Lipid moieties of Mycoplasma pneumoniae lipoproteins are the causative factor of vaccine-enhanced disease

Vaccine-enhanced disease (VED) occurs as a result of vaccination followed by infection with virulent Mycoplasma pneumoniae. To date VED has prevented development of an efficacious vaccine against this significant human respiratory pathogen. Herein we report that vaccination of BALB/c mice with M. pneumoniae lipid-associated membrane proteins (LAMPs) induces lung lesions consistent with exacerbated disease following challenge, without reducing bacterial loads. Removal of lipid moieties from LAMPs prior to vaccination eliminates VED and reduces bacterial loads after infection. Collectively, these data indicate that lipid moieties of lipoproteins are the causative factors of M. pneumoniae VED.


INTRODUCTION
Mycoplasma pneumoniae (Mp) is a highly contagious and widespread human respiratory pathogen causing over 2 million cases of community-acquired pneumonia (CAP) and~100,000 adult hospitalizations annually in the United States 1 . Additionally, innate resistance to commonly used β-lactam antibiotics, increasing resistance to macrolide antibiotics, and mortality in certain demographic groups establish Mp as a high-risk pathogen capable of causing acutely severe disease [2][3][4][5][6][7] . Critically, no vaccine against Mp is available, as experimental vaccine-enhanced disease (VED) after infection has stymied further development.
M. pneumoniae VED was first reported in human volunteers in the 1960s. In these studies, prisoners and military recruits were vaccinated with formalin-inactivated or a serially passaged liveattenuated strain of Mp, then subsequently challenged with a virulent strain of the bacterium. While some protection was observed, many vaccinated individuals in both studies exhibited more severe clinical symptoms than those receiving a placebo control 8,9 . Recent studies by our group (and others) have since recapitulated VED in an animal model by utilizing live-attenuated or crude extract Mp vaccine candidates, respectively [10][11][12][13] . In these murine models, VED is characterized by more severe histopathology post-challenge, when compared to sham vaccinated, Mp challenged mice. VED has also been observed in vaccine candidates against other Mycoplasma species, making the identification of the causative factors a crucially important task for vaccine development against these atypical pathogens.

RESULTS AND DISCUSSION
Lipoproteins are the most abundant components of the lipidassociated membrane protein (LAMP) fraction To identify the specific causative factor(s) for Mp VED, we utilized TX-114 phase partitioning to fractionate Mp proteins into three fractions: a LAMP hydrophobic detergent phase, an aqueous phase (Aq) harboring mostly hydrophilic cytosolic proteins, and an insoluble phase (Ins) containing the insoluble triton shell and a spontaneously forming, phospholipid-rich precipitate 14 . An initial pilot study indicated that vaccination of BALB/c mice with LAMPs resulted in VED after challenge with virulent Mp. We performed proteomic analysis to reveal lipoproteins, elongation factors, chaperones and chaperonins, and cytadherence proteins (previously found to induce strong antibody responses [15][16][17] ) as the most abundant immunogenic and antigenic components of the LAMP fraction ( Fig. 1a-f). Of these, lipoproteins are the most abundant (31.79%) (Fig. 1a) and include twenty-two lipoproteins representing all six Mp lipoprotein families (Fig. 1b, c). Family 2 lipoproteins, paralogs of the immunodominant Mycoplasma gallisepticum nucleotide-binding virulence factor MslA 18,19 , were the most abundant LAMP lipoproteins, making up 43.22% of the lipoprotein fraction (Fig. 1c). Lipoproteins not belonging to a multigene family were the second most abundant lipoproteins in the lipoprotein fraction (32.83%), followed by Family 4 lipoproteins (11.73%), Family 3 lipoproteins (5.53%) then Family 6 lipoproteins (4.36%) (Fig. 1c). Members of Family 5 lipoproteins (which are specific to Mp) were also identified in the fraction, albeit at minute relative quantities (0.67%) (Fig. 1c). Mycoplasma lipoprotein lipid moieties are known to be potent immunostimulators, inducing expression of inflammatory cytokines such as TNF-α, IL-6, and IL-1β following the recognition of their lipid moieties by Toll-like receptor complexes 20 . Given that these cytokines are frequently associated with immunopathology, we hypothesized that sensitization by Mp lipoprotein lipid moieties during vaccination induces VED upon challenge with virulent Mp.
Lipoprotein lipid moieties are responsible for VED To test our hypothesis, TX-114-derived Mp fractions were treated with exogenous lipoprotein lipase to generate delipidated fractions (dLAMPs, dAq, and dIns), which could be tested for immuno-stimulation in vitro and as vaccine candidates in vivo. Successful delipidation was assessed via a TLR2 bioassay. In murine macrophages, LAMPs stimulated higher levels of TNF-α than Aq and Ins fractions. Delipidation significantly reduced LAMPstimulated TNF-α production, notably to levels lower than those stimulated by Aq or Ins fractions (Fig. 2b). Surprisingly, delipidation also reduced Aq fraction-stimulated TNF-α production, suggesting that some lipoproteins may have sequestered to the Aq phase during TX-114 phase partitioning. Delipidation did not affect the ability of the Ins fraction to stimulate TNF-α production in vitro (Fig. 2b).
To test the effect of lipoprotein delipidation on VED, native and delipidated Mp fractions (LAMPs, dLAMPs, Aq, dAq, Ins, and dIns) were used to intraperitoneally vaccinate mice according to the schedule illustrated in Fig. 2a with 50 μg of protein from each fraction per dose. Shortly after vaccination, mice receiving nonlipase-treated fractions began to display clinical signs consistent with mild septic shock (Supplementary Table 1, Supplementary  Fig. 1), which we quantified as a group clinical score (Fig. 2c). While these clinical signs abated by 48-h post-vaccination, they point to potential immediate safety concerns for vaccine candidates containing intact lipoproteins. Notably, early clinical signs were absent in mice vaccinated with delipidated fractions. Given that in vivo data mirrored those obtained in vitro (Fig. 2b), early clinical signs may have resulted from the over-stimulation of peritoneal immune cells by the lipid moieties of native Mp lipoproteins.
In terms of VED, only mice vaccinated with LAMPs (containing native lipoproteins) exhibited histopathological lung lesions that were more severe than sham-vaccinated/challenged animals, consistent with LAMP-mediated VED (Fig. 2d, f). Furthermore, vaccination with LAMPs did not reduce bacterial loads, suggesting that the induced inflammation is solely pathologic and consistent with VED (Fig. 2e). Importantly, vaccination with delipidated LAMPs (dLAMPs) did not exacerbate lung pathology (Fig. 2d, f), demonstrating that the factors responsible for VED are the lipid moieties of Mp lipoproteins. Compared to controls, dLAMPvaccinated mice also had significantly lower bacterial loads (Fig. 2e). These data suggest that the dLAMP vaccine preparation includes protective antigens that improve bacterial clearance and can reduce Mp-induced lung disease when freed of the VED effects associated with Mp lipoprotein lipid moieties.
Mice vaccinated with MPLA-SM adjuvanted dLAMPs and challenged had significantly reduced lung lesion scores when compared to sham vaccinated, Mp challenged mice (Supplementary Fig. 2A, C), but did not differ in bacterial loads recovered ( Supplementary Fig. 2B). This was unsurprising as lesion scores did not always correlate with bacterial loads. These data further indicate that the choice of adjuvant will be critical in developing a vaccine that reduces bacterial loads while also protecting the host from unnecessary pulmonary inflammation.

