SM-MBIO210566 1..8 - mBio.01763-21.pdf

Model Systems to Study the Chronic, Polymicrobial Infections in

Cystic Fibrosis: Current Approaches and Exploring Future

Directions

George A. O

’

Toole

Aurélie Crabbé

Rolf Kümmerli

John J. LiPuma

Jennifer M. Bomberger

Jane C. Davies

Dominique Limoli

Vanessa V. Phelan

James B. Bliska

William H. DePas

Lars E. Dietrich

Thomas H. Hampton

Ryan Hunter

Cezar M. Khursigara

Alexa Price-Whelan

Alix Ashare

Robert A. Cramer

Joanna B. Goldberg

Freya Harrison

Deborah A. Hogan

Michael A. Henson

Dean R. Madden

Jared R. Mayers

Carey Nadell

Dianne Newman

Alice Prince

Damian W. Rivett

Joseph D. Schwartzman

Daniel Schultz

Donald C. Sheppard

Alan R. Smyth

Melanie A. Spero

Bruce A. Stanton

Paul E. Turner

Chris van der Gast

Fiona J. Whelan

Rachel Whitaker

Katrine Whiteson

Geisel School of Medicine at Dartmouth, Hanover, New Hampshire, USA

Ghent University, Ghent, Belgium

University of Zurich, Zurich, Switzerland

University of Michigan, Ann Arbor, Michigan, USA

University of Pittsburgh, Pittburgh, Pennsylvania, USA

Imperial College London and Royal Brompton Hospital, London, United Kingdom

University of Iowa, Iowa City, Iowa, USA

University of Colorado, Anschutz Medical Campus, Aurora, Colorado, USA

Columbia University, New York, New York, USA

University of Minnesota, Minneapolis, Minnesota, USA

University of Guelph, Guelph, Canada

Emory University School of Medicine, Atlanta, Georgia, USA

University of Warwick, Coventry, United Kingdom

University of Massachusetts and the Institute for Applied Life Sciences, Amherst, Massachusetts, USA

Brigham and Women

’

s Hospital and Harvard University, Boston, Massachusetts, USA

Dartmouth College, Hanover, New Hampshire, USA

Caltech, Pasadena, California, USA

Manchester Metropolitan University, Manchester, United Kingdom

McGill University, Montreal, Canada

University of Nottingham, Nottingham, United Kingdom

University of Oregon, Eugene, Oregon, USA

Yale University, New Haven, Connecticut, USA

University of Illinois, Urbana-Champaign, Illinois, USA

University of California, Irvine, California, USA

ABSTRACT

A recent workshop titled

“

Developing Models to Study Polymicrobial

Infections,

”

sponsored by the Dartmouth Cystic Fibrosis Center (DartCF), explored

the development of new models to study the polymicrobial infections associated

with the airways of persons with cystic

fi

brosis (CF). The workshop gathered 35

1

investigators over two virtual sessions. Here, we present the

fi

ndings of this work-

shop, summarize some of the challenges involved with developing such models, and

suggest three frameworks to tackle this complex problem. The frameworks proposed

here, we believe, could be generally useful in developing new model systems for

other infectious diseases. Developing and validating new approaches to study the

complex polymicrobial communities in the CF airway could open windows to new

therapeutics to treat these recalcitrant infections, as well as uncovering organizing

principles applicable to chronic polymicrobial infections more generally.

Citation

’

Toole GA, Crabbé A, Kümmerli R,

LiPuma JJ, Bomberger JM, Davies JC, Limoli D,

Phelan VV, Bliska JB, DePas WH, Dietrich LE,

Hampton TH, Hunter R, Khursigara CM, Price-

Whelan A, Ashare A, Cramer RA, Goldberg JB,

Harrison F, Hogan DA, Henson MA, Madden

DR, Mayers JR, Nadell C, Newman D, Prince A,

Rivett DW, Schwartzman JD, Schultz D,

Sheppard DC, Smyth AR, Spero MA, Stanton

BA, Turner PE, van der Gast C, Whelan FJ,

Whitaker R, Whiteson K. 2021. Model systems

to study the chronic, polymicrobial infections

in cystic

fi

brosis: current approaches and

exploring future directions. mBio 12:e01763-21.

https://doi.org/10.1128/mBio.01763-21

Editor

Matthew R. Parsek, University of

Washington

’

Toole et al. This is an

open-access article distributed under the terms

of the

Creative Commons Attribution 4.0

International license

Address correspondence to George A. O

’

Toole,

georgeo@dartmouth.edu.

