bacterial persisters in long-term infection: emergence and ... · 5 jennifer a. bartell 1*, david...
Post on 21-Sep-2020
3 Views
Preview:
TRANSCRIPT
1
Bacterial persisters in long-term infection: emergence and fitness in a complex host 1
environment 2
3
Authors 4
Jennifer A. Bartell1*, David R. Cameron2,3*, Biljana Mojsoska4,6*, Janus Anders Juul 5
Haagensen1, Lea M. Sommer4, Kim Lewis2§, Søren Molin1, Helle Krogh Johansen4,5§ 6
7
Affiliations: 8
1 The Novo Nordisk Foundation Center for Biosustainability, Technical University of 9
Denmark, Lyngby, Denmark 10
2 Antimicrobial Discovery Center, Northeastern University, Boston, USA 11
3 Department of Intensive Care Medicine, Inselspital, Bern University Hospital, University of 12
Bern, Switzerland 13
4 Department of Clinical Microbiology, Rigshospitalet, Copenhagen Ø, Denmark 14
5 Department of Clinical Medicine, University of Copenhagen, Copenhagen N, Denmark 15
6 Present address: Department of Science and Environment, Roskilde University, Roskilde, 16
Denmark 17
18
*Contributed equally to the work and listed alphabetically 19
§Corresponding authors 20
21
BIOLOGICAL SCIENCES: Microbiology 22
Keywords: persister cells, persistent infection, Pseudomonas aeruginosa, cystic fibrosis 23
24
25
26
27
28
29
30
31
32
33
34
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
2
Abstract 35
Despite intensive antibiotic treatment, Pseudomonas aeruginosa often persists in the airways 36
of cystic fibrosis (CF) patients for decades, and can do so without antibiotic resistance 37
development. Using high-throughput screening assays of bacterial survival after treatment with 38
high concentrations of ciprofloxacin and tobramycin, we have determined the prevalence of 39
persisters in a large patient cohort consisting of 460 longitudinal isolates of P. aeruginosa from 40
39 CF patients. Thirty patients exhibited high persister variants (Hip, defined by survival of at 41
least 75% of replicates) in at least one of the two antibiotic screens (25% of isolates in total). 42
Few bacterial lineages were dominated by Hip, but Hip emergence increased over lineage 43
colonization time. Furthermore, transient lineages were significantly less likely to exhibit Hips 44
than non-transient lineages, suggesting that the Hip phenotype is decisive for long-term 45
establishment of a lineage. While we observed no strong signal of adaptive genetic 46
convergence across all lineages with Hip emergence, Hip+ lineages were significantly 47
correlated with lineages with slow growing isolates. Finally, we evaluated Hips in a model CF 48
structured environment by testing the fitness properties of otherwise genotypically and 49
phenotypically similar low-persister (Lop) and Hip isolates in co-culture using a flow-cell 50
biofilm system with antibiotic dosing modelled on in vivo dynamics. Hip survived 51
ciprofloxacin treatment better than Lop. Our results strongly argue against the existence of a 52
single dominant molecular mechanism underlying bacterial antibiotic persistence. We instead 53
show that many routes, both phenotypic and genetic, are available for persister formation and 54
consequent increases in strain fitness in CF airways. 55
56
57
58
59
60
61
62
63
64
65
66
67
68
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
3
Introduction 69
Antibiotic-tolerant persister cells are suspected to be a significant clinical problem that has 70
been seriously neglected in favor of combating antibiotic-resistant bacteria, though persisters 71
were in fact described shortly after the clinical introduction of antibiotics (1). Persisters are 72
distinct from antibiotic-resistant mutants, as they do not grow in the presence of antibiotics. 73
Instead, they remain dormant during antibiotic exposure but retain the capacity to resuscitate 74
and restore the population when antibiotic concentrations drop (2–4). However, our 75
understanding of the physiology and clinical relevance of persister cells is limited, given the 76
difficulty in reliably isolating what is theorized to be a stochastic phenotype in vitro, much less 77
monitoring this phenotype in routine clinical care. Thus, while a few characterizations of small 78
environmental isolate collections have shown that formation of persisters varies across strains 79
(5, 6), few studies have assayed persister formation in clinical or other complex environmental 80
scenarios. One study of oral carriage (0-19 weeks) of Candida albicans isolates from 22 cancer 81
patients undergoing chemotherapy found that patients with carriage of greater than 8 weeks 82
had significantly higher persister levels than those with less than 8 weeks of carriage, but did 83
not address the underlying mechanisms of persistence in this pathogen (7). To examine the 84
underpinnings and long-term impact of the persister phenotype in a clinical scenario, both a 85
large, aligned patient cohort that places the bacteria under similar environmental stresses as 86
well as isolate sampling at a resolution that captures the emergence and longevity of the 87
phenotype are needed. 88
89
P. aeruginosa is the most frequent cause of chronic airway infections in patients with CF (8, 90
9). Mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene often 91
result in inefficient mucociliary clearance of bacteria from the airways, creating opportunities 92
for bacterial colonization (10, 11). Upon entering the host, environmental P. aeruginosa adapts 93
to the CF lung environment, ultimately establishing an incurable airway infection (12, 13). 94
Despite intensive antibiotic treatment from the first discovery of the bacterium in the lung, 95
resistance emergence in the first years of infection is surprisingly low (14, 15). In the absence 96
of clinically defined antibiotic resistance, survival of the bacteria is likely enabled by diverse 97
and co-occurring traits including slowed growth rate, biofilm formation, and the production of 98
small fractions of antibiotic tolerant subpopulations (16–18). How persister cells interrelate 99
with these complex and co-selected changes in vivo is rarely accounted for in in vitro persister 100
studies, but is likely clinically important. 101
102
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
4
These factors also complicate the search for genetic mechanisms of the persister phenotype 103
that are clinically impactful. While persister cells are stochastic phenotypic variants in any 104
bacterial population, genetic changes in bacterial populations have been shown to produce a 105
high persister state, producing increased numbers of antibiotic tolerant cells following exposure 106
to antibiotics in in vitro studies of pathogenic species (19, 20). Some of these genetic changes 107
have also been observed in clinical isolates; within a set of 477 commensal or urinary tract 108
infection isolates of Escherichia coli, 24 exhibited a mutation in the canonical persister gene 109
hipA, and the causality between a hipA7 mutation and a Hip phenotype confirmed by deleting 110
this allele from one of the clinical isolates (21). An investigation in young CF patients showed 111
an increase in persister phenotype in early/late infection isolate pairs from 14 patients. In this 112
study, 35 longitudinal P. aeruginosa isolates taken from one child over a 96-month period 113
showed increased levels of persister cells over time as well as an accumulation of 68 mutations 114
between the first and last isolate (16). However, the mutations in the single patient resembled 115
those known to accumulate in other CF patients over infection rather than any mutations 116
previously associated with the Hip phenotype in persister-focused in vitro studies. 117
118
To acquire a high-resolution pan-cohort perspective of persister emergence, genetic 119
mechanism, and impact in long-term infections, we have screened 460 longitudinal isolates of 120
P. aeruginosa collected from 39 young CF patients over a 10-year period from early 121
colonization onward for high persister variants (Hip, defined by survival of at least 75% of 122
replicates) tolerant to two unrelated antibiotics (ciprofloxacin and tobramycin). This unique 123
isolate collection allows us to determine Hip prevalence and dynamics during each colonizing 124
strain’s transition from environmental isolate to persistent pathogen. We describe relationships 125
between the Hip phenotype and response to each drug, the age of the isolate, and other adaptive 126
traits in longitudinal infections. We show that the Hip phenotype, defined in this study as a 127
strong and reliable recovery from antibiotic challenge that is a serious concern for the clinic, is 128
an independent and widespread trait. We further search for genetic and phenotypic changes 129
associated with the Hip phenotype in independent clonal lineages within distinct patients, 130
which may suggest adaptive routes to producing this phenotype. Finally, we show that the Hip 131
phenotype generally accumulates over time in patients via several archetypal patterns, appears 132
to contribute to long-term persistence of lineages, and increases the fitness of colonizing 133
populations of P. aeruginosa in antibiotic-treated CF patient lungs. 134
135
136
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
5
Results 137
The isolate collection 138
We examined a collection of 460 P. aeruginosa airway isolates obtained from 39 young CF 139
patients over a 10 year period while they were treated at the Copenhagen CF Centre at 140
Rigshospitalet (22). These patients represent a cohort aligned at the early infection stage and 141
undergoing similar treatment regimens per CF Centre guidelines, with repeated culture of P. 142
aeruginosa from their monthly sputum sampling within a time frame of 2-10 years (patient 143
inclusion was on a rolling basis over the study period in order to capture all early colonization 144
cases). Early isolates therefore represent bacteria that have not been exposed to substantial 145
antibiotic treatment before the study start excepting rare cases of strain transmission from 146
another patient. The bacterial CF isolates have been grouped into 52 genetically distinct clone 147
types (22), and while many patients retained a monoclonal infection during the entire course 148
of infection, half (n=20, 51.3 %) were infected at least transiently with another clone type. To 149
effectively account for these multi-clonal infections, clinical isolates are described by their 150
patient-specific lineage combining the clone type and the patient of origin (74 lineages in total). 151
