in the immune response to vaccines can limit their clinical effectiveness (15, 16). The determinants of vaccine response across individuals remain poorly understood, especially for nucleic acid vaccines, and likely involve a variety of immune-modulatory factors, such as age, genetics, nutrition, past microbial infections, and the microbiome (16–18). The microbiome is an important regulator of immune development and response (19, 20). Data from preclinical mouse models have demonstrated that components of the microbiome can influence the immunogenicity of several vaccine platforms (21), including trivalent inactivated influenza vaccine (22), canonical infant vaccines (23), adenoviral Mycobacterium tuberculosis vaccine (24), live-attenuated rabies vaccine (25), and proteins adjuvanted with CFA (26), alum (27), CpG (28), or cholera toxin (29). Associations between vaccine response and microbiota composition have also been observed in human cohorts (30–35), and microbiota-disrupting antibiotic treatment of human subjects has been associated with altered influenza and rotavirus vaccine immunogenicity (33, 36). However, whether the microbiome regulates RNA or DNA vaccine immunogenicity remains undescribed. Furthermore, the precise nature of microbiome control over existing vaccines remains to be resolved with contrasting effects reported, depending on factors such as the age of mice or the route of administration (23, 29), and for similar experiments reproduced across institutions (22, 23).

We explored the contribution of the microbiome to different vaccine technologies by comparing the immune responses of germ-free (GF) and specific pathogen-free (SPF) mice immunized with different protein-adjuvant vaccine regimens, a DNA-prime DNA-protein-boost vaccine regimen (10, 37), and a single-dose nucleoside-modified mRNA-LNP vaccine (38–40). Our data indicate contrasting roles for the microbiome in regulating protein, DNA-prime DNA-protein-boost, and mRNA-LNP vaccine responses and provide a basis for further interrogation of the microbiome-dependent mechanisms that regulate nucleic acid vaccine immunogenicity.

SPF C57B6/J mice were obtained from The Jackson Laboratory (Room RB17 or RB05) and housed at Fred Hutchinson Cancer Research Center. GF C57BL6/J mice were bred in-house at the University of Washington Gnotobiotic Animal Core and housed in hermetically sealed Tecniplast (West Chester, PA) cages as previously described for the duration of experiments (41). Stools were collected from GF mice longitudinally and at the experimental endpoints, and sterility was confirmed by Gram staining and sporadic aerobic and anaerobic culture on tryptic soy agar with 5% sheep’s blood and modified yeast casitone fatty acid agar (DSMZ, 1611). GF breeding colonies were further screened quarterly for the presence of bacterial contaminants by amplification of the 16S rRNA gene via PCR. All materials and equipment used with GF animals, such as restrainers for immunization, were sterilized by autoclaving or submersion in 1:3:1 Clidox disinfectant prior to use. Male and female mice were aged 6–16 wk at the start of immunization regimens. Experimental groups were age matched and distributed throughout cages during vaccination experiments to reduce the confounding influence of “cage effects” due to differences in mouse behavior or SPF microbiome composition between cages. SPF and GF mouse work was approved by the Fred Hutchinson Cancer Research Center and University of Washington institutional animal care and use committees, respectively, and all mice were euthanized following American Veterinary Medical Association guidelines for CO₂ overdose.

Akkermansia muciniphila (A. muc, American Type Culture Collection [ATCC] BAA-835), Collinsella aerofaciens (C. aero, ATCC 25986), Fusobacterium varium (F. var, ATCC8501/DSM 19868), and Prevotella copri (P. cop, DSM 18205) were grown at 37°C in liquid anaerobic yeast casitone fatty acid agar culture and diluted to OD₆₀₀ 0.5–1.1 ml of individual cultures. Next, they were combined, pelleted by centrifugation at 5000 × g, and resuspended in 1 ml sterile PBS, and 0.5 ml segmented filamentous bacteria (SFB)-containing fecal material (from SFB monocolonized mice) was mixed with the 1-ml bacterial suspension. Each GF mouse received a single 100-μl dose of the pooled consortia via oral gavage. CFUs gavaged per mouse were calculated by serial dilution plating (10⁷–10⁸ for A. muc, C. aero, F. var, and 10⁵–10⁶ for P. cop). Fecal material was collected from colonized mice throughout the experiment, DNA was extracted using the QIAamp Fast DNA Stool Mini Kit (Qiagen), and the genome copies of each bacterium were quantified by quantitative PCR using species-specific 16S rDNA primers (see Table I) relative to standard curves of pure genomic DNA from each strain. Quantitative PCRs were run using PowerUp SYBR Green Master Mix (Invitrogen) on a QuantStudio 3 thermocycler with the following conditions: 50°C for 2 min, 95°C for 2 min, 35 cycles of 95°C for 1 s, 60°C for 30 s (adjusted to 55°C for C. aero), and 72°C for 2 min. One week following inoculation, mice were immunized with OVA and AddaVax as described below. Wild mouse microbiome-rederived (WILD-R) and SPF gastrointestinal material from the terminal ileum and cecum was collected anaerobically, weighed, and cryopreserved at a 1:30 dilution (g/ml) in PBS containing 0.1% cysteine and 12% glycerol as previously described (42). GF mice were inoculated with 100 μl WILD-R or SPF material on 3 consecutive days. Two weeks following inoculation, blood samples were collected for immunophenotyping using fluorescence-conjugated Abs (see Table II) and flow cytometry. Four weeks following inoculation, mice were immunized with OVA/alum as described below. DNA was extracted from residual WILD-R and SPF donor inoculum and from the cecal contents of WILD-R–colonized and SPF-colonized (SPF-C) mice using the AllPrep PowerViral DNA/RNA Kit (Qiagen). 16S rDNA amplicon profiling was completed using whoi341 forward (5′-CCTACGGGNGGCWGCAG-3′) and whoi785 reverse (5′-GACTACHVGGGTATCTAATCC-3′) primers with 300-bp paired-end reads generated on the Illumina MiSeq platform (>20,000 reads per sample) (www.mrdnalab.com; MR DNA, Shallowater, TX). Operational taxonomic unit tables were generated using the MR DNA ribosomal and functional gene analysis pipeline, and α and β diversity metrics were calculated and visualized using the tidyMicro pipeline in R (43).

