In-vitro evolution of colistin resistance

Richard Neher
Biozentrum, University of Basel

slides at

How we got into this...

Bianca Regenbogen


  • polymyxin, active against gram negatives
  • interacts with outer membrane
  • old antibiotic, discontinued because of nephrotoxicity
  • today: last-resort antibiotic

mcr-1: "mobilized colistin resistance"

  • first found in China 2 years ago
  • now found in >30 countries
  • mainly E. coli, Klebsellia, and Enterobacter

How does colistin resistance emerge in situ?

  • How fast?
  • Which mutations?
  • Which order?
  • Genetic background?
  • Two P. aeruginosa strains from blood stream infections
  • Extensively drug resistant phenotypes
  • PA77:
    • Plasmid 40kb, chromosome 3.7Mb, 2.3Mb, 1Mb contigs
    • blaIMP-8, blaOXA-10, blaOXA-50
    • various aminoglycoside resistance determinants
    • fosfomycin resistance
    • mostly on plasmid
  • PA83:
    • Plasmid 400kb, chromosome 6.8Mb
    • blaVIM-2, blaOXA-50
    • various aminoglycoside resistance determinants
    • fosfomycin resistance
    • in several integrons in the chromsome

From human patients to in-vitro experiments

Morbidostat by Toprak et al.

  • Computer measures OD
  • Controls pumps to add medium or AB
  • Waste is removed
  • Morbidostat→ growth rate is kept constant
  • Chemostat → dilution is constant
  • Turbidostat → OD is constant

Our morbidostat

  • more flexible software
  • more compact design
  • cheaper pumps and controllers

Running the morbidostat

  • Typically 10 cultures in parallel
  • Measure OD every 30s
  • Adjust AB conc every 10min
  • Samples for sequencing every 2-3 days
  • Total duration: 3weeks
  • ~50'000 OD measurements per culture
  • Most parts can be autoclaved
  • Pumps sterilized with bleach/ethanol washes
  • Replace vials every few days to avoid biofilms
  • New piezo-pumps can be autoclaved as well

Colistin resistance emerges within 2 weeks

Mutation trajectories in strain PA77

  • Whole genome deep sequencing ($>200x$) with Illumina.
  • Track rare mutations, no matter where they are
  • Mutation frequencies to about 5% accuracy

Mutation trajectories in strain PA83

Recurrent mutations PA77

Genelocus tag PAO1 v01 v02 v03 v04 v05 v05a v08a v10a v11a
pmrB PA4777 V9A,L17Q L90Q,E320K V9A P216Q P254L P169X,M292I S257N N41I,P169X H261Y
pmrE PA2043 Y28N Y28C Y28N Y28C Y28N Y28C Y28C Y28N Y28N
lptDPA3559 Y803XL538R
  • pmrE: Most PA strains are 28C → reversion
  • pmrB: Many mutations that constitutively activate the gene
    → canonical colistin resistance gene
  • lptD: code for outer membrane protein, LPS transport.
    → has been associated with colistin resistance in Acinetobacter

Recurrent mutations in PA83

locus tag PAO1 v02 v03 v05 v06 v08 v11 v12 v14 v15
lpxC PA4406 P101S V222A,S106G V222A V164G,A107T A107T,G21W,F176S A107T,I131F M103I D232E,D232G,V217F,V217A V222A,S106G
pmrB PA4777 L96R L171P L87P F51L S8P,E320K V9A G123S E320K,A248T,L167P R259H,V361M
putative tranferase PA3853 C226G Y3C,G62S V34A,Y155C C226G R60C,Y216C,E185G C226G V122A,E185G
asparagine synthetase L365P frameshift L425P G32S frameshift W153* L365P,W153*,V286M
migA PA0705 H219P C25R,N27S D106G Q191R,V22A T196P,H123P H219P A168T
mutS PA3620 T51P T51P T51P T51P T51P T51P,T287P
lpxO2 PA0936 D163A D163N W209* D163A frameshift in-frame deletion
pmrA PA4776 L11Q L11P R159L,G15V,N172D
cupB5 PA4082 G260X,R26C P139P
pdtA PA0690 A3885V,A3885A G1527X
morA PA4601 R1199H G143D
lpxA PA3644 R96S R191C
priA PA5050 L38L R689R
traN W773* G912D
wbpM PA3141 E273K E273G
mscL PA4614 V86I S35P
  • lpxC: lipid A biosynthesis
  • pmrB: Many mutations that constitutively activate the gene
    → canonical colistin resistance gene
  • lptD: code for outer membrane protein, LPS transport.
    → has been associated with colistin resistance in Acinetobacter

Mutations in pmrAB

Previously observed and new mutations in pmrAB (blue: PA83, red: PA77)
Olaitan et al. Front. Micro., 2014

Mutator strains

  • 1/9 PA77 cultures developed mutator phenotypes
  • 6/9 PA83 cultures developed mutator phenotypes
  • about 10-100 fold more mutations, mostly unique
  • non-mutator: almost only non-synonymous mutations
  • mutator: pretty random mix
  • mutator/non-Mutator develop resistance at similar speed/similar mutations

Insertions and deletions

Stable genome!
  • few large deletions
  • a small number if gene-inactivating deletions
  • one notable progressive loss of parts of the resistance plasmid in one PA77 strain

Comparison of the two strains

  • few mutations
  • pmrB: universal, different positions
  • pmrE: reversion at specific position
  • one mutator
  • resistance increased around at day 8-10
  • messy -- many mutations
  • lpxC, pmrB: universal, different positions
  • migA, pmrA, lpxO2: common, different positions
  • 6 out of 9 mutators, possibly prexisting
  • resistance increased around at day 5-6
Take home:
  • pmrB is a common core of colistin resistance
    → 2-component system that regulates LPS modification
  • otherwise strain specific mutation trajectories, mostly involving LPS mods
  • Very reproducible in the same strain
  • Phenotypic trajectories are similar
  • No strict order, but pmrB tends to mutate first

High quality reference genomes and high coverage are essential

  • without good reference, lots of artifactual polymorphisms
  • duplicated regions are collapsed in mapping/assembly → spurious diversity
  • obtaining good references is straightforward with long read sequencing (PacBio/nanopore)
  • Illumina coverage is a poor guide of duplications. Only relative differences in coverage are meaningful.
  • mutations that change in frequency are most likely real than SNPs that remain at similar frequency (need high coverage)

Potential of morbidostat-like devices

Quantities that can be monitored:
  • features that can be captured by a camera or photodetector
  • optical density
  • fluoresence
  • ...
  • additions of several different solutions
  • each vial can be individually addressed
  • rapid feedback on measured parameters
  • custom schedules (e.g.~pharmacokinetics)
  • vary temperature, light, etc
  • electronic and mechanical workshop
  • a bit of programming
  • with flexibility comes the need to specify
  • lots of media
  • avoiding contamination
  • ...


  • Bianca Regenbogen, now Uni Hohenheim
  • Silke Peter, UKT Tübingen
  • Matthias Willmann, UKT Tübingen