Throughout this paper, we will also refer to ‘Time since first detection’ for each isolate, which 152
represents the length of time between first detection and subsequent isolations of the same 153
patient-specific lineage. 154
155
Identification of high persister (Hip) isolates by high-throughput screening 156
We screened the collection of P. aeruginosa isolates for the propensity to survive in the 157
presence of high concentrations of antibiotics. We chose two distinct antibiotics that matched 158
the following criteria; (i) they are frequently used to treat early P. aeruginosa infections in CF 159
patients; (ii) have contrasting targets and mechanisms of action; (iii) drive resistance 160
development with different dynamics in patients; and (iv) are each bactericidal toward 161
stationary phase P. aeruginosa (23, 24). On this basis, we performed two independent screens 162
using either the fluoroquinolone ciprofloxacin, which interacts with DNA gyrase, or the 163
aminoglycoside tobramycin, which acts upon bacterial ribosomes. Briefly, P. aeruginosa 164
subcultures in micro-titer plates were grown for 48 hours until they reached stationary phase, 165
after which they were challenged with antibiotics (100 μg/ml) for 24 hours before survival was 166
assessed (Fig. 1A). A standardised antibiotic concentration was used that was at least 25-times 167
higher than the European Committee on Antimicrobial Susceptibility Testing (EUCAST) 168
resistance breakpoint for each drug, minimising the chance that the screen selected for isolates 169
with modestly elevated minimum inhibitory concentrations (MICs). Within each antibiotic 170
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
6
screen, isolates were assayed eight times (technical quadruplicates performed in duplicate 171
biological experiments with a positive growth control for at least 3 of 4 replicates in each 172
experiment) and scored based on the capacity to re-grow after antibiotic treatment. An isolate 173
was given a score of 0 if it failed to re-grow in any replicate of an experiment, a score of 1 if it 174
grew once in both biological duplicates, a score of 2 if it grew in half of the technical replicates 175
in each experiment, a score of 3 if it grew in at least three replicates in each experiment, and a 176
score of 4 if it grew in all replicates (Fig. 1B, scores for the ciprofloxacin screen). To validate 177
our high-throughput screening approach, we selected 25 isolates, and enumerated CFUs 178
following 24 hours of ciprofloxacin treatment using standard survival assays. We observed a 179
significant positive correlation between persister score and CFU/ml following treatment (r2 180
0.5719, p < 0.0001, SI Appendix, Fig. S1), thus demonstrating the validity of our experimental 181
approach. 182
183
184
We defined high persister (Hip) isolates as those scoring either 3 or 4 (i.e. at least six of eight 185
technical replicates re-grew) and low persister (Lop) isolates as those scoring between 0 and 2. 186
This stringent scoring system was used to minimise the mis-classification of false Hips and 187
focus our analysis on isolates reliably producing persister cells, representing the most 188
concerning phenotype in a clinical environment. Isolates with a score of 4 made up the largest 189
Hip group, while the largest Lop group consisted of isolates scored as 0, failing to grow in any 190
replicate (Fig 1B). To validate this classification system, we selected six isolates from the same 191
patients, three of which were putative Hips, and three of which were classed as Lops and 192
performed time-dependent killing assays. The isolates displayed typical, biphasic killing (Fig. 193
1C). Each of the three putative Hip strains harboured a greater subpopulation of surviving 194
persister cells, thus they represent true Hips as defined by recent guidelines (25). In total, 195
24.8% of screened isolates (114 isolates) exhibited a persister phenotype in at least one of the 196
screens (Fig. 1D, AnyHip dataset). In Figure 2, we show the distribution of Hips across our 197
isolates with respect to (A) ciprofloxacin persistence (87 Cip Hip isolates), (B) tobramycin 198
persistence (60 Tob Hip isolates), and (C) multi-drug persistence (33 MD Hip isolates), where 199
an isolate scored 3 or more in both screens (Dataset 1). The bottom panel of Figure 2C shows 200
an explicit breakdown of dataset overlap. 201
202
The persister phenotype is antibiotic-specific and is dissociated from antibiotic resistance 203
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
7
As we were using antibiotics used in patient treatment to select for persister isolates, we 204
evaluated if we were simply selecting for resistant isolates as opposed to true Hips. 205
Ciprofloxacin and tobramycin MICs were determined using E-tests for isolates within the 206
collection. Most of the isolates were characterized as susceptible based on the EUCAST 207
breakpoints (Fig. 2A, B). Twenty-four ciprofloxacin susceptible isolates were Hip while 113 208
Lop variants were resistant. Similarly, 10 tobramycin resistant isolates were Lop while 50 Hip 209
isolates were tobramycin susceptible. We then contrasted the emergence of antibiotic resistance 210
versus Hip for all study lineages (Table S1). Of the 74 lineages assessed, resistance emerged 211
without Hip detection in 14 lineages (40%), while Hips emerged without resistance in 16 212
lineages. Resistance and persistence emerged simultaneously in 11 lineages. Resistance 213
preceded Hip in 9 lineages, while Hip preceded resistance in 5 lineages. If we used a less 214
stringent persister classification (persister score greater than 0 instead of 2), we observed that 215
resistance precedes persistence in 5 lineages, emerges simultaneously in 7 lineages, and follows 216
persistence in 15 lineages. Taken together, these data confirm that our screening approach 217
identified a persister phenotype separate from an antibiotic-resistant phenotype. 218
219
The persister phenotype is enriched in patient-specific lineages with slow-growing isolates 220
Along with changes in antibiotic susceptibility, the persister phenotype in our collection is not 221
arising in isolation from other adaptations. We and others have previously observed that CF 222
isolates adapt towards slow growth rates, increased resistance to antibiotics, and preference for 223
a biofilm lifestyle (18, 26, 27). A specific association between slowing growth rate and the Hip 224
phenotype has also been proposed (28). To probe interrelationships with other phenotypes, we 225
used a principle component analysis to evaluate the distribution of Hip (blue diamonds) versus 226
Lop (grey circles) variants by multiple traits under selection pressure in the CF lung. We see 227
that Any Hip variants group with isolates exhibiting more adapted traits (increased antibiotic 228
MICs and slowing growth), but they also appear across the full phenotypic space alongside 229
Lop isolates (Fig. 3A). 230
231
When comparing the ciprofloxacin versus tobramycin screen data (Fig. 3B-C), the data ellipses 232
enclosing the approximated majority of each population (68% of the population, t distribution) 233
show that Hip variants (blue ellipse) only partially separate from Lop variants (grey ellipse). If 234
we assess the first Hip variant of each lineage (FirstHip – blue ellipse with yellow fill), we see 235
FirstHips overlap substantially with both Lops and Hips. This variation of initial adaptive state 236
could be due to different adaptive trajectories with patients as well as lapses of time between 237
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
8
Hip emergence and isolation. The MD FirstHips show a more distinctive localization away 238
from Lops, which suggests further benefits of serious growth defects but could also be the 239
result of increased colonization time necessary to evolve both persister traits (Fig. 3D). If we 240
look at the likelihood of each lineage containing both Hips and slow growing isolates (where 241
the minimum growth rate of the lineage is less than 75% of the growth rate of PAO1), we see 242
a significant relationship between incidence of slow growing isolates and AnyHip isolates (Fig. 243
3E). In total, these results emphasize the complexity of selection pressures at play, resulting in 244
concurrent adaptation of distinct traits that may both influence each other and have related 245
genetic underpinnings. 246
247
Evolution of the persister phenotype is not genetically convergent across patient-specific 248
lineages 249
Sequencing and identification of genetic variations accumulating within each clone type have 250
been previously performed for most of the isolates used in this study (403 isolates, 46 lineages). 251
Genes targeted in convergent evolution were identified by the significant enrichment of 252
observed lineages with mutations in those genes compared to the number of lineages expected 253
to have mutations in the same genes according to genetic drift (derived from a simulated 254
evolution where lineages accumulate an equivalent number of mutations randomly for 1000 255
independent evolution simulations) (22). We split our dataset into Hip and Lop variants, and 256
then performed this same observed versus expected lineage enrichment analysis for each 257
population (see Materials and Methods for further details). The analysis was performed using 258
each of four datasets (ciprofloxacin Hip, tobramycin Hip, MD Hip and Any Hip), and the ratio 259
of lineage enrichment of mutated genes for Hip versus Lop variants allowed us to identify 260
candidate ‘hip’ genes for each set (Table 1 and Dataset 2). For completeness, we also performed 261
an additional genetic analysis focusing on mutations in non-coding sequences (Dataset 2). 262
263
In general, searching for hip genes accumulating non-synonymous mutations in Hip+ lineages 264
revealed only a weak signal for convergent evolution. Only one lineage assessed in the genetic 265
screen had only Hips present (the only isolate of the lineage assessed in our screen), so 266
practically all lineages with Hip isolates (Hip+ lineages, 29 included in the genetic study) also 267
contained Lops. Thus, mutated genes that were enriched 2-3 fold in independently evolved 268
Hip+ lineages were also frequently present in Lop isolates of the same lineage. Our lineage 269
enrichment ratio ultimately identified 12 mutated genes enriched in ciprofloxacin Hip+ 270
lineages (SI Appendix, Dataset 2), 13 mutated genes in tobramycin Hip+ lineages (SI 271