Published

September/October 2021 Volume 12 Issue 5 e01763-21

®

mbio.asm.org

OPINION/HYPOTHESIS

September

2021

Downloaded from https://journals.asm.org/journal/mbio on 08 November 2021 by 131.215.71.167.

KEYWORDS

cystic

fi

brosis, models, polymicrobial, airway, chronic infection

n 25 to 26 May 2021, over 35 investigators met virtually in a workshop titled

“

Developing Models to Study Polymicrobial Infections,

”

sponsored by the

Dartmouth Cystic Fibrosis Center (DartCF [

https://sites.dartmouth.edu/dartcf/

]). The

investigators who participated, including 30 Ph.D.s, 6 M.D.s, and an M.D./Ph.D., learned

of the meeting via direct email invitation, sharing of the emails, and word of mouth.

The initial size of the conference had been limited based on the available funding pro-

vided by a grant from the Cystic Fibrosis Foundation; however, the onset of the coro-

navirus disease 2019 (COVID-19) pandemic and the switch of the workshop to a virtual

format allowed us to open the event to a larger and more geographically diverse pop-

ulation. The goal of the workshop was to discuss current model systems for studying

polymicrobial infections in the airways of persons with cystic

fi

brosis (pwCF), to con-

sider their strengths and weaknesses, and to envision how new, improved models

could be developed. The workshop was held over 2 days. On each day there were

breakout sessions; the goal of each session was to identify 3 to 5 key concepts or ideas,

followed by the entire group gathering to discuss the breakout group

fi

ndings.

WHAT IS THE GOAL?

Maybe the

fi

rst question to ask is,

“

What is the goal of building new and better

model systems to study CF-associated polymicrobial infections?

”

The simple answer is

“

To help pwCF

”

; but how? Do we want to determine how to keep microbes from

fl

our-

ishing and what makes them thrive? That is, we could use these new models as drug

discovery platforms. Do we want to eradicate the communities or maintain the com-

munities but keep them stable? Are we trying to better understand microbial commu-

nity function, and do we want to uncover general rules that can be applied to CF-rele-

vant infections, and beyond? Do we want to build predictive models that will help

clinicians to know when a pwCF will take a turn for the worse and to intervene earlier

and more successfully? And will the new wave of CFTR-targeted therapeutics impact

what it means to build effective models? These points and others were addressed dur-

ing the workshop.

Three overarching themes emerged during the course of the workshop. The

fi

rst

theme is,

“

What is your question?

”

This query is an excellent

fi

rst step in designing any

research project, and here, it speci

fi

cally means that the particular scienti

fi

c problem

driving the investigator will likely dictate which model system they select. That is, in

some cases a relatively simple

in vitro

model might suf

fi

ce, versus a complex animal

model. This

fi

rst question is related to the second, a variation on one of our favorite axi-

oms,

“

All models are wrong, but some are useful,

”

often attributed to statistician

George Box, but generally applicable across all of science. Here, the phrase can be

taken to mean that no model system we develop to study chronic polymicrobial infec-

tions in the airways of pwCF will capture the true complexity or spatial/temporal heter-

ogeneity of the CF airway. Finally, and of a practical nature, is the idea that there is not

“

”

model system to study polymicrobial airway infections for pwCF, because there is

not

“

”

single microbiome associated with this disease. It will almost de

fi

nitely be the

case that a set of models will need to be developed (i.e., a microbial community domi-

nated by one pathogen versus an evenly distributed collection of microbes) to encom-

pass most/all of the variety of CF-associated airway infection communities.

Below we present three frameworks focused on developing models to study poly-

microbial airway infections in CF; these frameworks are equally relevant to other infec-

tious diseases. The

fi

rst framework is to achieve the

“

perfect

”

model system(s). The sec-

ond and third frameworks consider that on the road to the perfect model, there can be

the development of model systems which move beyond single-culture experiments in

lysogeny broth (LB) but which capture meaningful information of at least key aspects

“

reality.

