Designing Protective Microbial Consortia Requires Joint Optimization of Inhibition, Coverage, Safety, and Engraftability

Protective microbial consortia should be designed as multi-objective formulations rather than selected solely for maximal pathogen inhibition. Inhibition, coverage of the pathogen’s resource niche, safety constraints, compatibility among members, and probability of engraftment can favor different species or community sizes. The Rational Design of Protective Microbiome Formulations for Competitive Exclusion of *Pseudomonas aeruginosa* in Cystic Fibrosis Airways provides a cystic-fibrosis airway case study centered on suppressing pseudomonas-aeruginosa through metabolic competition and direct antagonism. ₁

The study supports metabolic competition as a real but incomplete planktonic mechanism, while establishing an evidence boundary: planktonic inhibition does not directly establish protection against the structured Pseudomonas aeruginosa biofilms that primarily occupy cystic-fibrosis airways. Pairwise interaction measurements further support a mixed-mechanism interpretation, but their sparsity and assay limitations make additive formulation scores and extrapolation to multi-species communities provisional. ₁

Evidence base and working theory

The analysis integrates six evidence streams: planktonic inhibition assays (220 isolates), carbon source utilization profiling (430 isolates × 21 substrates), growth kinetics (32 isolates), patient metagenomics (175 samples), pairwise interaction data, and pangenome analysis (499 genomes across 6 species). Across this base, metabolic overlap with PA14 significantly predicts inhibition (r = 0.384, p = 2.3×10⁻⁶); the approximately 27% of variance that the report’s discussion credits to metabolic competition is the training R² = 0.274 of a four-feature multivariate model, not the explanatory power of the overlap correlation on its own, and five-fold cross-validation of that model gives a lower out-of-sample CV R² = 0.145 ± 0.142. The report reads the strongest inhibitors as combining metabolic competition with direct antagonism mechanisms. ₁

The working theory is that Pseudomonas aeruginosa can be excluded by commensal communities that consume the same carbon sources it depends on — primarily amino acids abundant in cystic-fibrosis sputum (proline, histidine, ornithine, glutamate, aspartate, arginine; Palmer et al. 2007); this is stated as the study’s working theory rather than an established mechanism. The report further hypothesizes that prebiotics — substrates that feed commensals but not PA — could give the consortium a biomass advantage before competition begins, explicitly as a hypothesis. ₁

The report’s own summary states that no individual commensal outgrows PA14 on any tested substrate, which is why it argues that niche coverage must be achieved at the community level. This broad statement stands in unresolved tension with the kinetic comparisons below, which quantify a maximum growth rate advantage for commensals on 13.8% of substrate comparisons and an earlier start (lag advantage) on 43.1% of comparisons. Whether “outgrows” in the categorical statement refers to a different metric or a different comparison set than the 13.8% figure is never clarified in the source, so the discrepancy is recorded rather than reconciled; niche coverage, the community-level property the categorical statement is used to motivate, is in any case a distinct quantity from relative growth rate. ₁

The other reporter pathogens, acinetobacter baumannii and klebsiella-pneumoniae, show distinct profiles, so PA14’s amino acid specialization is not universal among cystic-fibrosis pathogens, suggesting formulations may need to be pathogen-specific. This bounds every result below to a PA14-targeted design rather than a general anti-pathogen formulation. ₁

Metabolic competition explains part of antagonism

In synthetic cystic-fibrosis sputum conditions, PA14 preferentially used amino acids: endpoint OD values were 0.60 for proline, 0.56 for histidine, 0.46 for ornithine, 0.40 for glutamate, 0.36 for aspartate, 0.36 for isoleucine, and 0.35 for arginine. Glucose supported OD 0.22, whereas threonine, methionine, cysteine, serine, and glycine each supported essentially no growth at <0.07. ₁

Across 142 isolates with both inhibition and carbon-utilization measurements, metabolic overlap with PA14, weighted by PA14’s substrate preferences, significantly predicted planktonic inhibition, with r = 0.384 and p = 2.3×10⁻⁶. A multivariate model incorporating metabolic overlap, total growth on PA-preferred substrates, metabolic breadth, and maximum growth explained R² = 0.274 of inhibition variance; adding genus-level taxonomy increased the fit to R² = 0.360. Five-fold cross-validation, in which the model is repeatedly evaluated on held-out data, yielded CV R² = 0.145 ± 0.142, showing that out-of-sample predictive power was lower than the training fit. The report reads the true out-of-sample predictive power of metabolic features as closer to 15% than 27%, and states that this does not invalidate the qualitative conclusion — metabolic overlap remains a statistically significant predictor (p = 2.3×10⁻⁶) — but that the effect size should be interpreted conservatively. ₁

This supports condition-specific-fitness: resource-use similarity contributed to competitive fitness under a defined assay condition rather than providing a universal antagonism score. It also refines a metabolism-only interpretation: approximately 73% of inhibition variance remained unexplained, while genus contributed an additional 8.6% of variance. The genus contribution is consistent with species-specific direct-antagonism mechanisms, although the analysis did not identify the responsible molecules or interactions. ₁

The discussion frames these same numbers slightly differently from the model section: it reports an additional 9% from taxonomy where the model section gives 8.6%, and nominates bacteriocins, secreted enzymes (Stubbendieck et al. 2023), or contact-dependent killing as the species-specific mechanisms behind the strong positive residuals of S. salivarius, N. mucosa, and G. sanguinis. Both percentages are recorded as stated rather than reconciled, and the named mechanisms are proposed explanations, not established ones. ₁