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
9
Appendix, Dataset 2), a solitary gene for MD Hip+ lineages (SI Appendix, Dataset 2), and 7 272
mutated genes for Any Hip+ lineages (Table 1). 273
274
Of note, there was a surprising lack of the most prominent ‘hip’ genes previously identified in 275
in vitro studies and screens of P. aeruginosa (SI Appendix, Table S2). None of the lineage 276
enrichment data pointed toward RNA endonuclease-type toxin-antitoxin systems under 277
adaptive selection, which supports recent research that has questioned the contribution of these 278
systems to persistence in numerous bacterial pathogens (29–31). Instead, they belonged to 279
diverse functional categories including transcriptional regulation/two component regulatory 280
systems (4 genes), energy metabolism (3 genes), and DNA replication and repair (2 genes). In 281
the Any Hip analysis as well as the MD Hip analysis, the shared top gene target, aceF, encodes 282
a component of the pyruvate dehydrogenase complex, a key player in central metabolism, for 283
which functional mutations are known to reduce growth rate (32–34). Other genes in Table 1 284
include major regulators rpoN, known to induce a growth defect when functionally mutated 285
(35), and retS, which when functionally mutated induces an array of phenotypic changes linked 286
to chronic infection such as a non-motile biofilm lifestyle (36, 37). These genes, as well as 287
hypothetical protein PA4311, also overlap with the ‘pathoadaptive’ mutationally enriched gene 288
list identified in our prior study of convergent evolution across all lineages (22). 289
290
Hip variants emerge via diverse incidence patterns 291
The lack of strong genetic signatures differentiating Hip from Lop isolates motivated us to 292
examine the temporal dynamics of high persister incidence using our comprehensive Any Hip 293
dataset. In half of the patients, the earliest bacterial isolate is also the first-ever identified P. 294
aeruginosa in the clinic and the other patients’ isolates also cover most of the initial 295
colonization phase. We can thus estimate the emergence of the Hip phenotype as P. aeruginosa 296
adapts from a wild type-similar naïve state into an adapted persistent pathogen. Previous 297
findings have indicated that the number of Hip variants from a lineage may increase over time 298
as the bacteria adapt to the antibiotic pressure in the host, and that once a Hip isolate is 299
observed, it is assumed to persist in the infecting population of the patient (7, 16). 300
301
To illustrate the range of persister dynamics we observe, we grouped each lineage by an array 302
of descriptors. The lineage descriptors include Hip presence versus absence (Hip+ vs Hip-), 303
transience of the lineage (whether it appears for less than 2 years, less than half the length of a 304
patient’s infection and is afterwards replaced by another lineage), continuity of Hip variants 305
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
10
(whether Hips are present for at least 3 sampling dates in a row), dominance of Hip variants 306
(whether Hips make up at least 2/3 of all isolates of a lineage), and whether a Hip variant 307
initiates the lineage. Figure 4A shows the ordered distribution of the lineages in 10 different 308
groups based on descriptor sets, illustrating the diversity of lineage Hip dynamics. We see that: 309
1) 34 of 74 lineages are Hip+, 2) 28 of 40 Hip- lineages are transient, while 4 Hip+ lineages 310
are transient, 3) 21 Hip+ lineages have Hip variants appear irregularly versus 7 lineages that 311
exhibit continuous periods and/or dominance of Hip variants, and 4) 12 lineages have initiating 312
Hip+ variants. Thus, the fraction (24.8%, Fig. 1D) of total isolates with a Hip phenotype 313
appears to be distributed over a subset of lineages (45.9%) in both stable (continuous/dominant) 314
and stochastic patterns of incidence, rather than present in every evolving lineage. 315
316
Hip variants accumulate over colonization time and occur rarely in transient lineages 317
We summarized the incidence of Any Hip variants over each lineage’s time of colonization in 318
Figure 4B. Here we plot the continuous counts of patients exhibiting the Hip phenotype (Hip+) 319
versus no Hip presence (Hip-) within the previous year (dashed lines) as well as the 320
accumulation of lineages that have exhibited a Hip variant at least once by a certain age of 321
colonization (solid line). The former illustrates both the number of lineages assessed at a given 322
colonization age and the increasing fraction of Hip+ versus Hip- lineages over time. The latter 323
shows how the likelihood of Hip emergence increases over time. Interestingly, when we plot 324
Hip accumulation for all four datasets (Fig. 4C), we also observe twice as many lineages with 325
initiating Hips in the Tob Hip dataset than the Cip Hip dataset, but Cip Hip incidence quickly 326
catches up with and eventually exceeds Tob Hip incidence over time. Overall, Any Hip variants 327
affect 34 lineages (Figure 4B-C) and 77% of the patients in our study cohort (Figure S2) by the 328
end of the study period. 329
330
Next, we evaluated the relationship between lineage transience and Hip presence. We first 331
mark the lineages present for less than 2 years in a patient at the end of their monitoring period 332
as ‘New’ lineages since we cannot determine transience without additional samples. Of the 56 333
remaining lineages, non-transient lineages are significantly associated with the Hip+ lineage 334
status, while transient lineages are significantly associated with the Hip- lineage status (Fig. 335
4D). Thus, a given patient often has multiple infecting lineages, but the Hip- lineages are much 336
more likely to disappear over the course of infection. In summary, we find that despite variable 337
incidence patterns, a clear majority of patients are infected by Hip+ lineages, and these lineages 338
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
11
have a significant persistence advantage in comparison to Hip- lineages over time, suggesting 339
that the Hip phenotype contributes to a fitness increase in antibiotic-treated patients. 340
341
Hip variants show increased fitness in patient-similar biofilms 342
We next asked if Hip isolates are able to survive antibiotic treatment better than Lop isolates 343
with similar antibiotic susceptibilities and growth properties in more complex conditions. We 344
simulated antibiotic treatment of CF patients in a recently developed biofilm 345
Pharmacokinetic/Pharmacodynamic (PK/PD) system, in which the bacteria are challenged 346
with antibiotics in much the same way as in patients (38). We chose this model because P. 347
aeruginosa often appears as biofilms in lungs of CF patients (39), because biofilms have been 348
shown to harbor increased levels of persister cells (40), and because our model mimics the 349
bacterial exposure to ciprofloxacin treatment as described for CF patients (38, 41). The isolates 350
that were chosen shared similar 1) time since their first detection in the CF lungs were similar, 351
2) MIC values for ciprofloxacin, 3) growth rates, and 4) belong to the same clone type. The 352
Hip/Lop pair was differentially tagged with yellow fluorescent protein, YFP (Hip), or cyan 353
fluorescent protein, CFP (Lop). Both strains formed biofilms with comparable biomasses in 354
the flow-cell system. Hip and Lop cells were then mixed 1:1 and allowed to form a mixed 355
biofilm. Representative images of the Hip/Lop biofilms are shown before and after treatment 356
with ciprofloxacin (Fig. 5A). 357
358
The majority of Lop bacteria were located close to the glass substratum with the Hip population 359
proliferating at the external surface of the biofilm, facing the liquid flow. The addition of 360
ciprofloxacin preferentially killed the Lop population leaving the Hip population relatively 361
unaffected by the antibiotic. COMSTAT analysis confirmed this changed population structure 362
after ciprofloxacin addition (Fig. 5B). This documentation of a Hip associated fitness increase 363
in an antibiotic containing environment is all the more striking, as it has been shown previously 364
that ciprofloxacin treatment of flow-cell biofilms preferentially kills the surface sub-365
populations of micro-colonies (42) – yet the Hip cells on the colony surfaces survive much 366
better than the internal Lop bacteria under treatment with ciprofloxacin. 367
368
Discussion 369
We have mapped the prevalence of persisters in a large, aligned cohort of patients under 370
intensive antibiotic treatment for a 10 year period (22, 41). Of 460 P. aeruginosa isolates from 371
the airways of 39 young CF patients (74 lineages in total), 24.8% of the isolates were scored 372
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
12
as robustly persisting Hip using a high-throughput screening approach to assay persistence 373
against ciprofloxacin or tobramycin (Fig. 1). We show that the isolates display different levels 374
of persisters, in accordance with the variance previously found between species and within 375
strains (5, 43, 44). Most adaptive changes occur during the first few years of colonization (18, 376
45), which matches our objective of searching for signs of increased fitness of Hip variants in 377
patients treated continuously with antibiotics. We show that in a young CF patient cohort 378
impacted by early longitudinal colonization by P. aeruginosa strains, Hip variants were 379
sampled from 77% of the patients (N=30) during a 10-year observation window. Our analysis 380
is a new and important comparative baseline for developing effective surveillance, impact 381
assessment, and eventual control of the persister phenotype in the clinic. 382
383
In the early years of infection after first detection of P. aeruginosa clone types, the Hip 384
phenotype appeared and disappeared over time in our routine clinical sampling (Fig. 4A). 385
While we see only partial overlap in Hip phenotype between our tobramycin and ciprofloxacin 386
screens, the number of lineages and patients that exhibit Hip variants increases over time for 387
all datasets (Fig. 4B-C, SI Appendix, Fig. S2), suggesting a selective advantage of this 388
phenotype during the continuation of antibiotic therapy. In general, the majority of lineages 389
that showed short-term colonization were made up of only Lop variants, which may partly 390
explain why they were unable to establish a persistent infection (Fig. 4D). These in-patient data 391