”

The second framework takes a question-driven approach, and the third

considers model systems as hypothesis generators. In this piece, we hope to convince

Opinion/Hypothesis

®

September/October 2021 Volume 12 Issue 5 e01763-21

mbio.asm.org

Downloaded from https://journals.asm.org/journal/mbio on 08 November 2021 by 131.215.71.167.

the reader that taking steps on the road toward the perfect model systems is a useful,

and we would say critical, exercise in working toward our goal to effectively reduce the

infection burden of pwCF. Finally, the very fact that we could gather such an illustrious

group of investigators from around the United States and the world in one (virtual)

place and at one time speaks to the dedication and engagement of the larger CF

research community toward this goal.

FRAMEWORK 1: WORKING TOWARD THE PERFECT MODEL SYSTEM(S)

Ideally, researchers would have access to a set of

“

do-it-all

”

model systems that cap-

ture much/all of the complexity of the

in vivo

environment. This is a lofty goal, and this

approach will get very complex, very quickly. Here, we put practicality aside and dis-

cuss the factors we should consider when building

“

perfect

”

model systems.

As part of this framework, it is critical to note upfront that CF is not a singular dis-

ease or pathological condition. Rather, as a chronic, lifelong (and life-shortening) con-

dition, its features (airway microbiota, airway environment, and host) are constantly

changing and evolving. To model polymicrobial infection in CF, we need to

fi

rst ask,

“

what CF disease are we trying to model?

”

In the short term, pwCF cycle between peri-

ods of relative clinical stability, punctuated by intermittent, unpredictable episodes of

poorer health (i.e., exacerbations) that may re

fl

ect acute changes in the airway micro-

biota, environment, and/or host. Over longer periods, as CF lung disease inevitably

progresses, the airway environment changes with concomitant changes in community

composition and activity, which can be (admittedly arbitrarily) de

fi

ned as disease

stage: early, intermediate, and advanced. Finally, aggressiveness describes the pace of

disease progression, which seems to be an innate feature of the host, who can have a

mild, moderate, or severe phenotype. Of interest, this phenotype is only roughly

related to CFTR genotype, and the etiologic underpinnings are unknown but could

include modi

fi

er genes. And of course, disease phenotype may have a bearing on air-

way microbial community composition and activity, and vice versa. So, in conclusion,

one must consider

“

which CF

”

to model: exacerbating versus stable? Early (pediatric)

versus advanced (adult) disease? Mild phenotype versus severe phenotype?

Exploring heterogeneity across time and space.

It is clear that the CF lung is a

heterogeneous environment across many dimensions, complicating our understanding

of the

in vivo

environment. It is important to note that any heterogeneity in the envi-

ronment may impact the microbe(s), the host (including the immune response), or the

host-microbe(s) interactions, adding yet another layer of complexity. Among the issues

one must consider are (i) the right spatial scale to study (bulk, single cell, or some-

where in the middle; micrometers versus nanometers); (ii) the right time scale for

studying communities (minutes, hours, days, or years) and whether these infection

communities are stable or unstable over time; (iii) how microbial growth rates vary

across the airway; (iv) local versus distance effects (cell-cell contact, diffusion of metab-

olites, proteases, and virulence factors); (v) density of microbes across the airway (i.e.,

proximal/distal airway, across mucus plugs); (vi) the de

fi

nition of

“

chronic,

”

whether it

is the same for all pathogens, and what

“

chronic

”

infection is in the context of a model

system (perhaps achieving stability of the microbial community or, in an animal model,

some indicator of host tissue damage or in

fl

ammatory markers or morphological evi-

dence); (vii) how antibiotic tolerance varies across the environment; (viii) how the envi-

ronment varies with host genotype; (ix) whether there is variation in available electron

acceptors (e.g., oxygen, nitrate); (x) how much genetic and/or phenotypic variation

exists within a species in a single lung and how many isolates need to be sequenced/

analyzed to appreciate such heterogeneity; (xi) the rate of

Pseudomonas aeruginosa

evolution (given the evidence that isolates vary over time) and whether there are rules

that govern the evolutionary trajectory of different pathogens; and

fi

nally (xii) how

much of the heterogeneity is

“

real

”

and how much is sampling error.

Documenting the

in vivo

environment.