On how much remains unexplained the source also gives two framings: the model section leaves approximately 73% of inhibition variance unexplained, while the discussion states that the remaining 64% of variance is likely attributable to unmeasured factors — biofilm dynamics, pH effects, iron competition, quorum sensing interference, and stochastic variation in the planktonic assay. Both figures are recorded without reconciliation, and the listed factors are candidate contributors rather than measured ones. ₁

Residual analysis identified inhibitors whose observed inhibition exceeded the metabolic model’s expectation: Streptococcus salivarius ASMA-737 had a positive residual of +74.1%, Gemella sanguinis ASMA-3044 had +62.2%, and Neisseria mucosa ASMA-3643 had +57.2%. These residuals support the hypothesis that some commensals combine resource competition with direct antagonism, rather than demonstrating a specific direct mechanism. ₁

Growth kinetics further separated possible mechanisms. PA14 was the fastest grower on preferred substrates in most comparisons; commensals exceeded its maximum growth rate in only 13.8% of substrate comparisons but began growing earlier in 43.1% of comparisons. Growth kinetic parameters were moderately correlated with endpoint OD at r ≈ 0.40, and adding kinetics increased the model fit to R² = 0.311 among the 29 isolates with all three assay types. ₁ From this asymmetry the report suggests that pre-establishing commensals before pathogen exposure — for example via prebiotic biomass pre-loading — could be more important than raw growth rate; that is a suggestion drawn from the lag-versus-rate contrast, not a demonstrated intervention, and it refines capability-versus-kinetic-predictability by separating metabolic capacity from timing. ₁

Alongside the kinetic result the report offers an illustrative community-biomass argument rather than a measurement: a 3-organism consortium each growing at half PA14’s rate still depletes resources 1.5× faster collectively, so PA’s fast growth on any single substrate is described as neutralized when all substrates are simultaneously contested, as in the k=3 formulation’s 100% niche coverage. This is an arithmetic illustration inside the source, not an experimental demonstration of community-level resource depletion. ₁

These results refine a single-score view of inhibition: endpoint growth, lag-time advantage, and direct antagonism can contribute distinct evidence, and a formulation should not be judged from any one assay alone. The condition-specific measurements do not demonstrate that the same advantages persist in a spatially structured airway biofilm. ₁

No selective prebiotic was found among the tested substrates

To ask whether a carbon source could selectively feed commensals and starve PA14, the report computed selectivity ratios (commensal mean OD / PA14 OD); the methods statement says these were computed for all 20 tested substrates. PA14 outgrew the average commensal on every tested substrate: the most selective substrates (cysteine, threonine, methionine) reached ratios of only 0.77–0.96, and there was no amino acid or simple sugar where commensals had a clear growth advantage. ₁

The report concludes that among the 22 tested substrates no selective prebiotic exists, and that competitive exclusion must instead work through community-level resource depletion — multiple organisms collectively consuming the resource pool faster than PA14 alone — rather than individual substrate advantage. This contradicts the selective-prebiotic hypothesis for the panel actually tested, while leaving it open for untested substrates. The substrate count is itself reported inconsistently inside the source (430 isolates × 21 substrates in the integration summary, 20 tested substrates in the selectivity methods, 22 tested substrates in this conclusion); the discrepancy is recorded, not reconciled. ₁

That gap motivated a genomic extension: pathway comparison nominates substrates that commensals can metabolize but PA14 cannot — the sugar alcohols xylitol and myoinositol, and the pentoses xylose and arabinose. The source’s integration summary groups all four under “sugar alcohols (xylitol, myoinositol, arabinose, xylose)”, whereas its detailed genomic section separates sugar alcohols (xylitol, myoinositol) from pentoses (xylose, arabinose); the detailed classification is used here. These are genomic predictions rather than demonstrated interventions, so the prebiotic component remains a hypothesis; it motivates computational-pathway-prediction-validation, which asks when predicted capacity is confirmed by measured growth. ₁

The genomic nomination is resolved at the pathway level by gapmind, a tool that scores the completeness of catabolic pathways from genome sequence. Six GapMind pathways are complete in at least one commensal species but absent in PA14, and each is carried by only 1–2 of the 5 core species, making them species-specific capabilities rather than consortium-wide ones: myoinositol, complete in 100% of R. dentocariosa against 0% of PA, selectivity 1.00; xylitol, 98% of S. salivarius and 100% of G. sanguinis against 0%, selectivity 1.00; xylose, 100% of N. mucosa and 100% of G. sanguinis against 0%, selectivity 1.00; arabinose, 100% of N. mucosa and 100% of G. sanguinis against 0%, selectivity 1.00; fucose, 100% of N. mucosa and 100% of G. sanguinis against 1% of PA, selectivity 0.99; and rhamnose, 100% of N. mucosa and 100% of G. sanguinis against 1% of PA, selectivity 0.99. ₁

Patient metatranscriptomics corroborates the genomic predictions: 47 KEGG pathways — KEGG being a reference database of metabolic pathways — show >2× commensal-to-PA expression ratio in vivo, dominated by PTS (phosphotransferase system) sugar transport systems for maltose, trehalose, and N-acetylmuramic acid that are exclusively commensal-expressed. This supports multi-omics-integration by adding expression-level agreement to a genome-derived prediction, while expression in patients is still not demonstrated selective growth. ₁