support the hypothesis that the Hip phenotype may generally have increased fitness in the 392
antibiotic-containing lung environment. It is, however, important to note that neither 393
dominance nor continuous presence of Hip variants is observed frequently (Fig. 4A). It is likely 394
that fitness trade-offs and clonal interference impact on the fitness properties and the 395
persistence level of Hip variants (14). 396
397
Multiple relationships between the Hip phenotype and other phenotypic traits such as growth 398
rate and antibiotic resistance have been suggested in the literature. While some studies point 399
out that there is no correlation between the mean growth rates of isolates and the observed Hip 400
phenotype (46–48), reduced growth rates have been associated with high persister phenotypes 401
in E. coli (28). A recent study in Salmonella enterica further supports that slow growth 402
(regardless of mechanism) promotes the persister phenotype (31). We see that lineages which 403
produce isolates with reduced growth rate are significantly more likely to also produce Hips. 404
Furthermore, multiple genes targeted for mutation at a higher rate in Hip+ lineages are known 405
to induce a growth defect (Table 1, SI Appendix, Dataset 2). However, we also observed the 406
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
13
Hip phenotype among naïve, fast growing clinical isolates of P. aeruginosa (Fig. 3), supporting 407
that other factors also influence the phenotype. Drug-tolerant cells have also been proposed to 408
facilitate evolution of true antibiotic resistance in E. coli in vitro (49). Intermittent antibiotic 409
exposure of a batch culture of E. coli selected for mutant clones harboring tolerance mutations 410
that increased the growth lag-time, during which tolerance to killing by ampicillin selected for 411
MIC-increasing mutations. Though P. aeruginosa in the CF lung is also exposed to fluctuating 412
concentrations of antibiotic, our stringently defined Hip phenotype emerges simultaneously or 413
after resistance in a majority of cases in contrast to these findings. We also observe more or 414
less an equal number of lineages where Hip variants and resistant clones evolve independently 415
in patients under antibiotic selection pressure, which has previously been suggested by 416
comparative studies of lab strains (50). In summary, our results suggest that the Hip phenotype 417
may be an early advantageous adaptation (18) arising stochastically in infected patients treated 418
with antibiotics. 419
420
In many ways, investigations of the genetic underpinnings of persisters have been performed 421
analogously to studies of antibiotic resistance, i.e. it was expected that a relatively limited set 422
of genes defines the phenotype. In a study of urinary tract infection E. coli isolates, a gain-of-423
function mutation in the HipA toxin was commonly observed (21). In contrast, a lack of 424
common targeted genes in a small collection of clinical Hip strains of Mycobacterium 425
tuberculosis suggested utilization of multiple genetic pathways (51). Working at a much larger 426
collection scale in a faster adapting organism, we do not see enrichment of mutations which 427
previously have been associated with Hip phenotypes in vitro. A role for the top proposed 428
persister target aceF has yet to be described, likely owing to its important role in growth; 429
mutants with severe growth defects are often overlooked in genome-wide analyses in vitro. In 430
support of this, an aceF transposon mutant is not available from the widely used PAO1 two-431
allele mutant library, and aceF did not appear in a recent persister screen of a pool of 100,000 432
unique PAO1 transposon mutants (20, 52). The aceF gene is mutated in 20% of the Hip+ 433
lineages, which have been genotypically evaluated in our large, longitudinal collection, but in 434
three of six lineages, it is also mutated in Lops. Meanwhile, no other enriched mutated genes 435
are affected in as many Hip+ lineages alone (where only Hip isolates are affected). We 436
therefore conclude that a Hip phenotype may derive from a diverse array of accumulating 437
genetic changes, and it is likely that more than one mutation often determines the persister level 438
in the respective bacterial populations. There are certainly many adaptive routes to slowed 439
growth rate, which we have previously demonstrated is a convergent adaptive outcome in this 440
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
14
early isolate collection (18). Our results likely reflect the multiple and dynamic selection 441
pressures in vivo, which challenge Hip variants in antibiotic-treated populations very 442
differently than those assessed in steady state in vitro conditions with only one selective force. 443
444
Many studies have shown the increased survival of persister cells under antibiotic treatment 445
(28, 53) and then screened for genetic determinants of persistence, but few have evaluated the 446
fitness of Hip versus Lop variants in direct competition experiments (43). In our study, we 447
tested a Lop/Hip pair of isolates matched by genotype, phenotype, and colonization age in 448
order to characterize the selective advantage of the Hip phenotype in a biofilm under treatment-449
replicating antibiotic exposure (38). We show that Hip cells survived ciprofloxacin treatment 450
far better than Lop isolates and this survival is potentially reliant on biofilm architecture. 451
Additionally, while homogeneous monoclonal P. aeruginosa biofilms treated with 452
ciprofloxacin show preferential killing of bacteria in the top layers (14, 42, 54), Lop bacteria 453
are preferentially killed in the deeper layers of the biofilm, showing an unexpected phenotype 454
worthy of further study. It is also striking that the in vitro biofilm fitness assessment shows 455
efficient elimination of the Lop strain in the presence of ciprofloxacin, whereas Hip variants 456
often coexist with Lop variants in vivo (Fig. 4A). This suggests that in the patient, direct 457
competition is likely limited by the large lung volume, many separate regional niches, and 458
influence of the host (55). 459
460
In summary, we have shown that Hip variants of P. aeruginosa emerge frequently in young 461
CF patients, and our results provide the first window into the evolving landscape of persistence 462
across a whole patient cohort. As pathogens increase their fitness in patients over time, they 463
clearly deploy the high persister phenotype as an important component in their survival 464
repertoire and can do so from the earliest stages of infection. It is still premature to conclude 465
that the high persister phenotype described here differs from what has been identified as Hip 466
in in vitro experimental conditions, but we consistently find a much broader bacterial repertoire 467
for survival in patient lungs. Hip variants do not seem to be mutated in genes previously found 468
from in vitro experiments to associate with Hip or in any strongly conserved genetic route. We 469
suggest that the difference in complexity of selection pressures when comparing in vitro and 470
in vivo environmental conditions results in highly different evolutionary trajectories. With our 471
investigation, we provide an important platform for broader clinically based studies and 472
contribute important new context for monitoring and one day hopefully preventing the high 473
persister phenotype in the clinic. 474
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
15
475
Materials and Methods 476
Strain collection. In total, we analyzed 460 P. aeruginosa airway isolates from young CF 477
patients followed at the Copenhagen CF-clinic at Rigshospitalet (Dataset 1). The local ethics 478
committee at the Capital Region of Denmark (Region Hovedstaden) approved the use of the 479
stored P. aeruginosa isolates: registration number H-4-2015-FSP. Phenotyping data for 434 480
isolates of this strain collection (growth rate in LB, adhesion in LB, and ciprofloxacin MIC) 481
have been previously published (18). We include additional tobramycin MIC measurements 482
and expand the complete trait dataset to 446 isolates. All available trait data is provided in 483
Dataset 1 along with the persister classification of each isolate and descriptive data. 484
485
Of the 460 isolates examined in the study, 403 isolates from 32 patients were described 486
previously in Marvig et. al. (22) and the remaining isolates were taken from seven previously 487
undescribed patients. The isolates were collected and stored at the Department of Clinical 488
Microbiology at Rigshospitalet, Copenhagen, Denmark, between 2002 and 2014. Of the 489
patients included in this study, 35.9% were diagnosed as chronically infected with P. 490
aeruginosa by the end of the study period. We defined chronicity based on the Copenhagen 491
CF Centre definition, whereby either P. aeruginosa has been detected in six consecutive 492
monthly sputum samples or fewer consecutive sputum samples combined with observation of 493
two or more P. aeruginosa-specific precipitating antibodies (41, 56). Intermittently colonized 494
patients were defined as patients where at least one isolate of P. aeruginosa is detected, and 495
normal levels of precipitating antibiotics against P. aeruginosa were observed. 496
497
High-throughput screening for Hip mutants. To determine the frequency at which P. 498
aeruginosa Hip mutants emerge in CF patients, we screened 460 isolates for the ‘persister’ 499
phenotype against either ciprofloxacin or tobramycin. Stock 96-well microtiter plates 500
containing 4 technical replicates of each isolate stored in glycerol (25 % v/v) were prepared 501
and stored at -80°C. Using a 96-well spot replicator, bacteria were transferred from the stock 502
plates into sterile 96-well microtiter plates containing 150 μl of Lysogeny Broth (LB) media. 503
Plates were incubated statically for 48 hours at 37°C until the bacteria reached the stationary 504
phase of growth. To determine the initial viability of bacteria in each well, the replicator was 505
used to spot bacteria onto LB agar plates. Subsequently, 100 μg/ml of either ciprofloxacin or 506
tobramycin was added to each well and the microtiter plates were incubated statically for a 507
further 20-24 hours at 37°C. Serial dilutions were performed in 96 well microtiter plates 508
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
16
containing 0.9 % NaCl using an automated fluid handling robot (Viaflo3844/ Integra 509