A critical

fi

rst step in developing a model

system is

fi

lling the key knowledge gap of understanding the nature of the environ-

ment to be modeled, in this case, the CF airway. Documenting this environment

Opinion/Hypothesis

®

September/October 2021 Volume 12 Issue 5 e01763-21

mbio.asm.org

Downloaded from https://journals.asm.org/journal/mbio on 08 November 2021 by 131.215.71.167.

involves the following. (i) First is measuring oxygen, pH, viscosity, nutrients (amino acids,

lipids, mucus, and eDNA) and micronutrients (metals and their various redox states, which

change with changing oxygen availability, for example). Importantly, the spatiotemporal

variability of these parameters needs to be characterized. In the case of nutrients, they

may be host or microbe derived, and when one can measure these compounds, concen-

trations might re

fl

ect bulk measurements or an average across patchy concentrations or

a gradient. (ii) Next is exploring whether the environment is static or characterized by

fl

ow, or if there is mixing. (iii) Third is visualizing the microbial communities and their sur-

roundings using approaches like Microbial Identi

fi

cation After Passive Clarity Technique

(MiPACT), combinatorial labeling and spectral imaging-

fl

uorescence

in situ

hybridization

(CLASI-FISH), or bioorthogonal noncanonical amino acid tagging (BONCAT) (1

–

3) to

appreciate their biogeography. Are the microbes in mixed communities or isolated, and if

they are isolated, is cross-diffusion of metabolites and other secreted molecules possible?

(iv) Last is appreciating that all of the above parameters likely change as the community

evolves, which in turn changes the infection environment, and so on. A clear challenge is

how one would make these measurements, as removing samples from the airway of a

pwCF and analyzing them invariably causes perturbations. One indirect method to under-

stand the

in situ

environment is to analyze the microbes and their

in vivo

state and infer

the environment from the response of the microbes (i.e., use microbes as biosensors).

Alternatively, developing or introducing new biological, chemical, or mechanical sensors

mightbeanecessitytodocumentthe

in vivo

environment accurately. A signi

fi

cant chal-

lenge to this approach is that even if one had all possible information, it is likely that we

would not really understand which components of this information are actually impor-

tant or relevant to CF airway infections. That is, what do we mean when we ask,

“

What

matters?

”

Using modeling or statistical approaches with data combined across many

patients may allow us to differentiate between factors that have a high versus low impact

on infections and infection dynamics.

Contributions of microbes, host, and interactions among these actors.

A model

system needs to capture three key factors driving the local infection environment

—

the contributions of the microbe(s), the host, and the host-microbe interactions. The

host-microbe interactions include core microbial metabolisms and virulence factors, as

well as the host

’

s defensive countermeasures. And while the focus of study is often on

bacteria, fungi and eukaryotic/bacterial viruses are also key players, and as mentioned

above, even within a single species, genetic and/or phenotypic variability can be sub-

stantial. Finally, the nature of the host immune response evoked by these organisms,

which is highly dependent upon the microbes

’

expressed phenotype at various stages

of infection, is also an important variable that must be considered.

Design factors for model communities.

As described above, it is likely that more

than one infection community would be needed to effectively capture the variety of

CF infections. How does one decide on such communities? Approaches to identify

communities could focus on (i) infection severity (mild, medium, or severe); (ii) cluster-

ing common community types based on 16S and/or 18S rRNA amplicon library

sequence or metagenomic data, metabolic potential, or other factors; (iii) leveraging

an ecological framework by considering stable/unstable communities, community suc-

cession, transition from one stable state to the next, or energy conservation as a driver

of overall community organization; (iv) establishing model communities with a domi-

nant pathogen versus an

“

even

”

distribution; or (v) using abundance versus prevalence

data. The challenge here is that some microbes are common across many pwCF

(

Pseudomonas

Staphylococcus

Streptococcus

, and

Prevotella

) while other pathogens

are abundant in a few patients but deadly in this context (i.e.,

Burkholderia

and the

mold

Aspergillus fumigatus

A priori

, it is not obvious which of these community types

should be analyzed to maximize

in vivo

relevance in the context of CF. Moreover, how

contemporary CFTR-targeted therapies will impact microbial communities is an open

question.

Validating the models.

Any model development would require validation, engag-

ing clinicians, microbiologists, immunologists, and researchers from other

fi

elds to

Opinion/Hypothesis

®

September/October 2021 Volume 12 Issue 5 e01763-21

mbio.asm.org

Downloaded from https://journals.asm.org/journal/mbio on 08 November 2021 by 131.215.71.167.

help with this process. There are multiple approaches to validating a model, including

but not limited to (i) using transcriptomics, proteomics, single-cell studies, or related

omics approaches to compare the model to

in vivo

samples; (ii) using phenotypic

assays, including antibiotic tolerance pro

fi

les or breath condensate pro

fi

les, to com-

pare the model to

in vivo

samples; (iii) utilizing validated biomarkers, as they become

available, to compare the model to

in vivo

samples; (iv) developing a quantitative accu-

racy score measure to help compare one model to another and to the

in vivo

setting;

(v) assessing whether the model reproduces across labs; and (vi) engaging a team

strategy where different labs leverage their speci

fi

c expertise to validate relevant

aspects of the model. Whatever metrics are used, validation should engage multiple

lines of evidence and falsi

fi

ability.