The report concludes that sugar alcohols (xylitol, myoinositol) and pentoses (xylose, arabinose, fucose, rhamnose) are strong prebiotic candidates on the grounds that they are genomically complete in the formulation species, absent from PA14’s metabolic repertoire, actively transported by commensals in patient airways, commercially available and FDA-GRAS, and — for xylitol — already used in cystic-fibrosis airway products for other indications, reframing the strategy from competing on PA14’s turf to feeding commensals on substrates PA14 cannot access. The report calls xylitol particularly attractive because it is already FDA-approved for cystic-fibrosis airway use for mucolytic/antimicrobial properties, which it reads as a dual-purpose prebiotic opportunity. Its absence claim concerns the PA14 strain; the 1% pathway completeness recorded for fucose and rhamnose refers to the broader PA genome collection and does not contradict PA14-specific absence. This remains a genomic and expression-based nomination rather than a demonstration of selective growth, and it refines rather than overturns the earlier null result that no selective amino acid prebiotic exists among the tested substrates, which the report attributes to PA14 being too metabolically versatile on amino acids. ₁

Because each nominated pathway sits in only one or two members, the report proposes species-targeted pairing — xylitol benefiting S. salivarius, the top inhibitor, while xylose, arabinose, and fucose benefit N. mucosa, the engraftment anchor, and G. sanguinis — and hypothesizes that a multi-prebiotic cocktail targeting different formulation members may be more effective than a single prebiotic. That is a prediction from pathway content, not a tested result. ₁

For delivery the report cites clinical precedent rather than its own data: Durairaj et al. (2007) demonstrated safety of inhaled xylitol in cystic-fibrosis patients, and Singh et al. (2020) conducted a randomized controlled trial of aerosolized hypertonic xylitol versus saline during cystic-fibrosis exacerbations, which the report reads as establishing the delivery route and safety profile its prebiotic strategy would use. ₁

Coverage, safety, and engraftability can oppose one another

The formulation analysis operationalized antagonism as a combination of metabolic coverage and measured inhibition. The multi-criterion scoring function evaluated formulations of 1–5 organisms on five criteria: PA14 niche coverage, the fraction of PA’s preferred substrates covered by at least one member; internal complementarity, the metabolic dissimilarity among members; mean inhibition; engraftability; and FDA safety. ₁

Optimization ran in two stages: permissive safety, excluding only well-known pathogens, and strict safety, additionally excluding all Pseudomonas, Enterobacteriaceae, and Staphylococcus. ₁ The permissive filter identified the organisms with the highest inhibition — Leclercia adecarboxylata (102%), Pseudomonas_E juntendi (94%), S. epidermidis (99%) — but the report flags all three as clinically problematic (Enterobacteriaceae, non-aeruginosa Pseudomonas, nosocomial Staphylococcus), so the eligible pool is fixed by safety rather than by the inhibition ranking. ₁ The report characterizes the strict filter as costing roughly 15% inhibition ceiling while doubling engraftability. ₁

Strict safety-filter optimization identified a five-species core of neisseria mucosa, streptococcus salivarius, micrococcus luteus, rothia dentocariosa, and gemella sanguinis. ₁

The best strict-safe formulations varied across objectives: the k=1 formulation containing N. mucosa [ASMA-3643] had 18% niche coverage, 88% inhibition, and engraftability 1.595 — the best single organism, combining the highest engraftability with strong inhibition, but with coverage too low for monotherapy; the k=2 formulation containing R. dentocariosa + N. mucosa had 18% coverage, 84% inhibition, and engraftability 0.820; the k=3 formulation containing M. luteus + N. mucosa + S. salivarius had 100% coverage, 75% inhibition, and engraftability 0.140; the k=4 formulation containing R. dentocariosa + M. luteus + N. mucosa + S. salivarius had 100% coverage, 76% inhibition, and engraftability 0.185; and the k=5 core had 100% coverage, 78% inhibition, and engraftability 0.188. ₁

The full strict-safety (FDA-safe) five-organism formulation — R. dentocariosa [ASMA-2935] + M. luteus [ASMA-2965] + G. sanguinis [ASMA-3044] + N. mucosa [ASMA-3643] + S. salivarius [ASMA-737] — is reported with composite 0.587, 100% coverage of PA14’s amino acid niche, 78% mean inhibition, and engraftability 0.188, and is described as the full formulation with maximum redundancy and inhibition depth. ₁

The report argues that the five-species core is consistent across all formulation sizes, indicating a robust solution rather than an optimization artifact, with exhaustive enumeration and bootstrap analysis further confirming the robustness of these rankings. This refines rather than removes the uncertainty recorded below: which species recur is stable across formulation size, while the composite scores for k=2 through k=5 are not statistically distinguishable. ₁

The k=3 formulation was the minimum size achieving complete PA14 niche coverage because M. luteus grew on 9 of PA14’s 11 preferred substrates; at k=3, 100% niche coverage means at least one member grows on every amino acid that PA14 can use. M. luteus itself had a measured best inhibition of 38%, an engraftability of 0.000, and no lung genomes in its species profile — these are M. luteus species-profile values, not formulation-level values, since the k=3 formulation is reported at 75% inhibition and engraftability 0.140. ₁ Thus niche overlap and direct inhibitory strength are distinct formulation criteria: M. luteus was central to complete coverage but was not a strong inhibitor or demonstrated engraftable species. This supports a coverage-versus-establishment interpretation of competitive-exclusion-consortium-design. ₁