Biosciences AG). Each dilution was spotted onto LB agar plates using the replicator and plates 510
were incubated at 37°C for at least 24 hours. The growth of the bacteria was compared by 511
counting colonies whenever possible and visually inspecting growth on the plates before and 512
after antibiotic treatment. Experiments were performed in duplicate for each antibiotic. 513
514
Persister assay validation. Time-kill experiments were performed for six isolates from the 515
same lineage (3 Hip and 3 Lop). P. aeruginosa were inoculated in 3 ml of LB media in 14 ml 516
culture tubes and incubated for 48 hours at 37°C with shaking at 250 rpm. Following 517
incubation, each culture was serially diluted using sterile 0.9 % NaCl, plated onto LB agar and 518
incubated at 37°C to determine the initial colony forming units (CFU). The remaining culture 519
was treated with 100 μg/ml of ciprofloxacin and incubated at 37°C with shaking. Cultures were 520
washed and diluted in sterile 0.9 % NaCl, then spot plated onto LB agar 6 and 24 hours after 521
the addition of antibiotic. Plates were incubated for 24 hours at 37°C. Bacteria survival was 522
measured by counting CFU per ml. For an additional 19 isolates, the same validation 523
experiment was performed, however cultures were only plated out after 24 hours, which was 524
within the ‘persister plateau’. 525
526
Phenotype screening. The same frozen library of isolates used in the persister screening was 527
also replicated for assay of minimum inhibitory concentrations (MICs), bacterial growth, and 528
adhesion as described below and in Bartell et al (18). MICs for ciprofloxacin and tobramycin 529
were determined using E-test methodology according to the manufacturer’s recommendations 530
(Liofilchem®, Italy). To assay growth rate, bacteria were replicated from frozen plates into 531
96 well plates containing 150µL of LB medium, and incubated for 20 hours at 37°C with 532
constant shaking. OD 630 nm measurements were taken every 20 minutes using a microplate 533
reader (Holm & Halby, Copenhagen, Denmark/Synergy H1). Generation times (Td) were 534
determined on the best-fit line of a minimum of 3 points during exponential growth of the 535
bacterial isolate. Growth rates (hr-1) were calculated using the formula log (2)/ Td x 60 using 536
semi-automated code described in Bartell et al (18). Adhesion was measured via attachment 537
assays in 96-well plates using NUNC peg lids and 96 well plates with 150µl Luria broth 538
medium. OD600nm was measured after incubation for 20 hours at 37°C and subsequently, a 539
“washing microtiter plate” with 180µl PBS was used to wash the peg lids and remove non-540
adhering cells. After transfer of the peg lids to a microtiter plate containing 160µl 0.01% 541
crystal violet (CV), they were left to stain for 15 min. To remove unbound crystal violet, the 542
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
17
lids were then washed again three times in three individual “washing microtiter plates” with 543
180µl PBS. Adhesion was measured by detaching adhering CV stained cells through washing 544
the peg lids in a microtiter plate containing 180µl 99% ethanol. An ELISA reader was then 545
used to measure the CV density at OD590nm. (Microtiter plates were bought at Fisher 546
Scientific, NUNC Cat no. 167008, peg lids cat no. 445497). 547
548
Pharmacokinetic/Pharmacodynamic (PK/PD) flow chamber biofilm model. For fitness 549
experiments, we used a PK/PD biofilm model system combined with confocal laser-scanning 550
microscopy. This system simulates the changing antibiotic concentrations in CF patients during 551
intravenous dosing in addition to retaining a similar profile of antibiotic decay as the one taking 552
place in CF patients (38). First, Hip and Lop isolates were differentially tagged with a yellow 553
fluorescent protein (YFP) or cyan fluorescent protein (CFP) respectively (14). Flow chambers 554
were inoculated with a 1:1 mixture of Hip and Lop bacteria (each isolate had an initial OD600 555
of 0.5). Bacteria were incubated for one hour at 30 °C, then nutrient flow was applied to each 556
chamber (40x diluted LB at a rate of 20 ml/h using a Watson Marlow 205S peristaltic pump). 557
Biofilms were allowed to form for 72 hours, at which point flow was stopped and medium 558
containing ciprofloxacin was added. Peak ciprofloxacin concentrations were calculated to be 4 559
mg/L based on PK parameters generated from healthy patients and CF patients (57). The 560
medium was pumped from the dilution flask through the antibiotic flask to the flow chambers 561
at a constant rate calculated to mimic the elimination rate constant of the antibiotic for 24 hrs. 562
A confocal laser-scanning microscope (Zeiss LSM 510) equipped with an argon/krypton laser 563
and detectors was used to monitor YFP (excitation 514 nm, emission 530 nm), CFP (excitation 564
458 nm, emission 490 nm), and dead cells (propidium iodine, excitation 543 nm, emission 565 565
nm). Multichannel simulated fluorescent projections (SFPs) and sections through the biofilms 566
were generated using Imaris software (Bitplane AG, Switzerland). The images were later 567
analyzed using COMSTAT (58). The PK/PD biofilm experiments were performed using two 568
independent Hip/Lop isolate pairs. Pairs were taken from the same patient at a similar time 569
since first detection and had similar growth rates and ciprofloxacin MICs (Table S2). The data 570
presented are from 2 biological experiments with 4 independent images taken from each 571
experiment. 572
573
Lineage-based genetic analysis. To generate a list of mutated genes associated with the Hip 574
phenotype, we used previously generated whole-genome sequencing data and variant calling 575
filtered to obtain nonsynonymous mutations that had accumulated within a lineage after the 576
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
18
first isolate (22) to evaluate differential mutation patterns for Lop and Hip variants for 403 577
sequenced isolates. In this filtering process, we also removed mutations associated with any 578
known ‘hypermutator’ isolates based on a mutation in mutS or mutL to avoid the influence of 579
high random mutation in these isolates on the analysis. To identify genes that were mutated 580
more than would have been expected by drift/random mutation while accounting for lineage-581
based mutation accumulation over time, we adapted a statistical analysis of the relative 582
mutation enrichment by lineage. After separating Lop and Hip variants, we compared the 583
mutated-gene lineage enrichment ratios for each group - the number of lineages with observed 584
mutation(s) in a given gene divided by the number of lineages expected to have mutations in 585
that gene according to random mutation. This enrichment metric was obtained as follows for 586
each group: we determined the observed number of lineages mutated (sum-obs) in each gene. 587
Then we estimated the average number of lineages (avg-exp) that would have been mutated in 588
each gene if mutations were spread out randomly over the PAO1 genome. Using a random-589
roulette algorithm, the number of genes that were observed to be mutated in a given lineage 590
was spread out over the PAO1 genome for 1000 iterations, providing a mgene by niteration matrix 591
of randomly mutated gene profiles for each lineage. For the same iteration n across all lineages, 592
it was noted whether a given gene was mutated. This allowed us to determine an average 593
number of lineages expected to be mutated over 1000 iterations. If a gene was hit by chance 594
more than once in a single iteration, this would still only be denoted as one hit; this is in 595
alignment with our observed mutation assessment, where multiple isolates could be hit in the 596
same gene but we only noted whether or not the lineage was hit by unique mutations in the 597
specific gene. After obtaining the relative enrichment by lineage, a Poisson distribution was 598
used to calculate the probability of the observed given random drift (expected). We also divided 599
the lineage enrichment metric for genes mutated in Hip variants by that for Lop variants to 600
obtain a lineage enrichment ratio to identify targeted genes particularly impactful in the 601
evolution of the Hip population. 602
603
Data analysis and statistics. Analyses were conducted in RStudio v. 1.0.143 and R v. 3.4.0 604
with visualization package ggplot2 v. 3.0.0. Lineage set analysis was performed using UpSetR 605
v. 1.3.3 in R (59). Principal component analysis was performed in R using ‘prcomp’ with 606
centered and scaled phenotype data (Dataset 1). Mosaic plots (visualizing multi-way 607
contingency tables) showing the association between two variables via the conditional relative 608
frequency and significant associations based on a Pearson X2 test were created using vcd v. 609
1.4-4 in R (60). 610
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
19
611
Data availability. Screen data for the persister phenotype is provided as a supplemental 612
dataset (Dataset 1) including trait data, which partially overlaps with trait data published 613
previously (18). Code in R for mosaic plots, PCA analysis, and set intersection analysis of 614
lineage characteristics is available on request. Lineage enrichment analysis was performed in 615
Anaconda3 v. 4.0.0 using custom scripts available on request. 616
617
Acknowledgments 618
This work was supported by Cystic Fibrosis Foundation Pilot and Feasibility Award to KL. 619
HKJ was supported by The Novo Nordisk Foundation (NNF12OC1015920 and 620
NNF15OC0017444), by Rigshospitalets Rammebevilling 2015-17 (R88-A3537), by 621
Lundbeckfonden (R167-2013-15229), by RegionH Rammebevilling (R144-A5287) and by 622
Independent Research Fund Denmark (DFF-4183-00051). JAB was supported by 623
postdoctoral fellowships from the Whitaker Foundation and the Cystic Fibrosis Foundation 624
(BARTEL18F0). We thank Katja Bloksted, Ulla Rydahl Johansen, Helle Nordbjerg 625
Andersen, Sarah Buhr Bendixen, Camilla Thranow, Pia Poss, Bonnie Horsted Erichsen and 626
Rakel Schiøtt for excellent technical assistance at Rigshospitalet. We thank Prof. Vasili 627
Hauryliuk for helpful comments. 628
629
Conflict of interest 630
The authors declare no competing financial interests. 631
632
Author Contributions 633
SM and KL designed the study. HKJ collected all the bacterial isolates. BM, DRC, JAB, and 634
JAJH performed all experiments. JAB and LMS performed genetic and lineage-based data 635
analysis. All authors contributed to the writing of the manuscript. All authors approved the 636