FRAMEWORK 2: THE RIGHT MODEL FOR THE QUESTION

Here, the concept is that a particular model might be suited to address a speci

fi

question, but the same model might have less utility in other contexts. The issues out-

lined in framework 1 would need to be considered, but these issues may be much

more easily addressed when only a speci

fi

c aspect of model utility needs to be tackled.

A list of models and their current or potential future utility is presented in Table 1; due

to lack of space, only a few selected models are brie

fl

y discussed here to illustrate the

concept underlying this framework.

Bacterium-host cell coculture models.

Bacterium-host cell coculture models have

been validated (4) and effectively used to study processes including iron cross feeding

from host cell to microbe (5), identi

fi

cation and analysis of virulence factors (6, 7), for-

mation of bio

fi

lms and the associated high-level bio

fi

lm antibiotic tolerance (8, 9), and

bacterial/viral/host interactions (10).

ASM.

The various formulations of arti

fi

cial sputum medium (ASM) provide excellent

(11, 12), and now validated (4), model media for assessing bacterial physiology in a nu-

trient environment re

fl

ective of the CF airway. However, this medium lacks a variety of

host-derived metabolites and likely does not re

fl

ect the variation in pH, viscosity, or

mucin/eDNA concentrations found

in vivo

WinCF.

The WinCF system uses the concept of a Winogradsky column, combined

with ASM, to generate gradients (oxygen, pH, and nutrients), with the potential to add

additional complexity, such as host metabolites (13). Again, while an improvement

over using LB liquid cultures, this system does lack, for example, immune cells and the

ability to assess the infections over long time frames.

FRAMEWORK 3: MODEL SYSTEMS AS HYPOTHESIS GENERATORS

The driving concept here is to develop relatively simple models, based on the best

information currently available, and use these models to explore mechanistic details of

microbe-microbe or microbe-host interactions. Once a speci

fi

c mechanistic interaction

is dissected, one then has a speci

fi

c question to answer via exploiting the more com-

plex model systems or human clinical samples (Fig. 1). This bottom-up approach is

attractive because it does not rely on a complete understanding of the infection envi-

ronment before embarking on a study. The danger is the dif

fi

culty of starting from a

low-level model and predicting what will happen with increasing complexity. The

other obvious risk, of course, is that one might discover a biologically interesting phe-

nomenon with no relevance to infection.

A simple polymicrobial model could start with 4 to 6 microbes; approaches for

selecting microbes may include modeling of 16S and/or 18S rRNA gene amplicon data

sets or mining transcriptomes with tools like ADAGE (analysis using denoising autoen-

coders of gene expression) (14) (additional considerations are presented in framework

1). The underlying premise is that the infection environment selects for stable com-

munities (pwCF show stable airway communities for often extended periods [15

–

17]),

and analyses of clinical data reveal the composition and features of such communities.

It might also be possible to evolve stable communities

in vitro

. Alternatively, one could

slowly build complexity

—

start with early colonizers of the CF airway like

Staphylococcus

Opinion/Hypothesis

®

September/October 2021 Volume 12 Issue 5 e01763-21

mbio.asm.org

Downloaded from https://journals.asm.org/journal/mbio on 08 November 2021 by 131.215.71.167.

TABLE 1

Utility of model systems for airway infection in CF

Model type

System

Use

Studies of

airway

infection

Chronic

airway

infection

Gut-lung

axis

studies

General

infection

studies

Studies of basic

microbial

biology

Studies

of host

response

New

antibiotic

discovery

New

antibiotic

validation

CF animal models

Mouse

YYYY

Rat

Piglet

Ferret

Rabbit

Other animal models

Zebra

fi

Caenorhabditis elegans

Wax moth larvae

Ex vivo

models

Porcine lung model

to 21 days

Piglet trachea

Human lung tissue

Human cell lines

Human primary cells

Organoid-derived 2D cultures

Human lung on chips

Sputum

In vitro

models

LB (standard lab medium)

Synthetic CF sputum medium

Conditioned medium from host cells

Conditioned bacterial supernatant

Micro

fl

uidic chambers

WinCF

Chemostats

Multiwell plates

In silico

models

Bioinformatics and modelling of existing data

YYYYY

Empirical machine learning models

Microbial metabolic models

Airway transport models

Agent-based models

Immune system models

Immune cells (neutrophils, monocytes, etc.)