Exhaustive enumeration of C(97,3) = 147,440 possible triples produced 127,598 valid unique-species formulations, and the k=3 combination was the global optimum with composite score 0.562. However, bootstrap resampling 1,000 times over the raw inhibition measurements — so the intervals express uncertainty in the inhibition inputs rather than in every scoring component — produced 95% composite-score confidence intervals of [0.753, 0.753] for k=1, [0.514, 0.588] for k=2, [0.551, 0.562] for k=3, [0.534, 0.578] for k=4, and [0.520, 0.587] for k=5; the k=2 through k=5 intervals overlapped. ₁

All k=2 through k=5 intervals overlap, so formulation sizes are not statistically distinguishable on the composite score, which the report takes as supporting the recommendation of k=2 as the practical primary candidate. ₁ The overlapping intervals therefore weaken any claim that larger formulations are statistically preferable on the composite score, even though larger formulations can improve niche coverage. The optimizer’s global k=3 result therefore identifies a candidate under the specified composite objective, not a validated community-level optimum. ₁

Engraftability is a prioritization proxy, not a measurement of establishment

Engraftability screening used paired metagenomic (DNA abundance) and metatranscriptomic (RNA activity) data from 175 patient samples to identify species that are both prevalent and metabolically active, with activity ratio = metaRS CPM / metaG CPM capturing transcriptional engagement per unit DNA — a computed proxy rather than observed therapeutic engraftment. ₁ Engraftability, defined as prevalence × log(activity ratio), was computed over 134 species detected in patient metagenomes; among inhibition-tested species, N. mucosa stands out with the highest engraftability at 1.595, combining high prevalence with strong transcriptional activity, while R. dentocariosa (0.422) and S. salivarius (0.172) are also above the median. ₁ The primary two-species formulation, R. dentocariosa + N. mucosa, had combined engraftability of 0.820, whereas the proposed three-species formulation, M. luteus + N. mucosa + S. salivarius, had engraftability of 0.140. ₁

The report recommends k=2, R. dentocariosa + N. mucosa, as the primary clinical candidate because both species are lung-adapted, provide 84% mean inhibition, and have combined engraftability 0.820, nearly 6× higher than the k=3 formulation’s 0.140. While k=2 achieves only 18% PA niche coverage, the report suggests that its high per-organism inhibition (84% mean) means direct antagonism may matter more than niche coverage for clinical efficacy — stated explicitly as a hypothesis testable in the mouse model, not as a demonstrated result, and clinical efficacy itself remains untested. ₁ This recommendation uses engraftability as one prioritization signal alongside inhibition and adaptation evidence, not as proof of post-administration establishment. ₁

The k=3 formulation is instead an aspirational second-line candidate contingent on demonstrating M. luteus engraftment in vivo: M. luteus had zero lung genomes and zero detected patient engraftability despite its central role in complete PA14 niche coverage. ₁ These results refine the optimizer’s interpretation: a species can be metabolically valuable in the design model without evidence that it establishes in the target airway environment. ₁

The report sets out three paths forward for this tension: (a) accept the risk and test k=3 in mice, since M. luteus’s broad carbon utilization may compensate for low natural prevalence if delivered at sufficient dose; (b) prioritize the k=2 formulation of R. dentocariosa + N. mucosa, both lung-adapted, with combined 84% inhibition and 0.820 engraftability, sacrificing niche coverage for colonization reliability; or (c) search for an alternative broad-spectrum metabolizer among lung-adapted species to replace M. luteus’s niche coverage role, screening oral and respiratory Actinobacteria with broad carbon utilization profiles — Kocuria, Dermacoccus, or other Micrococcaceae with respiratory isolation records — against the PA14 carbon utilization panel. Only the first path depends on M. luteus: the report pursues the k=3 formulation as an aspirational second-line candidate contingent on demonstrating M. luteus lung engraftment in vivo, so successful k=3 engraftment remains speculative, whereas paths (b) and (c) are routes that do not require it. ₁

Prevalence captures how often a species is detected across sampled patients, and the activity ratio captures its transcriptional activity relative to the comparison context, but neither quantity measures recovery of an administered strain after dosing. ₁ The score can reflect ecological compatibility or prior persistence in the sampled population, but cannot by itself distinguish successful engraftment from transient passage, repeated exposure, unequal sampling, resident strains, or activity caused by a short-lived response. The study inferred engraftability from patient prevalence and transcriptional activity rather than measuring it after administration. ₁

Only 21 lung genomes across 5 species were available for lung-adaptation comparisons, limiting direct genomic evidence for airway adaptation of the candidate set. ₁ The evidence therefore supports the hypothesis that higher prevalence and activity may improve the probability of engraftment, rather than an established relationship between the score and post-administration establishment. The proxy is most defensible as one component of a multi-criteria decision that also includes inhibition, niche coverage, pathway conservation, safety filtering, and direct testing in airway-relevant conditions. ₁