GF and SPF mice were immunized with a standard prime-boost schema with 10 µg EndoFit OVA protein (InvivoGen, vac-pova-100) s.c. at the base of the tail (50-µl injection volume) on study day 0 and day 21. OVA was formulated 1:1 with Alhydrogel (alum; InvivoGen, vac-alu-250) or AddaVax (InvivoGen, vac-1dx-10), or 1:2 with AS01 (kindly provided by GSK), or administered unadjuvanted with Ultra Pure PBS (VWR, K812-500ML). Negative control mice received Ultra Pure PBS without OVA by the same route. All vaccine formulations were handled under strict sterile conditions throughout. Mice were euthanized on day 42, and tissue was harvested aseptically for subsequent OVA-specific IgG quantification and ex vivo OVA restimulation assays.

An HIV-1 DNA vaccine, DNA-HIV-PT123, used in HIV Vaccine Trials Network (HVTN) clinical trials HVTN 108 and HVTN 111 was provided by EuroVacc for use in these experiments (44, 45). DNA-HIV-PT123 consists of three plasmids expressing (1) HIV-1 Subtype C ZM96 gag, (2) HIV-1 Subtype C ZM96 gp140 ENV, and (3) HIV-1 Subtype C CN54 pol-nef. Mice were immunized s.c. at the base of the tail with 100 μg DNA-HIV-PT123 formulated in Ultra Pure PBS (VWR, K812-500ML) on study day 0, day 14, and day 35. In select experimental groups, as specified in the text and Figs. 3 and 4, boost immunizations with 2 μg bivalent gp120 protein consisting of 1 μg TV.1C gp120 and 1 μg 1086.C gp120 formulated with 1:1 (v/v) MF59 (kindly provided by GSK) or in Ultra Pure PBS were administered on day 14 and day 35. When delivered on the same day, DNA-HIV-PT123 and bivalent gp120 protein vaccines were prepared separately and delivered contralaterally on either side of the tail. Negative control mice received Ultra Pure PBS by the same route on days 0, 14, and 35. Serum and spleen tissue was collected at day 56 for subsequent gp120- and gag-specific IgG quantification and ex vivo gp120 restimulation assays.

Codon-optimized coding sequences of Spike receptor binding domain (aa 1–14 fused with aa 319–541) of SARS-CoV-2 (Wuhan-Hi-1, GenBank accession no. MN908947.3) were synthesized and cloned into a proprietary mRNA production plasmid. mRNAs were produced to contain 101-nt-long polyadenylation tails. m1ψ-5′-triphosphate instead of UTP was used to generate modified nucleoside-containing mRNA. Capping of the in vitro transcribed mRNAs was performed cotranscriptionally using the trinucleotide cap1 analog, CleanCap. mRNA was purified by cellulose purification, as described previously (46). All mRNAs were analyzed by agarose gel electrophoresis and were stored frozen at −20°C. Cellulose-purified m1ψ-uridine–containing RNAs were encapsulated in LNPs using a self-assembly process as previously described wherein an ethanolic lipid mixture of ionizable cationic lipid, phosphatidylcholine, cholesterol, and polyethylene glycol lipid was rapidly mixed with an aqueous solution containing mRNA at acidic pH (47). The LNP formulation used in this study is proprietary to Acuitas Therapeutics; the proprietary lipid and LNP composition are described in U.S. patent no. US10221127. The RNA-loaded particles were characterized and subsequently stored at −80°C at an RNA concentration of 1 mg ml⁻¹ (in the case of loaded particles) and total lipid concentration of 30 mg ml⁻¹. The mean hydrodynamic diameter of mRNA-LNPs was 80 nm with a polydispersity index of 0.02–0.06 and an encapsulation efficiency of 95%. GF and SPF mice were immunized with 1 μg mRNA-Spike-LNP s.c. at the base of the tail in a 100-μl volume with Ultra Pure PBS (VWR, K812-500ML).

At the experimental endpoints of all experiments, blood was collected by cardiac puncture following euthanasia. Serum was isolated after blood clotted at room temperature for 30 min and was centrifuged at 3500 × g for 10 min. All serum was heat inactivated at 56°C for 30 min with intermittent mixing prior to use in ELISAs described in this article.

Half-area high-binding plates (Corning, 3690) were coated by overnight incubation with 200 ng OVA in 0.1 M NaHCO₃, pH 9.5, at 4°C. Wells were blocked for 2 h at 37°C with 10% nonfat milk (w/v, Research Products International, M17200-500.0) and 0.03% Tween 20 (Fisher Bioreagents, BP337-100) in PBS. Serum was diluted in PBS containing 10% nonfat milk and 0.03% Tween 20. OVA-specific mouse IgG1, IgG2b, and IgG2c mAbs (Chondrex) were used to create internal standard curves for quantification of these IgG subclass titers. Quantities of 50 μl diluted serum samples and standard Abs were added to wells in duplicate and incubated for 1 h at 37°C. For detection of anti-mouse IgG1, IgG2b, and IgG2c isotypes, HRP-conjugated goat anti-mouse secondary Abs were added at 0.5 μg/ml in 50 μl PBS containing 10% nonfat milk and 0.03% Tween 20, and plates were incubated for 1 h at 37°C. Plates were developed with 1× SureBlue Reserve TMB (VWR, 95059-290) for 7–15 min, depending on isotype, and reactions were stopped with 1 N H₂SO₄ (Fisher Chemicals, SA212-1). Absorbance at 450 nm was read on a SpectraMax i3x plate reader (Molecular Devices) and analyzed using SOFTmax Pro software version 6.5.1. After each incubation step, wells were washed four times with 140 μl PBS containing 0.02% Tween 20.