CONCLUSIONS
The identification of Mp lipoprotein lipid moieties as the cause of Mp VED is a breakthrough for the field, overcoming a major roadblock for Mp vaccine development. The findings presented here may be broadly applicable to other Mycoplasma species for which vaccine induced disease exacerbation has been reported. Vaccination of calves with inactivated whole cell vaccines and partially purified membrane proteins of Mycoplasma bovis resulted in increased pulmonary pathology upon virulent challenge 21,22 . Additionally, vaccination of cattle with inactivated whole cell Mycoplasma mycoides subsp. mycoides SC and a subunit lipoprotein Q (LppQ) vaccine also were reported to exacerbate the effects of Contagious Bovine Pleuropneumonia (CBPP) in challenged animals 23 . These studies did not identify the lipid moieties of lipoproteins or any other factors as the causative factors of VED in these cases. Given the parallels of these studies with our data, however, it is attractive to speculate that the lipid moieties of the lipoproteins included in these vaccine candidates may have been the maladaptive factors responsible for disease exacerbation. Indeed, vaccination of cattle with the N terminus of LppQ (which  5). Arrows point to perivascular and peribronchiolar leukocytic infiltrates. Error bars indicate mean and standard error of the mean (SEM) for parametric data (b) and median and interquartile ranges (IQR) for non-parametric data (d, e) (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).
includes the region where the lipid moieties attach) failed to be protective and was also associated with adverse events 24 . It is important, however, to note that more studies need to be conducted to demonstrate the maladaptive effect of lipoproteins in ruminant mycoplasmosis vaccines.
In terms of M. pneumoniae, our data indicate that caution should be used when considering the inclusion of native Mycoplasma lipoproteins in future vaccine formulations against this significant human pathogen. Furthermore, the findings here may be broadly applicable to other bacterial pathogens for which VED is observed, and suggest that the potential role of lipoproteins in this phenomenon should not be easily dismissed.

Proteomic analysis of Mp LAMPs
Purified LAMPs were prepared using a slightly modified filter-aided sample preparation (FASP) method 26 described in the supplementary methods (see Extended Methods). Purified peptides were injected onto a PepMap RSLC C18 column (Thermo Scientific) and separated using reversed phase gradient on a Dionex Ultimate 3000 RSLC UPLC instrument (Thermo Scientific) coupled directly to a Q Exactive HF mass spectrometer (Thermo Scientific) via electrospray ionization. MaxQuant (v1.6.1.0) was used to search raw files against a custom Mp proteome database (CP010538_faa) using the Andromeda search engine 27 and for peptide and protein quantification using the LFQ algorithm. Scaffold v4.9 (Proteome Software, Inc.) was used for visualization and analysis, and gene designations were identified through comparison to the Mycoplasma pneumoniae (T00006) KEGG genome database with 95-100% sequence identity and lowest Evalue used as cut-offs. Multigene Mp lipoproteins were categorized into the 6 numerically named families based on previously established criteria 28 . Lipoproteins not belonging to a multigene family were categorized as "No Defined Family".

Bacterial strains and animal studies
Mycoplasma pneumoniae strain PI1428 was utilized for all aspects of this study. All animal experiments were conducted in accordance with approved Institutional Animal Care and Use Committee protocol (A17-034) as described in the supplementary methods (see Extended Methods).

Statistical analyses
Results were analyzed via a non-parametric one-way ANOVA on ranks (Kruskal-Wallis) with a Dunn's post hoc test or a parametric ordinary oneway ANOVA with a Tukey's post hoc test for multiple pairwise comparisons between groups. Analysis of data with only two groups was conducted utilizing a one tailed Mann-Whitney U test for nonparametric data (α = 0.05). All data were analyzed using the GraphPad Prism software, version 8.02 (GraphPad Software, La Jolla, California, USA).

Reporting summary
Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.

DATA AVAILABILIY
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE 30 partner repository with the dataset identifier PXD016814. All other data are included in this manuscript and supplementary file.