This supports condition-specific-fitness, because candidate performance depends on airway substrate environment and assay condition rather than prevalence alone. It also motivates multi-omics-integration: metagenomic detection and metatranscriptomic activity should be combined with longitudinal strain-resolved measurements rather than treated as equivalent to engraftment. The gap is especially important for micrococcus luteus, whose predicted contribution to niche coverage was strong but whose patient and lung-genome evidence for establishment was absent. ₁

Pairwise evidence does not establish community-level effects

Pairwise measurements further support a mixed-mechanism interpretation but remain provisional. Mean synergy scores were +5.3% for N. mucosa + ASMA-2260, which showed 47% mean pair inhibition against 42% best-single inhibition; +1.4% for ASMA-3913 + ASMA-2260, at 52% mean pair against 51% best single; −2.2% for N. mucosa + ASMA-2464, at 40% mean pair against 42% best single; −14.2% for ASMA-3913 + ASMA-2464, an antagonistic pair at 37% mean pair against 51% best single; and −19.8% for ASMA-1478 + ASMA-1197, the strongest antagonism, at −3% mean pair inhibition against 17% best single. Overall mean synergy was −5.8% across only 8 comparisons and 5 unique pairs. ₁ The report reads this pattern as near-additive inhibition for N. mucosa combinations with modest antagonism in some pairs, used to inform which combinations to advance to mouse models. ₁

The study analyzed 8 pairwise comparisons across 5 unique pairs, including 3 A × 3 B isolate combinations, rather than the complete interaction structure of the proposed five-species core. ₁ The complete 10-pair interaction matrix for the five-species core had not been measured. ₁ Accordingly, a positive or near-additive result for one pair does not establish that the same species will remain compatible in a larger consortium, while a negative pairwise interaction may reflect isolate-specific or condition-specific behavior rather than universal species-level incompatibility. Additive formulation scoring was therefore provisional. ₁

The scoring function explicitly assumes additive inhibition — a combination’s effect equals the mean of its members — so member synergy would mean the score underestimates and antagonism would mean it overestimates. The source’s own account of how many pairs were tested is inconsistent: the assay rationale states that 3 commensal pairs were tested against PA14 at multiple inoculation densities, while the results table and conclusion refer to 8 comparisons across 5 unique pairs. Both figures are recorded without reconciliation, which supports adversarial-research-quality-assurance: the additive assumption rests on evidence whose extent the report itself describes in two incompatible ways. ₁

This refines cofitness-network-architecture: combining individually inhibitory organisms does not guarantee additive or synergistic community inhibition, and sparse pairwise edges cannot determine a larger interaction network when unmeasured edges may alter community behavior. It also supports the constraint that niche coverage and inhibition scores do not by themselves establish coexistence or joint suppression. ₁

The proposed five-species core comprised N. mucosa, Streptococcus salivarius, Micrococcus luteus, Rothia dentocariosa, and Gemella sanguinis, but its complete 10-pair interaction matrix had not been measured. ₁ This is especially consequential because the optimizer selected M. luteus + N. mucosa + S. salivarius as the global optimum among 127,598 valid unique-species formulations, while available pairwise measurements did not provide full interaction coverage for that community. ₁

The strict-safe two-species candidate, R. dentocariosa plus N. mucosa, provided 84% mean inhibition and combined engraftability of 0.820. Candidate isolates included strong planktonic performers, including 88% best inhibition for N. mucosa and 79% for R. dentocariosa. These measurements support the primary-candidate recommendation but do not establish protection against established airway biofilms. ₁

Measurement, representation, and translation limits

The pairwise database contained identical fact_pairwise_interaction and fact_carbon_utilization values, with correlation = 1.0 and mean difference = 0.0, so endpoint OD data could not assess per-substrate co-culture effects. ₁ Consequently, interaction conclusions relied on the RFU-based competition assay rather than endpoint-OD co-culture measurements. This supports interpreting the pairwise component as evidence of competition under the RFU assay, not as a complete per-substrate co-culture experiment. ₁

The interaction assays used planktonic cultures and PA14, whereas Pseudomonas aeruginosa in cystic-fibrosis lungs primarily occupies structured biofilms. ₁ Thus, even a more complete planktonic pairwise matrix would not establish multispecies performance in the relevant biofilm context. Strong planktonic candidate performance and metabolic prediction contradict neither the biofilm concern nor each other: the former is a direct assay result, while the latter identifies an unmeasured translation gap. ₁

The carbon panel contained 22 substrates and omitted mucins, lipids, iron, polyamines, and the sugar alcohols identified genomically. Consequently, measured metabolic overlap may not capture resource competition or direct antagonism in airway biofilms. ₁ The formulation report combined pairwise interactions with carbon utilization, growth kinetics, patient metagenomics, metatranscriptomics, and pangenome analysis, but the pairwise component remained the least complete evidence for assigning additive or synergistic community effects. This is a case for multi-omics-integration in which complementary data improve candidate selection but do not replace direct measurement of higher-order interactions. ₁

The metabolic model covered 142 isolates representing 62 of 211 species (29%), and growth kinetics were available for only 32 isolates. The core cohort was enriched for deeply characterized taxa, including Rothia, Streptococcus, Neisseria, and Gemella. Pairwise interaction data covered only 3 A × 3 B isolate combinations and 8 comparisons across 5 unique pairs; the complete 10-pair interaction matrix for the five-species core had not been measured. ₁

