Minimizing caging effects in murine lung microbiome studies

to the editor: We read with interest the work by Abdelgawad et al. (1) regarding the role of Toll-like receptors (TLRs) in shaping the pulmonary microbiome. We offer a clarification regarding our work (2, 3) that was addressed in the Editorial Focus and provide insights into strategies we have used to address the concern that we share with Abdelgawad et al. that environmental factors such as cohousing are sources of potential bias in the study of microbial communities in mouse lungs.

First, our inhaled Pam2 + oligodeoxynucleotide (ODN) therapy is composed of synergistic agonists of the TLR2/6 heterodimer and the TLR9 homodimer (4–6). It is incorrectly identified in the Editorial Focus as an inhibitor of TLRs 2, 6, and 9, potentially impacting readers’ understanding of our work.

On the other hand, we enthusiastically endorse their critical appraisal of caging effects as potential experimental confounders. Many environmental factors (e.g., vendor, shipment, cohousing, bedding, diet, and littermates) influence relationships between different exposures and gut microbiota (7, 8). Similarly, caging effects can alter causal relationships between an exposure of interest (E) and lung microbiota as an outcome (O) in healthy mice (Fig. 1A). Baseline lung microbial communities of genetically identical mice are most similar among previously cohoused mice and most dissimilar as increasing numbers of environmental factors differ upon arrival, with these baseline differences converging over time when mice are cohoused after initial delivery (9). Similarly, baseline differences in lung microbial communities of mice with different TLR knockout mutations converged when mice are randomly cohoused (10).

Figure 1.

Caging effects in mouse lung microbiome studies. A: schematic representation of direct and indirect housing effect modifications in causal pathways (9, 10). B: temporal associations of cohousing and experimental events. C: use of stratified random cohousing for modifiable (left) and nonmodifiable (right) exposures (2, 10). P + O, Pam2+ODN; t, time; TLR, Toll-like receptor; WT, wild type; X₁, direct effect modifications; X₀, indirect effect modifications.

These observations underscore the importance of caging effects in lung microbiome studies. Both indirect effect modifications (X₀) of the lung microbiota resulting from prior cohousing and direct effect modifications (X₁) from postdelivery cohousing emphasize the need to minimize caging effects using randomization and stratification strategies.

In our studies that determined that TLR manipulation did not durably alter the healthy lung microbiome, we intentionally sought to mitigate caging effects a priori by incorporating stratified random cohousing strategies in our experimental designs (Fig. 1B; 2, 3). In stratified random strategies, originally cohoused mice are randomly distributed for caging with mice from all experimental groups. Then, additional stratified randomization to new cages is performed following each intervention.

When interrogating the effects of a modifiable exposure, as in the case of our TLR treatment, a stratified random cohousing strategy is applied both preexposure and postexposure to minimize both X₀ and X₁ caging effects identified in lung microbiome studies (Fig. 1C, left). However, in studies of a nonmodifiable exposure, as in the case of genetic TLR mutations, only postexposure stratified random cohousing can be implemented to minimize X₁ caging effects (Fig. 1C, right), with the assumption that X₀ effects are also minimized (11). This assumption can be experimentally verified, if needed, by concurrently studying the effect of the nonmodifiable exposure without stratified cohousing (10).

Although current microbiome research guidelines seek to prevent confounding in animal and human studies during experimental design (12, 13), bioinformatic batch effect adjustment methods cannot account for cage effects if the exposures are confounded by the cages (e.g., multicollinearity) during data analysis (14, 15). Thus, we agree with Abdelgaward et al. (1) that mitigating caging effects is critical to high-quality lung microbiome investigations, and we propose stratified random cohousing strategies as validated means to minimize the potential for cohousing-related confounding.

GRANTS

This work was supported by ConTex Postdoctoral Fellowship 2019 & 2020, CONACYT, and University of Texas System (to J.P.G.); the Fellowship is a joined effort from both CONACYT and University of Texas System. The Fellowship is not part of the CONACYT Sistema Nacional de Investigadores program. This work was also supported by National Heart, Lung, and Blood Institute (NHLBI) Grant R01 HL144599 (to R.P.D.) and NHLBI Grants R01 HL117976, DP2 HL123229, and R35 HL144805 (to S.E.E.).

DISCLOSURES

S.E.E. is an author on U.S. patent 8,883,174 “Stimulation of Innate Resistance of the Lungs to Infection with Synthetic Ligands” and owns stock in Pulmotect Inc., which holds the commercial options on these patent disclosures. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

J.P.G., R.P.D., and S.E.E. prepared figures; J.P.G., R.P.D., and S.E.E. drafted manuscript; J.P.G., R.P.D., and S.E.E. edited and revised manuscript; J.P.G, R.P.D., and S.E.E. approved final version of manuscript.