Half-area plates (Corning, 3690) were coated overnight at room temperature with 100 ng bivalent gp120 or 100 ng SARS-CoV-2 Spike protein (BEI Resources, NR-52397) in 0.1 M NaHCO₃, pH 9.5. Wells were blocked with 10% nonfat milk (w/v; Research Products International, M17200-500.0) and 0.03% Tween 20 (Fisher Bioreagents, BP337-100) in PBS by incubating for 2 h at 37°C. Serum was diluted 1:10 in PBS containing 10% nonfat milk and 0.03% Tween 20 and loaded onto the plate in duplicate. Subsequently, serum was serially diluted 1:3 across the plate to provide an 11-point dilution curve, and plates were incubated for 1 h at 37°C. Anti-mouse IgG-HRP was added to each well in PBS containing 10% nonfat milk and 0.03% Tween 20 and incubated for 1 h at 37°C. Plates were developed and absorbance quantified as described above. Relative anti-gp120 IgG of each sample was calculated as the area under the curve of a four-parameter logistic regression of OD dilution curves.

Spleens were removed aseptically into sterile cRPMI (RPMI 1640 supplemented with 2 mM l-glutamine (Life Technologies, 25030081), 10% FBS (Life Technologies, 10437028), and 100 U/ml penicillin-streptomycin (Life Technologies, 15140122) and then disaggregated through a 100-µm mesh filter to generate a single-cell suspension. Cells were washed with 2% FBS in 1× PBS, and RBCs were lysed by incubation in sterile 1× RBC Lysis Buffer (Invitrogen, 501129751) for 3 min at room temperature. Cells were washed in 2% FBS in PBS, and concentrations of live cells were standardized in cRPMI by manual counting using a hemocytometer. For intracellular cytokine staining (ICS), 1 × 10⁶ splenocytes/well were added to a 96-well, round-bottomed plate and incubated in cRPMI alone (unstimulated negative control) or with 10 µg/ml Ag (OVA or bivalent gp120) for 24 h. After 19 h, brefeldin A (5 μg/ml; BioLegend, 420601) and monensin (2 μM; BioLegend, 420701) were added to block secretion pathways. To quantify secreted cytokines, 5 × 10⁶ cells/well were added to 12-well plates and incubated with cRPMI alone or 10 µg/ml Ag (OVA or bivalent gp120) in a 2-ml volume for 72 h. Supernatant was harvested following centrifugation at 400 × g for 5 min and stored at −80°C until analysis. For both ICS and secreted cytokine analyses, positive control wells containing anti-mouse CD3 and CD28 (0.5 µg/ml) were included for each individual mouse. All cell culture incubation was performed at 37°C with 5% CO₂.

Spleen tissue was collected aseptically, stored in 2-ml complete RPMI, and shipped overnight on ice for restimulation assays performed in the Weissman laboratory at the University of Pennsylvania. Splenocytes were collected, counted using a ViaCell automated cell counter, and seeded into a 5-ml polypropylene FACS tube at a final concentration of 20,000 cells/μl (100 μl/tube). Cells were stimulated with 80 μl cRPMI media containing overlapping peptide polls (15-mer, 11-aa overlap, 4-aa offset, >70% pure as per liquid chromatography–mass spectrometry) at a final concentration of 2.5 μl/ml and 1 μg/ml of anti-CD28/CD49d mixture (costimulatory signal) for 6 h. One hour after stimulation, brefeldin A and monensin at final concentrations of 5 μg/ml and 2 μM, respectively, were added to block cytokine secretion. All cell culture incubation was performed at 37°C with 5% CO₂.

Following incubation with or without Ag, cells were washed with PBS containing 2% FBS and incubated with Fc block (anti-mouse CD16/32) and LIVE/DEAD Aqua (1:500 dilution; Invitrogen, L34966) in PBS for 20 min on ice. Cells were washed in PBS and then incubated in a surface Ab panel (CD3, CD4, and CD8) in PBS containing 2% FBS for 30 min on ice. Cells were washed with PBS containing 2% FBS, fixed and permeabilized, and stained with an intracellular cytokine panel (IL-4, IL-10, IL-17, IFN-γ, and TNF-α) using a Cytofix/Cytoperm ICS kit (BD Biosciences, 554714) according to the manufacturer’s instructions. Details of antibodies used in these experiments are provided in Table II. Flow cytometry data for OVA- or gp120-stimulated ICS were acquired on an LSRFortessa X-50 flow cytometer (BD Biosciences) and analyzed using FlowJo version 10.7.1 (BD Biosciences). Flow cytometry data for Spike-stimulated ICS were acquired on a BD LSR II equipped with 4 laser lines and 18 photomultiplier tubes. Gates for proportions of cytokine-positive cells were set relative to fluorescence-minus-one controls containing all Abs except against the cytokine of interest.