The report recommends testing PAO1 and 3–5 mucoid clinical PA isolates because PA14-based inhibition measurements have not been validated against PAO1 or ExoS+ clinical strains, even though pangenome analysis found no amino-acid pathway differences between ExoU+ and ExoS+ PA. ₁ Genomic pathway conservation and patient transcriptional activity can refine candidate selection but do not demonstrate spatial colonization, biofilm competition, or protection from PA in vivo. ₁

Direct interactions, robustness, and safety-constrained optimization

The strict-safe optimization filtered candidate species before ranking formulations, so the reported optima represent tradeoffs within a safety-constrained candidate set rather than unconstrained maximization of inhibition. Species-level pathway conservation across 499 genomes supported robustness checks: M. luteus showed 18/18 amino-acid pathways and 39/39 carbon pathways conserved at >95%; S. salivarius showed 18/18 and 32/35; R. dentocariosa showed 14/18 and 39/41; N. mucosa showed 16/16 and 27/27; and G. sanguinis showed 7/18 and 37/39, respectively. ₁

These observations support pangenome-integration because conservation across genomes can test whether a proposed metabolic role is reproducible within a species, while not replacing direct safety or engraftment testing. They also refine the optimizer: pathway conservation can strengthen confidence in predicted capacity, but cannot supply missing evidence about higher-order compatibility, post-administration establishment, or biofilm performance. ₁

Core species profiles and lung adaptation

Per-species profiles report, in order, best inhibition, computed engraftability, pangenome size, conserved amino-acid pathways, and lung genomes: neisseria mucosa 88%, 1.595, 15 genomes (8 in PROTECT clade), 16/18 (18/18 in PROTECT clade), 5 (33%); streptococcus salivarius 98%, 0.172, 153 genomes, 18/18, 5 (4%); rothia dentocariosa 79%, 0.422, 29 genomes, 14/18, 10 (38%). ₁

S. salivarius carries the highest individual inhibition among the strict-safe core species (98%) and an established probiotic precedent (BLIS K12), and the report assigns it a dual mechanism — a residual +74% beyond metabolic prediction — calling it the direct antagonism backbone. The comparison holds only within the safety-eligible pool, since permissive-filter candidates excluded on safety grounds were reported at higher inhibition (for example Leclercia adecarboxylata at 102% and S. epidermidis at 99%). ₁

R. dentocariosa is described as lung-adapted (38% respiratory), strongly inhibitory (79%), and dual-mechanism (residual +57%), with published evidence for colonization resistance via a secreted endopeptidase (Stubbendieck et al. 2023). The report situates this alongside Rigauts et al. (2022) on Rothia mucilaginosa enrichment in healthy airways and its suppression of NF-κB inflammation in cystic-fibrosis airways, and Stubbendieck et al. (2023) on R. dentocariosa-mediated colonization resistance in the nasal tract via a peptidoglycan endopeptidase that inhibits moraxella catarrhalis — external context, not airway evidence generated here. The report proposes that its own R. dentocariosa isolates likely employ similar mechanisms, which would explain their high inhibition residuals; that mechanism is extrapolated from the cited M. catarrhalis result and was not demonstrated against PA here. ₁

The residual values attached to the species profiles do not line up cleanly with the isolate-level residual ranking: the ranking lists S. salivarius ASMA-737 at +74.1%, G. sanguinis ASMA-3044 at +62.2%, and N. mucosa ASMA-3643 at +57.2%, whereas the species profiles give +74% for S. salivarius, +62% for G. sanguinis, and +57% for R. dentocariosa. Both attributions are recorded as the source states them; which species the +57% residual belongs to is not resolved here. ₁

gemella sanguinis is reported with strong inhibition (85%), a dual mechanism (residual +62%), and a role adding depth at k=5, but it has the smallest pangenome (7 genomes), which the report flags as making strain selection most critical for this species. This supports sample-size-aware-phenotype-consensus: a species-level claim resting on 7 genomes carries less weight than one resting on 153. ₁

Of 21 lung/respiratory genomes identified, R. dentocariosa contributes 10 (38% of its species) and N. mucosa contributes 5 (33%), which the report reads as these two being naturally respiratory organisms rather than gut commensals being repurposed; micrococcus luteus has 0 lung genomes and is primarily skin/environmental, suggesting it may face engraftment challenges despite its metabolic contribution. Lung-adapted S. salivarius genomes show enrichment for L-malate (+0.39 score) and depletion for sorbitol (−0.82), suggesting metabolic adaptation to the airway carbon landscape. This supports cultivation-collection-bias-in-ecological-genomics, because habitat assignment here comes from deposited genome metadata rather than sampling designed to represent the airway. ₁

The report summarizes these metabolic capabilities as species-level traits (>95% conservation across hundreds of genomes) and the two anchor species, R. dentocariosa and N. mucosa, as naturally lung-adapted, with 33–38% of pangenome genomes from respiratory sources. ₁

PA-side invariance bounds what the formulation can target

The pathogen’s amino acid metabolism varies little across lung isolates. Across 1,796 lung PA genomes, amino acid catabolic pathways are 97.4% conserved by mean completeness, and even proline — PA14’s top substrate — is complete in 97% of lung isolates; the main variation among lung PA (PC1 = 79%) lies in carbon source pathways (sorbitol, mannitol, gluconate), substrates the formulation does not target. This supports within-species-conservation-between-species-functional-divergence: the competitive targets chosen here fall in the highly conserved rather than the variable part of the pathogen’s metabolism, with conservation quantified as a 97.4% mean completeness and not as literal invariance. ₁