Primary mouse immune cells

Primary human immune cells

YYYYY

Antibodies

Y, yes; F, future development of this use is likely.

Work that is not airway speci

fi

c but can help understand general virulence factors or their mechanism of action.

Agar bead model.

Classifying

ex vivo

versus

in vitro

models can sometimes be dif

fi

cult. Here, we classify models as

ex vivo

if they comprise complex tissues or human samples and use primary cells.

If adapted to use (arti

fi

cial) sputum.

Other

in vitro

bio

fi

lm models are also available (e.g., Calgary device, beads, tube bio

fi

lms, etc.).

Microbiome, metagenome, transcriptome, proteomes, chemical dynamics,

fl

uid

fl

ow, etc., alone or in combination with clinical metadata.

See also animal and

ex vivo

models.

Including peripheral and alveolar cells.

Opinion/Hypothesis

®

September/October 2021 Volume 12 Issue 5 e01763-21

mbio.asm.org

Downloaded from https://journals.asm.org/journal/mbio on 08 November 2021 by 131.215.71.167.

aureus

and

Haemophilus in

fl

uenzae

, then add a second microbe, a third, and a fourth,

and then alter the environment (add mucus, antibiotics, or immune cells), followed by

using phenotypic or omics analyses to ask when (if) the community reaches an equilib-

rium. With any strategy, using reference strains, with all the tools available to study such

strains, is an excellent start, which would then need to be supplemented with testing of

multiple clinical isolates of each genus, multiple species, and the variety of genotypes

that evolve

in vivo

(i.e.,

lasR

mutants, nonmotile strains, and antibiotic resistance). Using

the well-characterized ASM would be a reasonable place to start for a growth medium. A

clear advantage of such an approach is the ability to replicate

fi

ndings across labs and

perhaps even foster the adoption of such systems (with standard strains and media)

across multiple labs, generating a shared model system attacked by multiple researchers

using their best tools. Adopting common model systems for polymicrobial communities

could be akin to labs focusing on

Escherichia coli

Bacillus subtilis

Candida albicans

,and

P. aeruginosa

as single-organism models, which helped coalesce efforts and drive pro-

gress in understanding basic microbial biology.

Test hypotheses in more complex systems, including with animal models or

human clinical samples.

The idea here is to use the simple models to generate

hypotheses to test in more complex systems. Hypotheses generated could be quite

focused and assist in the

a priori

selection of the

“

right model for the job.

”

Framework

3 should be considered an iterative series of hypothesis-generating experiments, fol-

lowed by testing the hypotheses in more complex systems (i.e., animal models and

clinical samples), followed by improvement of the simple model and exploration to

generate the next round of hypotheses to be tested.

IT IS WELL WORTH THE EFFORT

Developing new and better model systems for CF is a clear and important chal-

lenge, but the bene

fi

ts could be large for pwCF in terms of understanding infection

biology and host responses to these infections, as well as developing discovery plat-

forms for new, more ef

fi

cacious antimicrobial agents. Despite the complexity of reca-

pitulating CF infections

in vitro

ex vivo

, or in animals, a concerted effort toward improv-

ing model systems raises hope for the emergence of some general organizing

principles applicable to chronic infections in CF. Finally, the general frameworks out-

lined here may provide a strategy for tackling other polymicrobial infections (e.g.,

chronic wounds and non-CF bronchiectasis), in terms of both studying community

structure/function and developing more ef

fi

cacious therapies.

FIG 1

Hypothesis generator cycle. The cycle starts with existing clinical data sets (i.e., 16S rRNA gene

amplicon sequences or metagenomes) combined with statistical analyses or predictive modeling to

inform the development of relatively simple

in vitro

ex vivo

models. Such models are used to learn

aspects of molecular mechanisms or ecological principles driving microbe-microbe or host-microbe

interactions; the data are then used to drive the next round of hypotheses that can be tested in

more complex models or human samples. This process can be repeated multiple times, with each

turn of the cycle providing additional insight into the complex polymicrobial infections in CF.