final version. 637
638
Ethics approval 639
The local ethics committee at the Capital Region of Denmark (Region Hovedstaden) 640
approved the use of the stored P. aeruginosa isolates: registration number H-4-2015-FSP. We 641
confirm that all methods were performed in accordance with the relevant guidelines and 642
regulations. 643
644
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
20
Supplementary information 645
Supplementary methods. Description of phenotyping methods, biofilm flow chamber 646
experiments, and lineage enrichment analysis. 647
Figure S1. Correlation between persister score and colony forming units following treatment 648
with ciprofloxacin. 649
Figure S2. Hip accumulation in patients. 650
Table S1. Emergence of resistance versus persistence across 74 lineages. 651
Table S2. Comparison of persister genes identified in previous P. aeruginosa studies to 652
mutated genes highlighted by our lineage analysis. 653
Dataset 1. Phenotypic dataset for all isolates. 654
Dataset 2. Lineage-based mutation enrichment analysis results for coding genes and noncoding 655
regions (separately) for all datasets. 656
657
References 658
1. Bigger JW (1944) Treatment of staphylococcal infections with penicillin. Lancet 659
2(6320):497–500. 660
2. Wood TK, Knabel SJ, Kwan BW (2013) Bacterial persister cell formation and 661
dormancy. Appl Environ Microbiol 79(23):7116–7121. 662
3. Rowe SE, Conlon BP, Keren I, Lewis K (2016) Persisters: Methods for Isolation and 663
Identifying Contributing Factors—A Review. Methods Mol Biol 1333:17–28. 664
4. Lewis K (2010) Persister Cells. Annu Rev Microbiol. 665
doi:10.1146/annurev.micro.112408.134306. 666
5. Stewart B, Rozen DE (2011) Genetic variation for antibiotic persistence in Escherichia 667
coli. Evolution (N Y) 66(3):933–939. 668
6. Hofsteenge N, van Nimwegen E, Silander OK (2013) Quantitative analysis of persister 669
fractions suggests different mechanisms of formation among environmental isolates of 670
E. coli. BMC Microbiol 13(1):25. 671
7. LaFleur MD, Qi Q, Lewis K (2010) Patients with long-term oral carriage harbor high-672
persister mutants of Candida albicans. Antimicrob Agents Chemother 54(1):39–44. 673
8. Salsgiver EL, et al. (2016) Changing Epidemiology of the Respiratory Bacteriology of 674
Patients With Cystic Fibrosis. Chest 149(2):390–400. 675
9. Parkins MD, Somayaji R, Waters VJ (2018) Epidemiology, Biology, and Impact of 676
Clonal Pseudomonas aeruginosa Infections in Cystic Fibrosis. Clin Microbiol Rev 677
31(4):1–38. 678
10. Dalemans W, et al. (1991) Altered chloride ion channel kinetics associated with the 679
ΔF508 cystic fibrosis mutation. Nature 354(6354):526–528. 680
11. Boucher RC (2007) Cystic fibrosis: a disease of vulnerability to airway surface 681
dehydration. Trends Mol Med 13(6):231–240. 682
12. Folkesson A, et al. (2012) Adaptation of Pseudomonas aeruginosa to the cystic fibrosis 683
airway: an evolutionary perspective. Nat Rev Microbiol 10(12):841–51. 684
13. Gibson RL, Burns JL, Ramsey BW (2003) Pathophysiology and management of 685
pulmonary infections in cystic fibrosis. Am J Respir Crit Care Med 168(8):918–951. 686
14. Frimodt-Møller J, et al. (2018) Mutations causing low level antibiotic resistance ensure 687
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
21
bacterial survival in antibiotic-treated hosts. Sci Rep 8(1):1–13. 688
15. Burns JL, et al. (1999) Effect of Chronic Intermittent Administration of Inhaled 689
Tobramycin on Respiratory Microbial Flora in Patients with Cystic Fibrosis. J Infect 690
Dis 179(5):1190–1196. 691
16. Mulcahy LR, Burns JL, Lory S, Lewis K (2010) Emergence of Pseudomonas 692
aeruginosa strains producing high levels of persister cells in patients with cystic 693
fibrosis. J Bacteriol 192(23):6191–9. 694
17. van Gestel J, Vlamakis H, Kolter R (2015) Division of labor in biofilms: the ecology 695
of cell differentiation. Microbiol Spectr 3(2):1–24. 696
18. Bartell JA, et al. (2019) Evolutionary highways to persistent bacterial infection. Nat 697
Commun 10(1):629. 698
19. Moyed HS, Bertrand KP (1983) hipA, a newly recognized gene of Escherichia coli K-699
12 that affects frequency of persistence after inhibition of murein synthesis. J Bacteriol 700
155(2):768–75. 701
20. Cameron DR, Shan Y, Zalis EA, Isabella V, Lewis K (2018) A Genetic Determinant 702
of Persister Cell Formation in Bacterial Pathogens. J Bacteriol 200(17):1–11. 703
21. Schumacher MA, et al. (2015) HipBA-promoter structures reveal the basis of heritable 704
multidrug tolerance. Nature 524(7563):59–66. 705
22. Marvig RL, Sommer LM, Molin S, Johansen HK (2015) Convergent evolution and 706
adaptation of Pseudomonas aeruginosa within patients with cystic fibrosis. Nat Genet 707
47(1):57–64. 708
23. Eliopoulos GM, Gardella A, Moellering RC (1984) In vitro activity of ciprofloxacin, a 709
new carboxyquinoline antimicrobial agent. Antimicrob Agents Chemother 25(3):331–710
335. 711
24. Su HC, et al. (2010) The development of ciprofloxacin resistance in Pseudomonas 712
aeruginosa involves multiple response stages and multiple proteins. Antimicrob Agents 713
Chemother 54(11):4626–4635. 714
25. Balaban NQ, et al. (2019) Definitions and guidelines for research on antibiotic 715
persistence. Nat Rev Microbiol 17(July):441–448. 716
26. La Rosa R, Johansen HK, Molin S (2018) Convergent Metabolic Specialization 717
through Distinct Evolutionary Paths in Pseudomonas aeruginosa. MBio 9(2):e00269-718
18. 719
27. Winstanley C, O’Brien S, Brockhurst MA (2016) Pseudomonas aeruginosa 720
Evolutionary Adaptation and Diversification in Cystic Fibrosis Chronic Lung 721
Infections. Trends Microbiol 24(5):327–337. 722
28. Balaban NQ, Merrin JM, Chait R, Kowalik L, Leibler S (2004) Bacterial Persistence 723
as a Phenotypic Switch. Science (80- ) 305(5690):1622–1625. 724
29. Goormaghtigh F, et al. (2018) Reassessing the Role of Type II Toxin-Antitoxin 725
Systems in Formation of Escherichia coli Type II Persister Cells. MBio 9(3):1–14. 726
30. Conlon BP, et al. (2016) Persister formation in Staphylococcus aureus is associated 727
with ATP depletion. Nat Microbiol 1(5):16051. 728
31. Pontes MH, Groisman EA (2019) Slow growth determines nonheritable antibiotic 729
resistance in Salmonella enterica. Sci Signal 12(592):eaax3938. 730
32. Schutte KM, et al. (2016) Escherichia coli Pyruvate Dehydrogenase Complex Is an 731
Important Component of CXCL10-Mediated Antimicrobial Activity. Infect Immun 732
84(1):320–328. 733
33. Kim Y, Ingram LO, Shanmugam KT (2007) Construction of an Escherichia coli K-12 734
Mutant for Homoethanologenic Fermentation of Glucose or Xylose without Foreign 735
Genes. Appl Environ Microbiol 73(6):1766–1771. 736
34. Chung JCS, Rzhepishevska O, Ramstedt M, Welch M (2013) Type III secretion 737
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
22
system expression in oxygen-limited Pseudomonas aeruginosa cultures is stimulated 738
by isocitrate lyase activity. Open Biol 3(1):120131. 739
35. Viducic D, Murakami K, Amoh T, Ono T, Miyake Y (2017) RpoN Promotes 740
Pseudomonas aeruginosa Survival in the Presence of Tobramycin. Front Microbiol 741
8(MAY):839. 742
36. Goodman AL, et al. (2004) A signaling network reciprocally regulates genes 743
associated with acute infection and chronic persistence in Pseudomonas aeruginosa. 744
Dev Cell 7(5):745–54. 745
37. Ventre I, et al. (2006) Multiple sensors control reciprocal expression of Pseudomonas 746
aeruginosa regulatory RNA and virulence genes. Proc Natl Acad Sci 103(1):171–176. 747
38. Haagensen JAJ, Verotta D, Huang L, Spormann A, Yang K (2015) New in vitro model 748
to study the effect of human simulated antibiotic concentrations on bacterial biofilms. 749
Antimicrob Agents Chemother 59(7):4074–4081. 750
39. Singh PK, et al. (2000) Quorum-sensing signals indicate that cystic fibrosis lungs are 751
infected with bacterial biofilms. Nature 407:762–764. 752
40. Spoering AL, Lewis K (2001) Biofilms and planktonic cells of Pseudomonas 753
aeruginosa have similar resistance to killing by antimicrobials. J Bacteriol 754
183(23):6746–6751. 755
41. Johansen HK, et al. (2004) Antibody response to Pseudomonas aeruginosa in cystic 756
fibrosis patients: a marker of therapeutic success? - A 30-year cohort study of survival 757
in Danish CF patients after onset of chronic P. aeruginosa lung infection. Pediatr 758
Pulmonol 37(5):427–432. 759
42. Pamp SJ, Gjermansen M, Johansen HK, Tolker-Nielsen T (2008) Tolerance to the 760
antimicrobial peptide colistin in Pseudomonas aeruginosa biofilms is linked to 761
metabolically active cells, and depends on the pmr and mexAB-oprM genes. Mol 762
Microbiol 68(1):223–240. 763
43. Stepanyan K, et al. (2015) Fitness trade-offs explain low levels of persister cells in the 764
opportunistic pathogen Pseudomonas aeruginosa. Mol Ecol 24(7):1572–1583. 765
44. Bink A, et al. (2011) Superoxide dismutases are involved in Candida albicans biofilm 766
persistence against miconazole. Antimicrob Agents Chemother 55(9):4033–4037. 767
45. Yang L, et al. (2011) Evolutionary dynamics of bacteria in a human host environment. 768
Proc Natl Acad Sci 108(18):7481–7486. 769
46. Fung DKC, Chan EWC, Chin ML, Chan RCY (2010) Delineation of a bacterial 770
starvation stress response network which can mediate antibiotic tolerance 771
development. Antimicrob Agents Chemother 54(3):1082–1093. 772
47. Keren I, Minami S, Rubin E, Lewis K (2011) Characterization and transcriptome 773
analysis of Mycobacterium tuberculosis persisters. MBio 2(3):3–12. 774
48. Wakamoto Y, et al. (2013) Dynamic persistence of antibiotic-stressed Mycobacteria. 775
Science (80- ) 339(6115):91–95. 776
49. Levin-Reisman I, et al. (2017) Antibiotic tolerance facilitates the evolution of 777
resistance. Science (80- ) 355(6327):826–830. 778
50. Vogwill T, Comfort AC, Furió V, MacLean RC (2016) Persistence and resistance as 779
complementary bacterial adaptations to antibiotics. J Evol Biol 29(6):1223–1233. 780
51. Torrey HL, Keren I, Via LE, Lee JS, Lewis K (2016) High Persister Mutants in 781
Mycobacterium tuberculosis. PLoS One 11(5):e0155127. 782
52. Held K, Ramage E, Jacobs M, Gallagher L, Manoil C (2012) Sequence-verified two-783
allele transposon mutant library for Pseudomonas aeruginosa PAO1. J Bacteriol 784
194(23):6387–6389. 785
53. Spoering AL, Vulić M, Lewis K (2006) GlpD and PlsB participate in persister cell 786
formation in Escherichia coli. J Bacteriol 188(14):5136–5144. 787
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
23
54. Haagensen JAJ, et al. (2007) Differentiation and distribution of colistin- and sodium 788