Read as a trajectory, lung PA genomes lose sugar utilization pathways (sorbitol, mannitol, gluconate) while amino acid catabolism remains comparatively conserved. The report treats this as confirming that competitive exclusion targets are evolutionarily stable; that stability is an inference from cross-sectional genome comparison rather than a measured evolutionary result. ₁

Cystic-fibrosis-derived PA shows 6 FDR-significant differences from non-CF lung PA — FDR being the false discovery rate, the expected proportion of false positives among calls declared significant — but all are in sugar-related pathways, with zero amino acid differences. The report infers from this null result that formulations designed for general lung PA should work equivalently for cystic-fibrosis PA; the equivalence is inferred from genome content, not measured in assays. ₁

The same pattern holds across virulence genotypes: amino acid catabolic pathways are identical between ExoU+ and ExoS+ PA, with GapMind score differences <0.03 on a 1–5 scale and zero FDR-significant pathways. This refines the report’s own recommendation to test PAO1 and ExoS+ clinical strains — the genomic case for transferability is strong while assay-level validation is still missing. ₁

Across strains, all 15 PROTECT strain groups retain the same core amino acid pathways at 97%+ conservation despite genome sizes ranging 6.1–7.4 Mb and 24% gene content variation, which the report reads as favorable for design: it predicts the formulation should be equally effective across PA variants because the metabolic competition targets do not vary, with strain-level variation expected to matter for virulence, biofilm formation, and antibiotic resistance instead. Equal efficacy across variants is a prediction, not a measured outcome. ₁

The report’s reasoning for discounting the extra gene content is that, since all strain groups retain the same amino acid catabolic pathways (97%+ conservation), the additional genes are likely conditionally expressed accessory functions — prophages, mobile elements, niche-specific adaptations — that affect virulence and persistence rather than amino acid growth rate, and that within-species genome size variation is a much weaker predictor of growth rate than cross-species variation (Vieira-Silva & Rocha 2010). This supports a two-speed reading of the PA genome (two-speed-bacterial-genome), while the predicted similarity of strain responses to the formulation remains untested. ₁

PA’s expression state may matter as much as its genome. Patient metatranscriptomics shows that during acute exacerbation PA downregulates 170 of 207 measured pathways, including amino acid biosynthesis (cysteine, methionine: 10× lower) and imipenem resistance (OprD: 13× lower), with only phosphate transport upregulated (64×). The report reads sick PA as metabolically quiescent and dependent on scavenging host-derived amino acids, and therefore potentially more vulnerable to competitive exclusion during acute episodes — a stated possibility rather than a tested one, and one that motivates timing an intervention to disease state. ₁

External context and probiotic precedent

The ecological framing draws on prior work rather than on this corpus: Widder et al. (2022) identified eight pulmotypes driven by ecological competition, which the report’s community-level niche coverage model is said to operationalize for therapeutic design, and Rogers et al. (2015) demonstrated competitive exclusion between PA and H. influenzae in bronchiectasis, cited as support for the ecological framework. ₁

On precedent, Anderson et al. (2017) reviewed cystic-fibrosis probiotic trials and found suggestive but inconclusive results, attributing this partly to a lack of rational strain selection — the gap this multi-criterion optimization is positioned to address — while S. salivarius BLIS K12 is cited as having established respiratory probiotic credentials (Tagg et al. 2025; Burton et al. 2011). ₁

The safety filter is not a formality: Drost & Fuhrmann (2025) showed that lactiplantibacillus-plantarum, a commonly proposed probiotic, actually enhanced PA pathogenicity in a cystic-fibrosis sputum–epithelial cell model. This supports treating safety as an independent objective rather than assuming any commensal probiotic is harmless in the airway. ₁

Weyand (2017) reviewed Neisseria colonization models in the upper respiratory tract, noting N. mucosa’s prevalence in oral tissue. This is consistent with the high engraftability proxy assigned to neisseria mucosa but, like the proxy itself, does not establish persistence after probiotic administration. ₁

Tensions and boundary conditions

The formulation ranking favors N. mucosa and R. dentocariosa because their prevalence-activity proxies are relatively high, while the complete-coverage design favors inclusion of M. luteus despite zero detected patient engraftability and zero lung genomes. ₁ This is a decision tension between predicted metabolic coverage and evidence of airway establishment, not a statistical contradiction between measurements. ₁

The tension should not be resolved by treating the composite formulation score as an engraftment measurement: its bootstrap intervals addressed composite-score uncertainty rather than post-administration colonization. ₁ Similarly, there is no direct disagreement between strong planktonic inhibition and the sparse-interaction warning. They answer different questions: inhibition measurements establish candidate performance in the tested assay, whereas incomplete pairwise data limit extrapolation to the full consortium. The planktonic assays and intended biofilm application likewise define a translation boundary rather than a contradiction. ₁

The formulation optimizer selected a global k=3 optimum with composite score 0.562, while bootstrap intervals for k=2 through k=5 overlapped and the complete interaction matrix for the proposed core was unavailable. ₁ These findings should not be reconciled by treating the k=3 formulation as either disproven or validated: optimization supports prioritization under the modeled objectives, while uncertainty, proxy-based engraftability, and sparse interaction evidence qualify its community-level interpretation. ₁