Opinion/Hypothesis

®

September/October 2021 Volume 12 Issue 5 e01763-21

mbio.asm.org

Downloaded from https://journals.asm.org/journal/mbio on 08 November 2021 by 131.215.71.167.

ACKNOWLEDGMENTS

We thank the Cystic Fibrosis Foundation

’

s willingness to support the original, in-

person version of this workshop, which needed to be canceled and rescheduled as a

virtual event due to the COVID-19 pandemic. We thank Madison Leonard and Cindy

Stewart of DartCF for their administrative support, Fabrice Jean-Pierre for his assistance,

and the workshop session moderators (A. Crabbé, J. Bomberger, V. Phelan, J. Davies,

D. Limoli, R. Kümmerli) and scribes (T. Hampton, R. Hunter, L. Dietrich, A. Price-Whelan,

C. Khursigara, J. Bliska, W. DePas) for their willingness to help both the planning of and

logistical support during the workshop.

DartCF is supported by the NIH (P30-DK117469) and the Cystic Fibrosis

Foundation Research Development Program (STANTO19R0). Additional support was

provided by NIH/R01AI155424, NIH/1R01HL152190, CFF/CRAMER19GO, CIHR/PJT-

156111, NIH/R35GM

128690, NIH/R01DK104847, R01/NIHAI127548, CFF/TURNER19PO,

NIH/R01HL136919, CFF/ASHARE21G0, NIH/R35HL18500, NIH/P20GM130454, NIHR

Systematic Reviews Program, CFF/WHITEL20A0, NIH/R01AI103369, University of Nottingham

Anne McLaren Fellowship, Swiss National Science Foundation/31003A_182499, MRC/MR/

R001898/1, NIH/R01HL151385, NIH/R01AI155424, CFF/BOMBER21R3, CFF/SPERO19F0, NIH/

R01HL136647, Paul G. Allen Frontiers Investigator Award, NIH/R01HL136647, Cystic Fibrosis

Trust Strategic Research Centre Award, CF Trust/VIA078, CF Trust/VIAO77, CIHR/FDN/

159902, and CFF/BOMBER18G0.

REFERENCES

1. DePas WH, Starwalt-Lee R, Van Sambeek L, Ravindra Kumar S, Gradinaru

V, Newman DK. 2016. Exposing the three-dimensional biogeography and

metabolic states of pathogens in cystic

fi

brosis sputum via hydrogel

embedding, clearing, and rRNA labeling. mBio 7:e00796-16.

https://doi

.org/10.1128/mBio.00796-16

2. Valm AM, Mark Welch JL, Rieken CW, Hasegawa Y, Sogin ML, Oldenbourg

R, Dewhirst FE, Borisy GG. 2011. Systems-level analysis of microbial com-

munity organization through combinatorial labeling and spectral imag-

ing. Proc Natl Acad Sci U S A 108:4152

–

4157.

https://doi.org/10.1073/

pnas.1101134108

3. Valentini TD, Lucas SK, Binder KA, Cameron LC, Motl JA, Dunitz JM, Hunter

RC. 2020. Bioorthogonal non-canonical amino acid tagging reveals trans-

lationally active subpopulations of the cystic

fi

brosis lung microbiota. Nat

Commun 11:2287.

https://doi.org/10.1038/s41467-020-16163-2

4. Cornforth DM, Diggle FL, Melvin JA, Bomberger JM, Whiteley M. 2020.

Quantitative framework for model evaluation in microbiology research

using

Pseudomonas aeruginosa

and cystic

fi

brosis infection as a test case.

mBio 11:e03042-19.

https://doi.org/10.1128/mBio.03042-19

5. Moreau-Marquis S, Bomberger JM, Anderson GG, Swiatecka-Urban A, Ye

S, O'Toole GA, Stanton BA. 2008. The DeltaF508-CFTR mutation results in

increased bio

fi

lm formation by

Pseudomonas aeruginosa

by increasing

iron availability. Am J Physiol Lung Cell Mol Physiol 295:L25

–

L37.

https://

doi.org/10.1152/ajplung.00391.2007

6. MacEachran DP, Ye S, Bomberger JM, Hogan DA, Swiatecka-Urban A,

Stanton BA, O'Toole GA. 2007. The

Pseudomonas aeruginosa

secreted pro-

tein PA2934 decreases apical membrane expression of the cystic

fi

brosis

transmembrane conductance regulator. Infect Immun 75:3902

–

3912.