Concentrations of IL-4, IL-10, and IFN-γ were measured in splenocyte culture media supernatant using a Th1/Th2 Mouse Uncoated ELISA Kit (Invitrogen, 88-711-44) following the manufacturer’s protocols. Unstimulated and anti-CD3/CD28 samples were included as negative and positive controls, respectively.

Twenty-four hours after mRNA-Spike-LNP or PBS s.c. base-of-the-tail administration, spleen tissue was collected aseptically. A small portion of spleen (3–5 mm in length) was preserved in RNAlater (Thermo Fisher) and stored at −80°C. Remaining spleen tissue was fragmented with a sterile razor blade; placed in 4.5 ml RPMI 1640, 10% FCS, 7.5 mM HEPES, with 0.75 mg/ml collagenase II (Sigma-Aldrich); and incubated for 35 min at 37°C with 200 rpm agitation. Collagenase activity was quenched with 200 μl 0.5 M EDTA, and tissue was disaggregated through a 70-μm cell strainer. Cells were washed with 2% FBS in 1× PBS, and RBCs were lysed by incubation in sterile 1× RBC lysis buffer (Invitrogen, 501129751) for 3 min at room temperature. Cells were washed in 2% FBS in PBS, and concentrations of live cells were standardized by manual counting using a hemocytometer. A total of 2 × 10⁶ cells were incubated with Fc block (anti-mouse CD16/32) and LIVE/DEAD Blue (1:500 dilution; Invitrogen, L23105) in PBS for 20 min on ice. Cells were washed in PBS and then incubated in a surface Ab panel in PBS containing 2% FBS for 30 min on ice (Table II). Cells were fixed (BD Biosciences, 554714) and stored in PBS in the dark at 4°C overnight prior to data acquisition on a BD LSRFortessa X-50 flow cytometer (BD Biosciences) and analysis using FlowJo version 10.7.1 (BD Biosciences). RNA was extracted from preserved spleen tissue using the RNeasy Mini Plus kit (Qiagen), integrity was confirmed using Agilent 4200 Tape Station (Agilent), and mRNA transcripts were counted in 100 ng total RNA using NanoString nCounter hybridization and quantification protocols. Count data were analyzed using nSolver 4.0 software (NanoString) with the advanced analysis package installed for pathway scoring.

Prism version 9.0.0 (GraphPad Software, La Jolla, CA) was used to complete statistical analyses. Statistics were calculated for ELISA and ICS results using tests specified in the figure legends and were considered statistically significant if p ≤ 0.05.

GF and SPF C57BL/6J mice were immunized s.c. on day 0 and day 21 with OVA protein formulated with either Alhydrogel (OVA/Alum), AddaVax (OVA/AddaVax), or AS01 (OVA/AS01) adjuvants. GF and SPF mice injected with OVA in PBS (OVA/PBS), or PBS alone (naive) were employed as negative controls. Three weeks after the final immunization, anti-OVA IgG1, IgG2c, and IgG2b were quantified in serum by ELISA (Fig. 1A). Anti-OVA-IgG1, -IgG2c, and -IgG2b titers were all highest in OVA/AS01-immunized mice, followed by OVA/Alum- and OVA/AddaVax-immunized mice. This effect was strongest for anti-OVA IgG2c consistent with enhanced type 1 polarization by AS01 in mice compared with Alum or AddaVax. There were no significant differences in anti-OVA IgG1, -IgG2c, or -IgG2b between vaccine-matched GF and SPF mice, indicating a minimal contribution of the SPF microbiome to the humoral responses of these protein-adjuvant vaccines. Mice receiving OVA/PBS displayed anti-OVA-IgG1, -IgG2c, and -IgG2b titers 4–5 log units lower than adjuvanted mice, as expected. Anti-OVA-IgG2b titers were slightly but significantly higher in OVA/PBS-injected SPF mice than in OVA/PBS-injected GF mice, consistent with a role of the SPF microbiome in supporting T-independent humoral responses to unadjuvanted Ag.

In order to assess vaccine-induced T cell responses, we restimulated splenocytes from OVA/AS01-, OVA/Alum-, and OVA/PBS-immunized mice with OVA protein ex vivo. T cell responses were quantified both by ICS flow cytometry and by the concentration of cytokine released into the culture supernatant relative to unstimulated controls (Fig. 1B–1D, Supplemental Fig. 1). OVA-stimulated IFN-γ⁺ CD4⁺ T cell proportions and IFN-γ concentration in the culture supernatant were markedly increased in OVA-stimulated splenocyte cultures from OVA/AS01-immunized mice compared with OVA/Alum- or OVA/PBS-immunized mice, consistent with the enhanced type 1 polarization activity of AS01 (Fig. 1B). OVA-stimulated IL-4 concentration was also increased in splenocyte cultures from OVA/AS01- but not OVA/Alum-immunized mice compared with OVA/PBS controls, although these IL-4 responses were low (less than 100 pg/ml) and were not detectable by ICS. IFN-γ and IL-4 concentrations were not significantly different in OVA-stimulated splenocyte cultures from vaccine-matched GF and SPF mice (Fig. 1B, 1C). Finally, both IL-10 cytokine concentration and the proportion of IL-10⁺ CD4⁺ T cells were increased in OVA-stimulated splenocyte cultures from GF OVA/AS01-immunized mice, and these IL-10 responses were significantly higher than those of SPF OVA/AS01-immunized mice (Fig. 1D), suggesting that the microbiome may play a role in suppressing IL-10 responses in OVA/AS01-immunized mice.

