micr microbiolog obiologie ie Lulla Opatowski - - PowerPoint PPT Presentation
Modlisa Modlisation en tion en micr microbiolog obiologie ie Lulla Opatowski Pharmaco-Epidemiology and Infectious Diseases unit UMR1181 Institut Pasteur / Inserm / Universit de Versailles St Quentin RICAI Dcembre 2019 Emergence,
Pharmaco-Epidemiology and Infectious Diseases unit UMR1181 Institut Pasteur / Inserm / Université de Versailles St Quentin
RICAI – Décembre 2019
2
Ecology of pathogens: Colonies Interactions, competition and synergy Genetic evolution : Mutations, genes transfer Microorganism Colonization of ecosystems (gut, skin, naso-, oro-pharynx) Infection Immunity Drug exposure (vaccine, antibiotics) Individuals Between-host transmission (heterogeneous) Public health policies Population
– High density of individuals and contacts – High antibiotic exposure => high frequency of antibiotic resistant bacteria (ARB) – At-risk patients, eg. intensive care or post surgery – Specificities
– Efficient control may require the set up of heavy procedure
3
Controlling this dissemination in these settings requires understanding bacterial spread
and in particular ARB
4
Routes of acquisition? Microbiota, within-host pathogens interactions? Impact of antibiotic exposure? Pathogen’s intrinsic epidemicity? Factors of clearance?
manner
=> Set up ad-hoc epidemiological studies => development of biostatistics, mathematical modelling tools
Natural clearance Susceptible Colonized with a S bacteria Colonized with a R bacteria Emergence (p) Contacts 0 1 1 1
ds dt = 0 − s(cS + cR) + (cS + cR) − 1s dcS dt = +scS − cS − pcS − 1cS dcR dt = +scR − pcR − cR − 1cR s + cS + cR =1
1 homogeneous population 1 drug, 1 simple mechanism => Useful but poorly representative => more specificity is required
5
Force of infection
𝜇 𝑢 = 𝛾(𝐷𝑇 + 𝐷𝑆)
data from hospital database records, and inter-individual proximities from log sensors
6
[Assab, Nekkab et al, Current Op Inf Dis 2017]
Ward 1 Ward 2 Ward 5 Ward 8 Ward 7
7
Detection of all Close proximity interactions (CPI) every 30sec with a distance less than 1m50
Berck sur Mer Long term care hospital 329 patients 263 hospital staff 5 months 2 740 728 contacts
I-BIRD study : « Individual-based Investigation of Resistance Dissemination » → Prospective longitudinal study →Principal investigator Didier Guillemot →European project MOSAR →June-October 2009
Daily Contact data Staphylococcus aureus Enterobacteriaceae
rectal Weekly swab
N~6620 N~3500
Antibiotics
9
Frequency of contacts over a day Duration of contacts
[Duval A, et al. Scientific report 2018]
10
Frequency of contacts over a day Duration of contacts
[Duval A, et al. Scientific report 2018]
11
12
New acquisition Potential transmitter
Real data
Newly colonized Candidate Uncolonized
New acquisition Randomly pick up transmitter on contact network
Randomly permuted data (null hyp) Comparaison (Wilcoxon test)
[Obadia et al, Plos Comp Biol 2015] [Duval et al, Plos Comp Biol 2019]
13
ESBL Escherichia coli
p-value = 0.25 35 acquisitions
ESBL K. pneumoniae
p-value = 0.025 20 acquisitions
p-value < 0.001
[Obadia et al, Plos Comp Biol 2015] [Duval et al, Plos Comp Biol 2019]
S. aureus and K. pneumoniae transmission supported by the proximity network Not clear for E. coli, probably other important sources
Shorter distance found in observed data compare to permutated data
Distance to the nearest potential transmitter
14
Likelihood-based parameter estimation MCMC Contact data, swabs, antibiotics … Transmission rate per contact.day for the different strains
15
P0 P0
P0 P0 tn : last positive swab (strain s) t0 : first positive swab (strain s)