https://doi.org/10.1128/IAI.00338-07

7. Kiedrowski MR, Paharik AE, Ackermann LW, Shelton AU, Singh SB, Starner

TD, Horswill AR. 2016. Development of an in vitro colonization model to

investigate

Staphylococcus aureus

interactions with airway epithelia. Cell

Microbiol 18:720

–

732.

https://doi.org/10.1111/cmi.12543

8. Anderson GG, Moreau-Marquis S, Stanton BA, O'Toole GA. 2008. In vitro

analysis of tobramycin-treated

Pseudomonas aeruginosa

bio

fi

lms on

cystic

fi

brosis-derived airway epithelial cells. Infect Immun 76:1423

–

1433.

https://doi.org/10.1128/IAI.01373-07

9. Starner TD, Zhang N, Kim G, Apicella MA, McCray PB, Jr. 2006.

Haemophi-

lus in

fl

uenzae

forms bio

fi

lms on airway epithelia: implications in cystic

fi

brosis. Am J Respir Crit Care Med 174:213

–

220.

https://doi.org/10.1164/

rccm.200509-1459OC

10. Hendricks MR, Lashua LP, Fischer DK, Flitter BA, Eichinger KM, Durbin JE,

Sarkar SN, Coyne CB, Empey KM, Bomberger JM. 2016. Respiratory syncy-

tial virus infection enhances

Pseudomonas aeruginosa

bio

fi

lm growth

through dysregulation of nutritional immunity. Proc Natl Acad Sci U S A

113:1642

–

1647.

https://doi.org/10.1073/pnas.1516979113

11. Palmer KL, Brown SA, Whiteley M. 2007. Membrane-bound nitrate reduc-

tase is required for anaerobic growth in cystic

fi

brosis sputum. J Bacteriol

189:4449

–

4455.

https://doi.org/10.1128/JB.00162-07

12. Palmer KL, Mashburn LM, Singh PK, Whiteley M. 2005. Cystic

fi

brosis spu-

tum supports growth and cues key aspects of

Pseudomonas aeruginosa

physiology. J Bacteriol 187:5267

–

5277.

https://doi.org/10.1128/JB.187.15

.5267-5277.2005

13. Comstock WJ, Huh E, Weekes R, Watson C, Xu T, Dorrestein PC, Quinn RA.

2017. The WinCF model

—

an inexpensive and tractable microcosm of a

mucus plugged bronchiole to study the microbiology of lung infections. J

Vis Exp 8.

https://doi.org/10.3791/55532

14. Tan J, Hammond JH, Hogan DA, Greene CS. 2016. ADAGE-based integra-

tion of publicly available

Pseudomonas aeruginosa

gene expression data

with denoising autoencoders illuminates microbe-host interactions.

mSystems 1:e00025-15.

https://doi.org/10.1128/mSystems.00025-15

15. Gifford AH, Alexandru DM, Li Z, Dorman DB, Moulton LA, Price KE,

Hampton TH, Sogin ML, Zuckerman JB, Parker HW, Stanton BA, O'Toole

GA. 2014. Iron supplementation does not worsen respiratory health or al-

ter the sputum microbiome in cystic

fi

brosis. J Cyst Fibros 13:311

–

318.

https://doi.org/10.1016/j.jcf.2013.11.004

16. Carmody LA, Zhao J, Kalikin LM, LeBar W, Simon RH, Venkataraman A,

Schmidt TM, Abdo Z, Schloss PD, LiPuma JJ. 2015. The daily dynamics of

cystic

fi

brosis airway microbiota during clinical stability and at exacerba-

tion. Microbiome 3:12.

https://doi.org/10.1186/s40168-015-0074-9

17. Zhao J, Schloss PD, Kalikin LM, Carmody LA, Foster BK, Petrosino JF,

Cavalcoli JD, VanDevanter DR, Murray S, Li JZ, Young VB, LiPuma JJ. 2012.

Decade-long bacterial community dynamics in cystic

fi

brosis airways.

Proc Natl Acad Sci U S A 109:5809

–

5814.

https://doi.org/10.1073/pnas

.1120577109

Opinion/Hypothesis

®

September/October 2021 Volume 12 Issue 5 e01763-21

mbio.asm.org

Downloaded from https://journals.asm.org/journal/mbio on 08 November 2021 by 131.215.71.167.