In order to extend our assessment of microbiome influence on protein-adjuvant immunization beyond the conventional SPF microbiome of C57BL6/J mice, we employed two distinct colonization approaches. First, we infected GF mice with a five-member microbial consortia (PRO-consortia) consisting of Akkermansia muciniphila, Prevotella copri, Fusobacterium varium, Collinsella aerofaciens, and SFB selected for their reported immune-stimulatory activities in vivo (48–50). GF mice were infected by oral gavage beginning 7 d prior to immunization with OVA/AddaVax in a prime-boost regimen. Quantification of bacterial 16S rDNA copy number in the stool of colonized mice with species-specific primers indicated that P. copri failed to colonize, whereas the remaining four members of the consortia displayed stable colonization levels over the course of the experiment (Supplemental Fig. 2A). There was no difference in OVA-specific IgG titers between PRO-consortia-colonized mice and uncolonized GF controls (Supplemental Fig. 2B). Second, we infected GF mice with cryopreserved cecal material obtained from WILD-R mice. GF mice were orally gavaged on 3 consecutive days with either WILD-R or SPF material or remained uncolonized (GF), whereas standard SPF-raised mice were employed as controls. Two weeks after colonization, blood samples were collected from GF, WILD-R-colonized, SPF-C, and SPF-raised mice, and the proportions of circulating immune cells were measured by flow cytometry. No significant differences detected between groups (Supplemental Fig. 2C), suggesting that no sustained inflammatory reaction was taking place as a result of exposure to WILD-R or SPF material. Mice were immunized 4 wk after colonization with OVA/Alum in a prime-boost regimen. No differences between OVA-specific IgG1, IgG2b, or IgG2c were detected between GF, SPF-C-colonized, or WILD-R-colonized mice (Supplemental Fig. 2D). 16S rDNA sequencing of cecal contents of WILD-R and SPF-C mice indicated substantial differences in the microbiome communities between WILD-R and SPF-C mice, as shown by principal coordinate analysis using the Bray-Curtis distance metric at the species level (Supplemental Fig. 2E). Several families of bacteria were differentially abundant between WILD-R-colonized and SPF-C mice with Bifidobacteriaceae, Desulfovibrionaceae, Eggerthellaceae, Marinilabiliaceae, Odoribacteraceae, Sphingobacteriaceae, Streptococcaceae, Tannerellaceae, and the candidate phylum Candidatus Saccharibacteria significantly increased in WILD-R-colonized mice, and Bacteroidaceae, Clostridiaceae, Eubacteriaceae, and Vallitaleaceae were significantly increased in SPF-C mice (Supplemental Fig. 3F). The α diversity, as measured by Simpson or Shannon index, of WILD-R-colonized mice trended toward being increased compared with SPF-C mice (Supplemental Fig. 3G). However, the α diversities of WILD-R-colonized mice were notably lower than the donor WILD-R material, indicating incomplete transplant of WILD-R microbiome diversity by this procedure. Indeed, the phyla most differentially abundant between donor WILD-R and SPF material, Proteobacteria consisting mostly of Helicobacter species, failed to colonize GF mice, indicating the challenges in using this transplant approach to fully recapitulate the WILD-R microbiome community (Supplemental Fig. 3H). Collectively, these data show that addition of select immunostimulatory species or reconstitution with a WILD-R microbiome community was insufficient to modulate OVA/AddaVax or OVA/Alum vaccine immunogenicity in mice.

The vaccine DNA-HIV-PT123 consists of three plasmids encoding HIV gp140 env, gag, and pol-nef fusion proteins and has been employed in several clinical studies (44, 45). We employed DNA-HIV-PT123 in GF and SPF mice using a DNA-HIV-PT123-prime, DNA-HIV-PT123 + MF59-adjuvanted gp120 protein-boost regimen that mirrors the regimen employed in the HVTN 108 and 111 clinical trials (Fig. 2A). Mice were immunized s.c. by needle syringe on day 0 with DNA-HIV-PT123 alone and on day 14 and day 35 with both DNA-HIV-PT123 and gp120 protein adjuvanted with MF59 (Fig. 2A). Three weeks after the final immunization, anti-gp120 IgG was quantified in serum by ELISA, and splenocytes were restimulated with gp120 protein to measure T cell responses by ICS flow cytometry and by the concentration of cytokine released into the supernatant relative to unstimulated controls. GF mice receiving this immunization regimen displayed significantly higher mean anti-gp120 IgG Ab titers than SPF mice (Fig. 2B). A post hoc Levene test also indicated that anti-gp120 IgG titers of GF mice were significantly less variable than those of SPF mice (p = 0.0102). GF mice displayed significantly higher IFN-γ⁺ and TNF-α⁺ CD4⁺ T cell responses than SPF mice measured by ICS (Fig. 2C) following gp120 splenocyte restimulation. Consistent with this, IFN-γ concentrations were also increased in the supernatant of gp120-restimulated splenocyte cultures from GF compared with SPF mice (Fig. 2D). These data indicate that the microbiome suppresses both humoral and cellular immunogenicity to this DNA-HIV-PT123-prime, DNA-HIV-PT123 + gp120/MF59-boost regimen.