Negative swab Positive swab Recorded proximity
For any non-colonized individual i, we define the force of acquisition on day t :
𝜇𝑗 𝑢, 𝑡 =
𝑘≠𝑗, 𝑡𝑘 𝑢,𝑡 =1 𝑗∈𝐷𝑘(𝑢,𝑡)
𝛾(𝑦
𝑘, 𝑦𝑗, 𝑢)
– xi, xj, denote for individuals i and j, the vectors of individual variables eg. status [HCW versus patient] – 𝛾(𝑦𝑘, 𝑦𝑗, 𝑢), the probability of transmission of strain s from individual j to individual i on a given day of contact t – Case transmission network of j
𝐷
𝑘 𝑢, 𝑡 = ቊ 𝑗, 𝑗, 𝑘 ∈ 𝐹 𝑢 , 𝑡𝑗 𝑢, 𝑡 = 0 if 𝑢 ∈ [𝑢1, 𝑢𝑜]
∅ otherwise
16
17
=> significantly higher epidemicity for MRSA t777 => More generally, higher epidemicity for MRSA
MRSA MSSA
[Opatowski et al, in prep]
→Some species spread over interindividual networks but lack of evidence for
→In MRSA: different epidemicities according to spatype and resistance to methicillin
– differences in modes of transmission? – differences de transmissibility between MRSA and MSSA for t777 ?
18
20
Dynamics of relative abundance of different species/groups in the microbial community
N: the number of different species 𝑦𝑗(𝑢): the abundance of species i at time t 𝑠𝑗: growth rate of species i (can be negative when receiving antibiotic or under immune response) 𝑏𝑘𝑗: interaction strength between species j and i (positive or negative) K: carrying capacity of the entire community
[Xuefeng Gao et al, Frontiers in Microbiology 2018]
Ι
When longitudinal microbiota data is available …
21
[Xuefeng Gao et al, Frontiers in Microbiology 2018]
When longitudinal microbiota data is available …
22
Dynamics of relative abundance of different groups in the microbial community
[Xuefeng Gao et al, Frontiers in Microbiology 2018]
Ι
Antibiotics Symptoms Feeding Colonization with ESBL
When longitudinal microbiota data is available…
Natural clearance Susceptible Colonized with a S bacteria Colonized with a R bacteria Emergence (p) Contacts 0 1 1 1
ds dt = 0 − s(cS + cR) + (cS + cR) − 1s dcS dt = +scS − cS − pcS − 1cS dcR dt = +scR − pcR − cR − 1cR s + cS + cR =1
1 homogeneous population 1 species 1 drug
23
Force of infection
𝜇 𝑢 = 𝛾(𝐷𝑇 + 𝐷𝑆)
FOR MICROBIOME- PATHOGEN CO- COLONIZATION
model
healthcare setting
patients receiving antibiotics
(qr)
antibiotics against which pathogen resistant Hospital
High-D flora (H) Low-D flora (L)
Pathogen (P) No pathogen
1 1 2 2 3 4 3 4
1. Flora disruption (𝜏) 2. Flora recovery (𝜀) 3. Pathogen acquisition (𝛾+𝛽) 4. Pathogen clearance (𝛿+𝜏)
[Smith et al, in prep]
Reduced pathogen force of infection (𝜇)
bacteriocins inhibit C. difficile and E. coli colonization
Affected pathogen clearance rate (𝛿)
duration in E. faecalis, Listeria spp.; Bacterioides metabolites enhance colonization duration in E. coli
Kamada et al. Nat Rev Immunol 2013 Pacheco et al. Nature 2012
Enhanced endogenous acquisition rate (𝛽)
acids prevent
subdominant C. difficile colonies
Buffie & Pamer Nat Rev Immunol 2014 Buffie et al. Nature 2014
5.Horizontal gene transfer (HGT)
HGT especially important among Gram-negative pathogens (ESBL- Enterobacteria)
Rarely considered in models
K. pneumoniae
Community prevalence (fP) Medium Low High Medium Transmission rate (𝛾) High Medium Low High Endogenous acquisition rate (𝛽) Low Low High Medium Clearance rate (𝛿) Medium Medium Low Medium Barrier protection (𝜁) Low High Medium Medium Niche competition (𝜃) Medium None High None Ecological release (𝜚) None High Medium Low HGT rate (𝜓) None None Medium High Resistance proportion (qr) Medium Low Medium High
Narrow spectrum Broad spectrum
Proportion of all antibiotics used (qn) High Low Pathogen clearance rate (𝜏P) Medium Medium Flora disruption rate (𝜏F) Low High