dodecyl sulfate-tolerant cells in Pseudomonas aeruginosa biofilms. J Bacteriol 789
189(1):28–37. 790
55. Jorth P, et al. (2015) Regional Isolation Drives Bacterial Diversification within Cystic 791
Fibrosis Lungs. Cell Host Microbe 18(3):307–319. 792
56. Høiby N, et al. (1977) Pseudomonas aeruginosa infection in cystic fibrosis - 793
Diagnostic and prognostic significance of Pseudomonas aeruginosa precipitins 794
determined by means of crossed immunoelectrophoresis. Scand J Respir Dis 58(2):65–795
79. 796
57. Touw DJ, Knox AJ, Smyth A (2007) Population pharmacokinetics of tobramycin 797
administered thrice daily and once daily in children and adults with cystic fibrosis. J 798
Cyst Fibros 6(5):327–333. 799
58. Heydorn A, et al. (2002) Statistical Analysis of Pseudomonas aeruginosa Biofilm 800
Development: Impact of Mutations in Genes Involved in Twitching Motility, Cell-to-801
Cell Signaling, and Stationary-Phase Sigma Factor Expression. Appl Environ 802
Microbiol 68(4):2008–2017. 803
59. Conway JR, Lex A, Gehlenborg N (2017) UpSetR: An R package for the visualization 804
of intersecting sets and their properties. Bioinformatics 33(18):2938–2940. 805
60. Meyer D, Zeileis A, Hornik K (2006) The Strucplot Framework: Visualizing Multi-806
way Contingency Tables with vcd. J Stat Softw 17(3):1–48. 807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
24
Figures 829
830
831
Figure 1. High-throughput screening approach for isolates with a high persister (Hip) 832
phenotype. (A) A large collection of Pseudomonas aeruginosa clinical isolates were grown to 833
stationary phase in quadruplicate wells for two biological replicate (BR) experiments (each 834
isolate tested 8 times in total). The screen was performed twice, using either ciprofloxacin or 835
tobramycin. Each isolate was treated with 100 g/ml of antibiotic for 24 hours, while growth 836
was assessed by plating on LB agar. Following antibiotic treatment, cultures were diluted then 837
plated on agar, at which point survival was assessed. Each isolate was given a persister score 838
based on consistent replicate survival following treatment. Isolates for which 3-4 replicates 839
survived for each BR were given a score of 3-4, respectively, and were considered high 840
persisters (Hip). Isolates with a respective score of 0-2 were considered low persisters (Lop). 841
(B) Score distribution of P. aeruginosa Hip (blue) and Lop (grey) isolates against ciprofloxacin. 842
(C) Traditional time-kill assays were performed for three Hip (HipIso 1, 2, 3) and three Lop 843
isolates (LopIso 1, 2, 3) from the same patient to validate the high throughput screen. Colony 844
forming units (CFU) per ml were determined following treatment with 100 g/ml of 845
ciprofloxacin. Data are the mean of 6 independent cultures, bars represent SEM. (D) 846
Distribution of P. aeruginosa Hip (blue) and Lop (grey) isolates for the AnyHip dataset, where 847
Hips were given a score of 3 or 4 in at least one of the antibiotic screens. 848
849
850
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
25
851
Figure 2. Persister screening results. (A) Distribution of Hip (blue) and Lop (grey) isolates 852
following ciprofloxacin (CipHip,) or (B) tobramycin treatment (TobHip), and (C) the overlap 853
between these datasets (MDHip - multi-drug Hip). For A and B, the bottom panel shows a 854
mosaic plot (multi-way contingency table) for the comparison of isolate persister class versus 855
resistance class for each antibiotic. Hip and Lop persister isolates were classified as susceptible 856
(S) or resistant (R) according to EUCAST breakpoints (cip: S≤0.5µg/ml, tob: S≤4µg/ml) based 857
on their MIC obtained via E-test. The area of each cell is proportional to the frequency of 858
isolates with the indicated combination of Hip and resistance classification, and resistance-859
associated cells are further highlighted by orange borders. The bottom panel of C shows the 860
overlap of Hip and Lop isolates under ciprofloxacin versus tobramycin treatment. MDHip are 861
highlighted by a black border. 862
863
864
865
866
867
868
869
870
871
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
26
872
Figure 3. High persisters in the multi-trait landscape. (A) Lop (grey circle), Cip Hip (blue 873
squares), Tob Hip (blue triangles) and MD Hip isolates (black diamonds) were analyzed via 874
principle component analysis with respect to their similarity with other infection-linked traits: 875
growth rate (GR_LB), adhesion, ciprofloxacin MIC (cip) and tobramycin MIC (tob). 446 876
isolates with complete trait sets were included. Hip isolates do not consistently cluster with any 877
one additional trait. For (B) Cip Hip, (C) Tob Hip and (D) MDHip, the first Hip isolates from 878
a lineage (FirstHip, yellow triangles) were highlighted as Hip variants with mitigated effects 879
of other accumulating mutations within the lineage to improve cross-lineage comparison. In 880
each case, FirstHip and the remaining Hip isolates shift to various degrees from ‘naïve’ towards 881
‘adapted’ levels given the particular Hip dataset. We illustrate this using data ellipse enclosing 882
samples approximately within the first standard deviation (t distribution, 68% of the set) for 883
isolate sets characterized as FirstHip (yellow ellipse), and the remaining Hips (blue ellipse). 884
(E) We visualized the association between lineages that produced Hips versus slow growing 885
isolates (identified by the minimum growth rate of lineage isolates falling below 75% of the P. 886
aeruginosa PAO1 growth rate based on a 45 minute generation time in LB). Association 887
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
27
between variables is illustrated by a mosaic plot (multi-way contingency table visualization) 888
where color indicates significant deviation from the expected frequency of lineages in each cell 889
under trait independence using Pearson’s chi-squared test. 890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
28
922
Figure 4. Hip Persister incidence patterns from a lineage-based perspective. (A) Lineages 923
were classed according to several nested characteristics: transient versus non-transient 924
lineages, Hip presence (AnyHip, meaning Cip Hip, Tob Hip or MD Hip), continuous periods 925
of isolated Hips, lineage-initiating Hips, and Hips dominating a lineage. Lineages representing 926
each combination of traits are shown on the left (Hip blue diamonds, Cip grey circles), while 927
characteristic sets are identified and enumerated for the entire collection on the right. 928
(B) For the AnyHip dataset, the continuous patient count of Hip- patients (grey circles) versus 929
Hip+ patients (blue diamonds) for the prior year of colonization is plotted, while the 930
accumulating count of Hip+ patients from time 0 is shown by black diamonds. (C) The 931
accumulating count of Hip+ patients is shown for all four datasets (AnyHip as black circles, 932
CipHip as transparent blue squares, TobHip as transparent blue triangles, and MDHip as dark 933
blue diamonds). (D) Transient lineages (lineages of shorter than 2 years duration, less than 934
50% of total patient infection length, and which are followed by the appearance of a new 935
lineage) are significantly associated with lineages lacking Hips (Hip-), while non-transient 936
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
29
lineages are associated with the presence of Hips based on Pearson's chi-squared test (via a 937
mosaic plot visualizing a multi-way contingency table). Transience-unclassifiable lineages of 938
shorter than 2 years’ duration at the end of a patient’s collection period are shown for context 939
(‘New’). 940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
30
971
Figure 5. Fitness comparison of Lop and Hip isolates in biofilm conditions. (A) A 972
representative Lop and Hip isolate with similar characteristics (Lop: cip MIC 1.0 µg/ml; 973
growth rate 0.27 hr-1; time since first detection 4.28 years. Hip: cip MIC 0.75 µg/ml; growth 974
rate 0.25 hr-1; time since first detection 5.49 years) were differentially tagged with CFP (Lop) 975
or YFP (Hip). Tagged isolates were cocultured and allowed to form biofilms in a flow-cell 976
model for 72 hours. Mixed biofilms were treated for 24 hours with ciprofloxacin (4 µg/ml). 977
Propidium iodine (PI) was added to visualise dead cells (red). (B) Biomass was quantified for 978
each population. Significant differences in biomass following treatment were determined using 979
unpaired t-test (*** p <0.001). 980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
31
Tables 1006
1007
Table 1. Lineage-based mutation enrichment analysis. Mutated genes enriched in AnyHip 1008
versus Lop dataset as assessed from a convergent evolution perspective accounting for 1009
lineage adaptation. Lineage enrichment ratio was calculated by dividing lineage-based gene 1010
mutation enrichment within Hip variants by that within Lop variants for each gene. Top Hip-1011
linked genes were selected via the following criteria: greater than 2 lineages presenting 1012
mutations in that gene in the Hip population and a lineage enrichment ratio greater than 2. 1013
The only gene also enriched in MDHip lineages, aceF, is in boldface. 1014
1015
Locus Gene Product Hip+ Lineage Count
Hip- Lineage Count
Lineage Enrichment
Ratio
Total Lineages
Hit
PA5016 aceF dihydrolipoamide acetyltransferase
6 3 3.65 6
PA0339 hypothetical protein 3 2 2.75 3
PA4462 rpoN RNA polymerase sigma-54 factor
3 2 2.58 4
PA3640 dnaE DNA polymerase III, alpha chain
4 3 2.41 4
PA4311 conserved hypothetical protein
4 3 2.15 5
PA4856 retS RetS (Regulator of Exopolysaccharide and Type III Secretion)
6 5 2.06 6
PA2894 hypothetical protein 4 3 2.01 4
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
32
Supplementary Information for 1037
1038
Bacterial persisters in long-term infection: emergence and fitness in a complex host 1039
environment 1040
1041
Authors 1042
Jennifer A. Bartell1*, David R. Cameron2,3*, Biljana Mojsoska4,6*, Janus Anders Juul 1043
Haagensen1, Lea M. Sommer4, Kim Lewis2§, Søren Molin1, Helle Krogh Johansen4,5§ 1044
1045
Affiliations: 1046
1 The Novo Nordisk Foundation Center for Biosustainability, Technical University of 1047
Denmark, Lyngby, Denmark 1048
2 Antimicrobial Discovery Center, Northeastern University, Boston, USA 1049
3 Department of Intensive Care Medicine, Inselspital, Bern University Hospital, University of 1050
Bern, Switzerland 1051
4 Department of Clinical Microbiology, Rigshospitalet, Copenhagen, Denmark 1052