In order to define more precisely which components of this DNA-HIV-PT123-prime, DNA-HIV-PT123 + gp120/MF59-boost regimen the microbiota regulates, we compared the humoral and cellular responses of GF and SPF mice immunized with either (1) DNA-HIV-PT123 alone, (2) DNA-HIV-PT123 with MF59, (3) DNA-HIV-PT123 with unadjuvanted gp120, (4) gp120 with MF59, or (5) gp120 alone. No significant differences were observed in serum anti-gp120 IgG titers, gp120-responsive IFN-γ⁺ and TNF-α⁺ splenic CD4⁺ T cells, or gp120-induced splenocyte IFN-γ secretion between vaccine-matched GF and SPF mice for any of these partial regimens (Fig. 3), indicating that the full DNA-HIV-PT123 prime, DNA-HIV-PT123 + gp120/MF59 boost regimen is necessary to reveal microbiome-dependent regulation in this model. Notably, anti-gp120 IgG and CD4⁺ T cell ICS responses were generally low in mice receiving DNA-HIV-PT123 alone, DNA-HIV-PT123 with MF59, or DNA-HIV-PT123 with unadjuvanted gp120 protein, with some of the mice undergoing these regimens failing to mount a response above background even after three 100-μg doses of DNA-HIV-PT123 vaccine (Fig. 3). Serum anti-gag IgG titers were also not detectable in nearly all DNA-HIV-PT123-immunized mice, despite plasmids encoding the Gag immunogen being present (Supplemental Fig. 3). This indicates that in the absence of subsequent protein-adjuvant boost, s.c. needle syringe delivery of DNA-HIV-PT123 is poorly immunogenic in mice, regardless of gnotobiotic condition, thus limiting our ability to assess the contribution of the microbiome to regulation of DNA-HIV-PT123 immunization alone in this model.

To investigate whether the microbiome contributes to the immunogenicity of a nucleoside-modified mRNA-LNP vaccine, as used in Moderna and Pfizer/BioNTech COVID-19 vaccines, we immunized GF and SPF mice s.c. with a single immunization of mRNA-LNP vaccine encoding SARS-CoV-2 Spike protein; mice were euthanized 14 d later. Anti-Spike IgG titers were determined in serum, and CD4⁺ and CD8⁺ T cell responses were quantified by ICS flow cytometry following splenocyte restimulation with a spike peptide pool. There were no significant differences in anti-Spike IgG titers or in the proportions of CD4⁺ T cells producing IFN-γ, IL-2, IL-4, IL-5, TNF-α, or IL-17 between GF or SPF mRNA-LNP-immunized mice (Fig. 4A, 4B). However, there were significant reductions in the proportions of CD8⁺ T cells producing IFN-γ, IL-2, or TNF-α and coexpressing IFN-γ⁺ and CD107α⁺ in GF compared with SPF mice following mRNA-LNP immunization (Fig. 4C). These data indicate that CD8⁺ T cell responses to the mRNA-LNP immunization used in this study are enhanced by the presence of the microbiome.

To provide insight into the potential mechanisms by which the microbiome contributes to mRNA-LNP immunization, we completed cellular immune profiling of the early innate response to mRNA-LNP 24 h after immunization in spleen tissue of GF and SPF mice. There was a dramatic activation of CD11b⁺ Ly6C^hi monocytes following mRNA-LNP immunization, with nearly all Ly6C^hi monocytes in the spleens of immunized SPF mice displaying upregulation of the dendritic cell (DC)/macrophage marker CD11c (Fig. 5A), as well as upregulation of CD80 and CD86 (Fig. 5D). Interestingly, the total proportion of Ly6C^hi monocytes, as well as expression of CD11c and CD86 on Ly6C^hi monocytes, was reduced in GF compared with SPF mRNA-LNP-immunized mice (Fig. 5A, 5D), indicative of reduced Ly6C^hi monocyte activation in GF animals. We reproduced this experiment in a complementary model of microbiome depletion using broad-spectrum antibiotic treatment. There was a dramatic decrease in stool microbial 16s rDNA copy number in the first week of antibiotic treatment, which was sustained for the 4 wk until mRNA-LNP immunization (Supplemental Fig. 4A), and, at the time of immunization, no bacteria could be cultured either aerobically or anaerobically from stool collected from antibiotic-treated mice (Supplemental Fig. 4B). Consistent with the experiment in GF and SPF mice, there was a dramatic upregulation in CD11c expression on monocytes 24 h after mRNA-LNP immunization in this model, which was reduced in antibiotic-treated compared with untreated mRNA-LNP immunized mice (Supplemental Fig. 4C).

We employed a gating strategy to assess neutrophil and DC/macrophage populations based on expression of CD11b, Ly6G, CD64, Ly6C, CD11c, and MHC class II (Fig. 5B). mRNA-LNP immunization resulted in an accumulation of neutrophils (CD11b⁺Ly6G^hi), which was reduced in GF mice (Fig. 5C). We focused our initial analysis on distinct populations of Ly6C⁺ and CD64⁺ DC/macrophage cells, which have previously been associated with inflammation and the response to vaccination (51, 52), and Ly6C⁻CD64⁻ conventional DCs defined as follows: Ly6C⁺CD11c⁺MHC-II⁺ cells (Ly6C⁺ DCs), CD64⁺ Ly6C^−/lo CD11c⁺ MHC-II⁺ cells (CD64⁺Ly6C^−/lo DCs/mononuclear phagocytes), and Ly6C⁻CD64⁻CD11c⁺MHC-II⁺ cells (CD64⁻Ly6C⁻ conventional DCs [cDCs]) (Fig. 5B). The proportion of Ly6C⁺ DCs in the spleen was increased by mRNA-LNP immunization in both SPF and GF animals (Fig. 5C). Interestingly, there was also an increase in the proportion of CD64⁺Ly6C^−/lo DCs/mononuclear phagocytes following mRNA-LNP immunization in SPF mice that was not evident in GF mice (Fig. 5C). Meanwhile, proportions of CD64⁻Ly6C⁻ cDCs were not changed by mRNA-LNP immunization in either SPF or GF settings (Fig. 5C). Further analyses also demonstrated that proportions of XCR1⁺, CD8a⁺, and CD11b⁺ cDCs were not altered by mRNA-LNP immunization (data not shown). CD80 and CD86 were already expressed on these DC/macrophage populations prior to immunization, and although there were some small increases in CD86 expression 24 h after mRNA-LNP immunization on CD64⁺ DCs/mononuclear phagocytes, this was not significantly different between GF and SPF settings (Fig. 5D). Collectively, these data demonstrate the significant innate immune activation that occurs following mRNA-LNP immunization and highlight differences in both monocyte activation and CD64⁺Ly6C^−/lo DC/macrophage accumulation between SPF and GF mice.