Design principle

A defensible consortium-ranking framework should retain separate objectives for pathogen inhibition, resource-niche coverage, safety eligibility, and engraftability, while representing uncertainty in pairwise and higher-order compatibility and distinguishing inferred engraftability from measured establishment. It should report the tradeoff surface rather than treating one composite score as definitive. This extends multi-omics-integration because the objectives draw on inhibition assays, carbon utilization, growth kinetics, patient metagenomics, metatranscriptomics, and pangenome pathway comparisons. ₁

For this case, the evidence favors R. dentocariosa + N. mucosa as the primary candidate because it combines 84% mean inhibition with combined engraftability 0.820, while the M. luteus-containing k=3 formulation remains valuable as a coverage-maximizing hypothesis rather than a clinically established solution. ₁ The available data support the hypothesis that pairwise competition can contribute to multispecies formulation design, but do not establish that pairwise synergy or antagonism is transferable to the full consortium. The strongest defensible conclusion is therefore candidate prioritization, not validated prediction of community-level suppression or formulation engraftment. ₁

Open Directions

• Measure the complete 10-pair interaction matrix for the five-species core with RFU-based competition assays and substrate-resolved co-cultures, including spatially structured co-cultures, to determine whether effects are additive, synergistic, or antagonistic and whether observed edges predict the k=3 and k=5 formulation outcomes. ₁
• Repeat the complete pairwise matrix and candidate formulations in structured biofilm models, asking whether planktonic synergy scores predict biofilm inhibition of P. aeruginosa. ₁
• Repeat inhibition and carbon-utilization assays with PAO1 and 3–5 mucoid clinical PA isolates, then test the k=2 and k=3 formulations in mixed-species biofilm models to determine whether planktonic inhibition predicts biofilm biomass reduction or PA exclusion. ₁
• Separate per-substrate co-culture effects from endpoint carbon-utilization measurements by generating substrate-resolved co-culture data, asking whether competition is additive, synergistic, or antagonistic on the 22 tested substrates. ₁
• Test biofilm models using mucins, lipids, iron, polyamines, xylitol, myoinositol, xylose, arabinose, and other genomically nominated sugar alcohols and pentoses; compare planktonic and biofilm phenotypes to quantify which condition-specific fitness measurements transfer. ₁
• Expand the isolate-by-isolate interaction design beyond the 3 A × 3 B combinations and use a hierarchical model to test whether interaction effects are species-level, isolate-specific, or condition-specific. ₁
• Track formulation species by strain-resolved metagenomics and metatranscriptomics during biofilm growth or an airway-relevant model, and administer the k=2 and k=3 formulations in an in vivo engraftment model, to test whether inferred engraftability predicts actual establishment and activity. ₁
• Administer barcoded or otherwise strain-resolvable candidate organisms in an airway-relevant model, then use longitudinal quantitative metagenomics to test whether the prevalence × log(activity ratio) score predicts recovery, persistence, and dose-normalized abundance. ₁
• Measure candidate abundance and transcription before administration and at multiple post-administration time points in cystic-fibrosis airway samples to distinguish transient detection from sustained engraftment. ₁
• Generate additional lung genomes and strain-resolved metagenomes for N. mucosa, R. dentocariosa, S. salivarius, G. sanguinis, and M. luteus to test whether genomic lung adaptation improves prediction beyond prevalence and transcriptional activity. ₁
• Compare proxy scores with measured engraftment using mixed-effects or survival models incorporating dose, sampling time, patient, formulation, and airway environment, thereby testing whether the score predicts establishment independently of these factors. ₁
• Expand the matched isolate cohort and use held-out validation to determine whether genus-level residuals predict direct antagonism after controlling for metabolic overlap, growth kinetics, and phylogenetic relatedness. ₁
• Combine metabolite profiling, transcriptomics, and targeted inhibition assays for S. salivarius ASMA-737, G. sanguinis ASMA-3044, and N. mucosa ASMA-3643 to identify mechanisms underlying their +74.1%, +62.2%, and +57.2% positive residuals. ₁
• Experimentally test xylitol, myoinositol, xylose, and arabinose supplementation with the candidate commensals and PA, asking whether predicted pathway gaps create selective commensal growth without increasing PA growth. ₁
• Recompute the multi-objective ranking with uncertainty intervals for inhibition, coverage, safety, and measured engraftment, asking whether the recommended formulation remains optimal when inferred objectives are replaced by direct measurements. ₁
• Test the species-targeted prebiotic pairing directly — xylitol with S. salivarius, xylose, arabinose, or fucose with N. mucosa and G. sanguinis — and compare a multi-prebiotic cocktail against single substrates, asking whether per-species pathway completeness predicts selective commensal growth. ₁
• Screen oral and respiratory Actinobacteria with broad carbon utilization, including Kocuria, Dermacoccus, and other Micrococcaceae with respiratory isolation records, against the PA14 carbon utilization panel to test whether a lung-adapted replacement for M. luteus can preserve 100% niche coverage. ₁
• Measure formulation inhibition against PA in an exacerbation-like quiescent state, in which 170 of 207 measured pathways are downregulated, to determine whether the proposed increase in vulnerability to competitive exclusion is real. ₁