5 Department of Clinical Medicine, University of Copenhagen, Copenhagen, Denmark 1053
6 Present address: Department of Science and Environment, Roskilde University, Roskilde, 1054
Denmark 1055
1056
*Contributed equally to the work and listed alphabetically 1057
§Corresponding authors 1058
1059
1060
This PDF file includes: 1061
Supplementary Figures 1 and 2 1062
Tables S1-S2 1063
Captions for Datasets 1 and 2 1064
References for SI reference citations 1065
1066
Other supplementary materials for this manuscript include the following: 1067
Datasets 1 and 2 1068
1069
1070
1071
1072
1073
1074 1075
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
33
Supplementary Figures 1076
1077 1078 Figure S1. Correlation between persister score and colony forming units following treatment 1079 with ciprofloxacin. P. aeruginosa isolates representing each of the scores possible from the high-1080 throughput screen (0-4) were treated with 100 µg/ml of ciprofloxacin for 24 hours, then plated on agar 1081 for surviving CFU determination. Each isolate was tested independently at least 4 times. The data are 1082 represented by the mean and SEM. 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107
0 1 2 3 4
2
4
6
8
Persister score
log
(CF
U/m
l)
r2= 0.5719p < 0.0001
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
34
1108 1109 Figure S2. Hip accumulation in patients. (A) Hip- (gray squares) and Hip+ (blue diamonds) show 1110 the continuous count of patients with Hip- lineage(s) versus Hip+ lineage(s) for the prior year of 1111 colonization, while the accumulating count of patients with Hip+ lineages from time 0 is shown by 1112 black circles (AnyHip). (B) Accumulated patients with Hip+ lineages are shown for all persister 1113 datasets. 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
35
Supplementary Tables 1148 1149
Table S1. Emergence of resistant versus persister isolates across 74 lineages via assessment 1150
of AnyHip isolates (persister score greater than 2) versus AnyHip plus transition isolates 1151
(persister score greater than 0). 1152
1153
Persister score
Category # Lineages affected Cip Tob Any
AnyHips (Persister score > 2)
Persisters alone 2 16 16
Resistance alone 14 4 14 Simultaneous emergence 11 1 11
Persisters before Resistance 2 3 5 Resistance before Persisters 8 2 9
AnyHips plus
transition variants
(Persister score > 0)
Persisters alone 10 45 48
Resistance alone 8 1 8 Simultaneous emergence 7 2 7
Persisters before Resistance 15 7 15
Resistance before Persisters 5 0 5
1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
36
Table S2. Persister genes identified in previous P. aeruginosa studies which we did not observe as 1185 targets in our lineage enrichment analysis. 1186
Locus Gene Product Function Reference
PA5338 spoT Guanosine-3',5'-
bis(diphosphate) 3'-
phyrophosphohydrolase
Adaptation, protection;
Nucleotide
biosynthesis and
metabolism
(8)
PA0934 relA GTP pyrophokinase Adaptation, protection (8)
PA3622 rpoS Sigma factor RpoS Transcription (9, 10)
PA4723 dksA DksA
Transcription; DNA
replication
(8)
PA5002 dnpA de-N-acetylase Membrane proteins (11)
PA1045 Hypothetical protein Unknown (11)
PA0299 spuC Polyamine:pyruvate
transaminase
Carbon compound
catabolism
(11)
PA3589 Probable acyl-CoA
thiolase
Carbon compound
catabolism
(11)
PA3883 Probable short-chain
dehydrogenase
Putative enzyme (11)
PA5261 algR AlgR Two-component
regulatory system;
secreted factors
(11)
PA0318 Hypothetical protein Putative enzymes (11)
PA0409 pilH PilH Motility, chemotaxis;
Two-component
regulatory system
(11)
PA3166 pheA Chorismate mutase Amino acid
biosynthesis and
metabolism
(11)
PA5332 crc Catabolite repression
control protein
Carbon compound
catabolism; energy
metabolism
(12)
PA4756 carB Carbamoylphosphate
synthase
Nucleotide
metabolism; Amino
(13)
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
37
acid biosynthesis and
metabolism
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
38
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
39
Dataset Descriptions 1215
1216 Dataset 1. Isolate collection data, metadata, and screen results. Description of all isolates 1217 including age, genotypic, and phenotypic information used in the analysis. Identification of Hips in 1218 CipHip, TobHip, MDHip, and AnyHip datasets as well as Persister Scores in each antibiotic screen. 1219 Specific legend included below and in dataset file. 1220 1221 ID: The id of the isolate. 1222 Sequence_ID: Sequence id of isolate for internal use and can be referenced to the isolate collection 1223 published and analyzed in Marvig et al.: "Convergent evolution and adaptation of Pseudomonas 1224 aeruginosa within patients with cystic fibrosis", Nature Genetics 47, (2015). 1225 SRA_accNo: Accession number for sequences published previously in Marvig et al.: "Convergent 1226 evolution and adaptation of Pseudomonas aeruginosa within patients with cystic fibrosis", Nature 1227 Genetics 47, (2015). 1228 Genotype: The clone type/genotype of the isolate. 1229 Patient: The patient wherefrom the isolate has been sampled. 1230 Lineage: Patient-Genotype combination distinguishing specific clone types evolving in a given 1231 patient. 1232 Date_no: Date of sampling, using fraction of year. 1233 IageCT: The "Infection age of the Clone Type", the time (in years) since the clone type of the specific 1234 isolate was first detected in the patient from which the specific isolate was sampled. This is referred to 1235 in the study as "the time since first detection" or "colonization time" or "age of the lineage". 1236 Length1PA: The time (in years) since P. aeruginosa was first positively cultured from the patient's 1237 lungs, regardless of clone type. 1238 PersisterScore: Score of 0-4 according to minimum number of spotted cultures that grew after 1239 antibiotic exposure (minimum in 2 biological replicates of 4 spots, 8 total spots). 1240 cip: MIC of Ciprofloxacin. 1241 tob: MIC of Tobramycin. 1242 AdhesionN: OD of crystal violet normalised against 20h of growth. 1243 GR_LB: Growth rate in LB, h-1. 1244 cipR: Isolate classification as ciprofloxacin sensitive or resistent based on EUCAST breakpoint of .5. 1245 tobR: Isolate classification as tobramycin sensitive or resistent based on EUCAST breakpoint of 4. 1246 hypermutator: designation of known hypermutator isolates based on mutations in mutL or mutS. 1247 Persister: Hip versus Lop classification as 1 versus 0 using a cutoff of at least 3 spots growth in each 1248 biological replicate. 1249 Persister_CIP: persister class from ciprofloxacin screen alone. 1250 Persister_TOB: persister class from tobramycin screen alone. 1251 Persister_MD: persister class from ciprofloxacin screen and tobramycin screen. Hips must have 1252 scored as 3 or greater in both screens, representing a multidrug persister. 1253 Persister_ANY: persister class from ciprofloxacin screen and tobramycin screen. Hips must have 1254 scored as 3 or greater in at least one screen. 1255 1256 1257 1258 Dataset 2. Lineage-based mutation enrichment analysis for coding genes and noncoding regions 1259 for each dataset (CipHip, TobHip, MDHip, AnyHip). Mutated genes enriched in Hip versus Lop 1260 isolates as assessed from a convergent evolution perspective accounting for lineage adaptation. 1261 Lineage enrichment ratio was calculated by dividing lineage-based gene mutation enrichment within 1262 Hip variants by that within Lop variants for each gene. Top Hip-linked genes were selected via the 1263 following criteria: greater than 2 lineages presenting mutations in that gene in the Hip population and 1264 a lineage enrichment ratio greater than 2. 1265
1266
1267
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
40
Supplemental Information References 1268
1. Marvig RL, Sommer LM, Molin S, Johansen HK (2015) Convergent evolution and 1269
adaptation of Pseudomonas aeruginosa within patients with cystic fibrosis. Nat Genet 1270
47(1):57–64. 1271
2. Johansen HK, et al. (2004) Antibody response to Pseudomonas aeruginosa in cystic 1272
fibrosis patients: a marker of therapeutic success? - A 30-year cohort study of survival 1273
in Danish CF patients after onset of chronic P. aeruginosa lung infection. Pediatr 1274
Pulmonol 37(5):427–432. 1275
3. Høiby N, et al. (1977) Pseudomonas aeruginosa infection in cystic fibrosis - 1276
Diagnostic and prognostic significance of Pseudomonas aeruginosa precipitins 1277
determined by means of crossed immunoelectrophoresis. Scand J Respir Dis 58(2):65–1278
79. 1279
4. Haagensen JAJ, Verotta D, Huang L, Spormann A, Yang K (2015) New in vitro model 1280
to study the effect of human simulated antibiotic concentrations on bacterial biofilms. 1281
Antimicrob Agents Chemother 59(7):4074–4081. 1282
5. Frimodt-Møller J, et al. (2018) Mutations causing low level antibiotic resistance ensure 1283
bacterial survival in antibiotic-treated hosts. Sci Rep 8(1):1–13. 1284
6. Touw DJ, Knox AJ, Smyth A (2007) Population pharmacokinetics of tobramycin 1285
administered thrice daily and once daily in children and adults with cystic fibrosis. J 1286
Cyst Fibros 6(5):327–333. 1287
7. Heydorn A, et al. (2002) Statistical Analysis of Pseudomonas aeruginosa Biofilm 1288
Development: Impact of Mutations in Genes Involved in Twitching Motility, Cell-to-1289
Cell Signaling, and Stationary-Phase Sigma Factor Expression. Appl Environ 1290
Microbiol 68(4):2008–2017. 1291
8. Darija V, et al. (2013) Functional Analysis of spoT, relA and dksA Genes on 1292
Quinolone Tolerance in Pseudomonas aeruginosa under Nongrowing Condition. 1293
Microbiol Immunol 50(4):349–357. 1294
9. Murakami K, et al. (2005) Role for rpoS gene of Pseudomonas aeruginosa in antibiotic 1295
tolerance. FEMS Microbiol Lett 242(1):161–167. 1296
10. Nguyen D, et al. (2011) Active Starvation Responses Mediate Antibiotic Tolerance in 1297
Biofilms and Nutrient-Limited Bacteria. Science (80- ) 334(6058):982–986. 1298
11. De Groote VN, et al. (2009) Novel persistence genes in Pseudomonas aeruginosa 1299
identified by high-throughput screening. FEMS Microbiol Lett 297(1):73–79. 1300
12. Zhang L, et al. (2012) The catabolite repression control protein Crc plays a role in the 1301
development of antimicrobial-tolerant subpopulations in Pseudomonas aeruginosa 1302
biofilms. Microbiology 158(12):3014–3019. 1303
13. Cameron DR, Shan Y, Zalis EA, Isabella V, Lewis K (2018) A Genetic Determinant 1304
of Persister Cell Formation in Bacterial Pathogens. J Bacteriol 200(17):1–11. 1305
1306
1307
was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted December 17, 2019. ; https://doi.org/10.1101/561589doi: bioRxiv preprint
top related