To explore further the possibility that innate immune activation to mRNA-LNP immunization is altered in GF compared with SPF mice, we extracted RNA from the spleen tissue of mRNA-LNP-immunized GF and SPF mice and assessed the expression of a panel of 785 host response genes. Principal component PC analysis and unsupervised clustering illustrated that GF and SPF samples clustered separately, primarily along principal component 2, which explains ∼17% of variation in the dataset (Fig. 6A). Pathways associated with myeloid activation, host defense peptides, lysosomes, and phagocytosis were enriched in mRNA-LNP-immunized SPF mice, whereas pathways associated with lymphocytes, such as TCR and BCR signaling, were enriched in GF mice (Fig. 6B). These data provide further evidence that the innate myeloid response to mRNA-LNP immunization is reduced in GF compared with SPF mice, consistent with the flow cytometric analysis. In order to gain insight into what may be driving these differences in myeloid activation, we compared the relative expression of genes associated with different innate immune signaling pathways in SPF and GF mRNA-LNP-immunized mice. The top signaling pathway that was relatively enriched in SPF samples in this dataset was type I IFN signaling, whereas pathways of immune sensing, such as TLR or NOD-like receptor signaling, were comparatively enriched in GF settings (Fig. 6C).

We have identified distinct and contrasting ways in which the microbiome impacts DNA and mRNA-LNP vaccine responses. Specifically, the data shown here indicate that the microbiome suppresses humoral and cellular responses to DNA-prime, DNA-protein-boost vaccination, whereas the microbiome enhances CD8⁺ T cell responses to mRNA-LNP immunization. To our knowledge, these are the first data demonstrating regulation of these novel vaccine platforms by the microbiome and have implications for our understanding of how differences in vaccine response may be associated with microbiome heterogeneity. Future work will necessarily focus on understanding the full mechanistic basis underlying these different effects.

Our data indicate that the IFN-I response to mRNA-LNP immunization is reduced in GF compared with SPF mice, which is consistent with studies showing that IFN-I at homeostasis and following microbial challenge is microbiome dependent (53–59). The mRNA molecule in this vaccine is nucleoside modified and stringently purified to eliminate intrinsic immune-stimulatory activity (60–62), and therefore we assume the majority of this IFN-I response is being driven by the LNP component. IFN-I is an important regulator of both DNA and mRNA vaccine response, where its pleiotropic nature can result in opposing effects on immunogenicity (63–68). Prominent effects of IFN-I include enhancing T cell responses through costimulation, and recent data have indicated that myeloid activation and CD8⁺ T cell responses to a nucleoside-modified mRNA-LNP vaccine similar to the one used here are IFN-I dependent (69). Consistent with this, GF mice also displayed reduced CD8⁺ T cell responses and myeloid activation to mRNA-LNP vaccination, and therefore it is likely that an absence of microbiome-dependent IFN-I responses is at least part of the mechanism underlying this effect. A limitation of this study is that we were unable to quantify CD8⁺ T cell responses or interrogate innate immune responses to the DNA-HIV-PT123 vaccine to see if the same effect of the microbiome is observed in that setting. This is likely due to the low immunogenicity of the DNA vaccine delivered by an s.c. needle syringe, which is consistent with data obtained from human trials with this vaccine (37), and also the use of whole gp120 protein in the restimulation assay rather than fully optimized peptide pools.

In opposition to such immune-stimulatory IFN-I effects, IFN-I induced by DNA or unmodified mRNA vaccines can also suppress immunogen expression, which inhibits their immunogenicity (62, 65, 68, 70, 71). We hypothesize that microbiome-primed IFN-I may be suppressing DNA transgene expression in our model, providing a potential mechanism for the reduced DNA-prime DNA-protein-boost vaccine responses observed in SPF compared with GF mice. Given that nucleoside-modified mRNA-LNP vaccines are notably less sensitive to IFN-I–mediated suppression (60–62), such an outcome would also explain the contrasting contributions of the microbiome to DNA-prime DNA-protein-boost and mRNA-LNP vaccine-induced humoral and CD4⁺ T cell responses. The low immunogenicity of DNA-HIV-PT123 in mice when delivered s.c. by needle syringe has prevented us from directly testing this hypothesis in our current system. Further work will employ alternative DNA vaccine constructs and delivery methods to alleviate this limitation, facilitating further assessment of the sensitivity of DNA vaccines to microbiome regulation and to interrogate underlying IFN-dependent or -independent mechanisms. RNA vaccines that require cytosolic trafficking or amplification or that activate IFN-I responses similar to those of DNA vaccines may also be sensitive to IFN-I–mediated suppression, and therefore it will be informative to extend these studies to include unmodified and self-amplifying mRNA vaccine platforms (60–62, 70–73).