INTERVENTION 1: REDUCE PRESCRIPTIONS BY 30%
−20 −10 10
Change in colonization incidence (%)
Gram-negatives
INTERVENTION 2: RESTRICT BROAD-SPECTRUM ANTIBIOTICS BY 30%
−20 −10 10
Change in colonization incidence (%)
ecological release (C. difficile) and/or HGT (Gram- negatives)
(ineffective) narrow spectrum antibiotics leads to more selection for resistant pathogen
Preliminary, requires expert elicitation…
[Smith et al, in prep]
→Model analysis can help understand better the impact of complex mechanisms at stake and assess the global effect of these mechanisms at the population level →Pathogen’s interaction with flora may affect transmission routes and impact of antibiotics on resistance selection at the population level →Antibiotic stewardship interventions may be able to more effectively target specific pathogens when taking flora into account
28
models, it is possible to provide novel understanding of the mechanisms at stake
– Analyse data by considering dynamic and mechanistic aspects, estimate parameters – We showed E. coli does not spread over contacts network and that different S. aureus does, but clones have different epidemicities
processes and optimize control measures => direct applications in terms of public health
– biological origin ? => look for associations with genes / biological functions
29
microbiologists
30
Question biologique Design experimentation épidémiologique / microbiologique Analyse par des modèles dynamiques Interprétation biologique
31
Institut Pasteur)
With increasing pathogen clearance:
pathogens resistant With increasing flora disruption:
(hosts more susceptible)
Intermediate antibiotic resistance (qr) Intermediate interaction strength (𝜁 + 𝜃 + 𝜚 + 𝜓)
0.2 0.4 0.6 0.8 2 4
0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.0 0.1 0.2 0.0 0.1 0.2
Proportion of patients on antibiotics (qa) Daily colonization incidence of resistant pathogen (PR)
Acquisition route
Endogenous HGT Transmission
Overall similar incidence despite very different interaction strengths But large shift in acquisition routes High interaction strength may help explain patterns observed in ESBL E. coli:
acquisition despite low evidence of cross- transmission
interventions such as hygiene
Low-strength interactions High-strength interactions Pathogen resistant to greater share of antibiotics (qr)
−20 −10 10
Change in colonization incidence (%)
Endogenous Transmission HGT Total
−30 −20 −10 10 −30 −20 −10 10 −30 −20 −10 10 −30 −20 −10 10
Change in colonization incidence (%)
Gram-negatives
−20 −10 10
Change in colonization incidence (%)
ecological release (C. difficile) and/or HGT (Gram-negatives)
(ineffective) narrow spectrum antibiotics leads to more selection for resistant pathogen
Endogenous Transmission HGT Total
−30 −20 −10 10 −30 −20 −10 10 −30 −20 −10 10 −30 −20 −10 10
Change in colonization incidence (%)
36
wikipedia
mechanistic process
Need to take into account unobserved phenomenon => Consider the problem as whole
populations
=> in silico experiments
Antibiotics affect distinct bacteria at different rates – Rate of antibiotic-induced pathogen clearance (𝜏P) – Rate of antibiotic-induced flora disruption (𝜏F) Example: – Narrow-spectrum antibiotics
– Broad-spectrum antibiotics
1. Analyze data taking into account the mechanistic and dynamic aspects - using statistical inference – Assess transmission routes, at risk behaviors, natural history of disease, pathogens interactions 2. Analyze model behaviour – Assess role of parameters or mechanisms
– Eg. Antibiotic stewardship or cohorting in hospitals – Support for decision making in Public health
38