Further work is required to understand the innate immune responses elicited by DNA or mRNA-LNP immunization and detail the precise mechanisms that elicit humoral and cellular immunity. Our data report dramatic monocyte activation occurring within 24 h of mRNA-LNP immunization, as well as accumulation of Ly6C⁺ and CD64⁺ DC/macrophage populations that are likely mostly monocyte derived. This is consistent with recent data reporting myeloid activation within 24 h of mRNA-LNP vaccination in mouse draining lymph nodes and human peripheral blood (69, 74). The data shown here indicate that this mRNA-LNP–induced myeloid cell activation is reduced in GF mice, indicating regulation by the microbiome and warranting further investigation and validation in independent cohorts. This is consistent with profound regulation of mononuclear phagocyte activation by the microbiome in settings of viral infection, which is at least in part IFN mediated and epigenetically imprinted (53, 54). The mRNA expression profiling data also indicate relative enrichment of lymphocyte pathways in GF mRNA-LNP–immunized mice, which we interpret as being a reflection of reduced innate immune activation at this early 24-h time point rather than true induction of these pathways. However, further detailed kinetic profiling of the innate and adaptive response to mRNA-LNP activation in different gnotobiotic settings, including at the single-cell level, will help to elucidate this in more detail. T follicular helper cell), germinal center B cell, and CD8⁺ T cell responses to nucleoside-modified mRNA-LNP immunization are independent of TLR- 2, -3, -4, -5, and -7 and the TLR adaptor protein MyD88. T follicular helper and B cell responses are also independent of RIG-I/MDA-5 adaptor protein mitochondrial antiviral signaling but are dependent on IL-6 (40). Meanwhile, CD8⁺ T cell responses to mRNA-LNP are at least partially MDA-5 dependent (69). A myriad of innate immune pathways are capable of sensing microbial ligands, exogenous nucleic acid and/or LNP components (75), including TLRs (e.g., TLR9), and cytosolic sensors, such as cGAS-STING, RIG-I, and AIM2, although the degree to which these various receptors are influenced by the microbiome remains to be determined.

In contrast to our work with nucleic acid vaccines, our studies with protein-adjuvant combinations show that the endogenous mouse microbiome is not necessary for eliciting immune responses toward s.c. administered OVA vaccines adjuvanted with alum, AddaVax or AS01, or gp120 protein administered with MF59. This is consistent with published data investigating the contribution of the microbiome to systemically administered alum- or cholera toxin–adjuvanted vaccines in broad-spectrum antibiotic–treated mice (23, 29). Published data have suggested that the microbiome is necessary for immune responses to OVA adjuvanted with CFA and to other protein-based vaccines delivered mucosally (26, 29). Therefore, collectively, this work emphasizes that the contribution of the microbiome to protein-based vaccine responses is likely dependent on the nature of the adjuvant as well as the route of delivery. Recent work has also highlighted the microbiome as a potential source of Ag that cross-reacts with pathogen-, tumor-, or self-derived Ag (76–86). The presence of cross-reactive Ags has been hypothesized to divert HIV vaccine responses away from those that are protective (85, 86). Future work would be advantaged by extending beyond traditional model Ags, such as OVA, to consider microbiome–vaccine Ag cross-reactivity.

As is the case for the majority of microbiome studies in mice, our work is limited to the SPF microbiome of the C57BL/6 mouse lines sourced as detailed in the Materials and Methods. By definition, SPF microbiomes are constrained to a limited diversity, maintained under careful environmental control, and exclusive to potentially pathogenic agents. Thus, our work does not exclude vaccine modulatory contributions of microorganisms that are not represented in the mouse colony studied here but that may be present in the microbiome of other mouse lines or in the human population. Recent data have emphasized the differential responses of environmentally exposed “dirty” mice compared with SPF mice following live-attenuated or killed split influenza vaccination (88). We completed exploratory vaccine experiments colonizing adult SPF or GF mice with select immunostimulatory species or with WILD-R mouse gastrointestinal material, which increased the diversity of the microbiome community but was insufficient to modulate the protein-adjuvant vaccine responses tested. We observed that certain taxa of interest, such as WILD-R-derived Proteobacteria, failed to colonize GF mice using the approach described. Future work seeking to further explore the presence of vaccine modulatory microorganisms in the microbiome using gnotobiotic models should extend beyond the C57BL6/J sources investigated in the present study and consider approaches that mitigate the limitations of GF colonization (e.g., by colonizing earlier in life).

In conclusion, there is an urgent need to understand the determinants of nucleic acid vaccine immunogenicity and implement this understanding to maximize the potential for clinical success. The work reported in this article emphasizes the importance of considering the microbiome as a determinant of nucleic acid vaccine immunogenicity. We hypothesize several potential points whereby nucleic acid vaccines may be regulated by the microbiome, including at the level of transgene expression, innate immune response to adjuvant components, and lymphocyte activation/polarization. This establishes a foundation for investigation of microbiome features that may be associated with nucleic acid vaccine response and of candidate mechanisms that may be employed to predict or enhance protective immunity.

P.J.C.L. and Y.K.T. are employees of Acuitas Therapeutics, a company involved in the development of mRNA-LNP therapeutics. Y.K.T., D.W., and M.-G.A. are named on patents that describe LNPs for delivery of nucleic acid therapeutics, including mRNA and the use of modified mRNA in LNPs as a vaccine platform. The other authors have no financial conflicts of interest.

We are grateful to Drs. Song Ding and Giuseppe Pantaleo (EuroVacc) for provision of the DNA-HIV-PT123 vaccine, Dr. Alex Maue (Taconic) for provision of WILD-R gastrointestinal contents, and Dr. Clarisse Lorin (GSK) for provision of AS01. We are also grateful to the University of Washington Gnotobiotic Animal Core, Fred Hutchinson Cancer Center (FHCC) Comparative Medicine, the HVTN flow cytometry core, and FHCC genomics shared resource for experiment support.