Genetic determinants of antibiotic resistance and the evolution of trade-offs during adaptation in a single patient

The processes by which pathogens evolve within single hosts dictate the efficacy of treatment strategies designed to slow antibiotic resistance evolution and influence the population-wide resistance levels. The aim of this study is to describe the underlying genetic and phenotypic changes leading to antibiotic resistance within a single patient who died as resistance evolved to available antibiotics. We assess whether robust patterns of collateral sensitivity and response to combinations exist that might have been leveraged to improve therapy. Whole-genome sequencing was completed for nine isolates taken from this patient over 279 days of chronic infection with Enterobacter hormaechei, along with systematic measurements of changes in resistance against five of the most relevant drugs considered for treatment. The entirety of the genetic change is consistent with de novo mutations and plasmid loss events, without the acquisition of foreign genetic material via horizontal gene transfer. The isolates formed three genetically distinct lineages, with early evolutionary trajectories being supplanted by previously unobserved multi-step evolutionary trajectories. Importantly, no single isolate evolved resistance to all of the antibiotics considered for treatment against E. hormaechei (i.e., none was pan-resistant). Patterns of collateral sensitivity and response to combination therapy revealed contrasting patterns across this diversifying population. Translating antibiotic resistance management strategies from theoretical and laboratory data to clinical situations, such as this, may require managing diverse populations with unpredictable resistance trajectories.


24
The processes by which pathogens evolve within single hosts dictate the efficacy of treatment 25 strategies designed to slow antibiotic resistance evolution and influence the population-wide 26 resistance levels. The aim of this study is to describe the underlying genetic and phenotypic 27 changes leading to antibiotic resistance within a single patient who died as resistance evolved to 28 available antibiotics. We assess whether robust patterns of collateral sensitivity and response to 29 combinations exist that might have been leveraged to improve therapy. Whole-genome The effective treatment of bacterial infections has transformed modern medicine, but the ability of 45 bacteria to evolve resistance to essentially all antibiotics now threatens these gains (CDC, 2019).

46
For certain infections, the evolution of resistance within individual patients undergoing therapy can lead directly to treatment failure and worse patient outcomes (Folkesson et

57
However, the prevailing low levels of predictability of resistance evolution at the single patient 58 level are a major challenge to physicians on a daily basis.

60
A previous case study reported a patient suffering from a multispecies infection for more than 500 61 days that could not be resolved, resulting in the death of the patient as drug resistance progressed 62 (Woods and Read, 2015). The infection developed in two phases ( Figure 1A

91
We identified the genetic changes and population diversity leading to antibiotic resistance using 92 whole genome sequencing of the nine E. hormaechei isolates. We also evaluated evolutionary

106
We considered nine E. hormaechei isolates obtained during the second phase of infection within 107 a single patient that was previously described in detail ( Figure 1A; see (Woods and Read, 2015) 108 for a full account of the clinical case description). We obtained the isolates at times when the 109 patient visited the hospital and further analyzed them by using whole genome sequencing,   (Figure 2A and B). The two lineages including the early isolates were not detected 118 subsequently, so they were potentially lost during the course of treatment. Importantly, the use of 119 vancomycin and TMP/SMX during treatment was not targeted at E. hormaechei but were included 120 to reduce the risk of a MRSA re-emergence ( Fig 1B). Yet, these antibiotics, particularly the 121 TMP/SMX, could still have affected the E. hormaechei population structure. Indeed, the late 122 isolates revealed loss of TMP/SMX resistance associated with loss of the plasmid encoded dfrA12 123 and sulphonamide resistance sul1 genes ( Figure 1C

151
Of all mutations, we only found a single synonymous and seven intergenic variants while most of 152 the remaining variants were strongly disruptive ( Figure 2A). This pattern strongly suggests that 153 non-neutral evolution was predominant in this particular context.

155
Some of the identified mutations are associated with resistance against particular drugs that were 156 not used during treatment, suggesting the potential evolution of collateral resistance. For instance, 157 isolate E_134 was resistant to tigecycline ( Figure 1B and C, and 2A), an antibiotic to which the 158 patient was not exposed. This isolate had a frameshift variant in ramR, a TetR-like transcriptional

241
Similarly, the pairwise correlations across all nine isolates for all combinations of these five 243 antibiotics revealed both positive and negative correlations ( Figure 3A and B). However, none of 244 these associations were statistically significant (-0.5 < ρs < 0.78, P > 0.28; Figure 3B), and careful 245 inspection of the genetic changes reveal a more complex picture. First, sensitivity against 246 gentamicin remained below the initial levels of resistance of the earliest isolate, suggesting that 247 increases in resistance against any other drug did not co-select for resistance against this drug.   Figure 1 and 2). Evolution among the early isolates reshape the growth surface 261 area by expanding growth into higher concentration of both drugs, such that clones with increased 262 resistance to meropenem also have increased resistance to cefepime at all concentrations 263 (Isolates E_1, E_43, E_100 and E_134; Figure 4A). However, in the later five isolates, there 264 evolved a one-directional inhibition of antibiotic efficacy, whereby small amounts of meropenem 265 allow E. hormaechei to grow in much higher concentrations of cefepime. Among these later five 266 isolate the shape is similar, but there is a re-scaling along both the cefepime and meropenem 267 axes. Thus, the opportunity for selection inversion only exists between these two clades, but 268 within each clade there is a consistent pattern of collateral resistance. The patterns observed with 269 the remaining three drugs similarly reveal that the majority of the phenotypic evolution is stretching

303
The genetic evolution leading to resistance within this patient was complex. Resistance was a 304 multi-step evolutionary process, as we see at least three distinct lineages accumulating multiple 305 mutations. With the exception of ampD, the resistance was all in different genes, indicating that 306 the population was able to explore the fitness landscape by taking multiple steps in multiple

382
We recovered nine isolates from the patient during the phase when E. hormaechei was detected.

383
Labels, collection dates and culture sites for each isolate are given in Table 1

410
We identified genetic variants by mapping the short-read sequencing back to the closed genomes.

411
Finally, we evaluated the presence of foreign genetic elements via horizontal gene transfer by 412 aligning the assembled genomes of the first and last isolates using bwa (Li and Durbin, 2010).

413
Any unmapped reads were then extracted with samtools (Li et al., 2009) and de novo assembled 414 using spades (Prjibelski et al., 2020). We examined the obtained contigs and found for all isolates 415 a ~5kb phage phiX178 which is added as a control in the Illumina sequencing, there were typically 416 less than 9 contigs per sample, most of the contigs were less than 250bp long and they 417 predominantly consisted of homopolymers. This suggests that no foreign genetic material was 418 transferred horizontally among the isolates.

420
Changes in resistance

421
We determined the concentration inhibiting 90% of growth (IC90) using the broth microdilution 422 method detailed in the CLSI standard M07 for each of the E. hormaechei isolates. We added each 423 of the isolates to microdilution plates containing 2-fold dilutions of each of the antibiotics in 424 triplicate and incubated them at 36 ˚C for 21 h in LB media (Bertani, 1951). At this time, we 425 measured optical density (OD600) using a plate reader (FLUOstar Omega from BMG Labtech).

426
We then fitted Hill-Curves to the OD600 data to calculate the IC90 using the R platform (R Core To evaluate susceptibility of the isolates to antibiotic combinations we used checkerboard assays.

432
We diluted increasing concentrations of any two drugs along the X-and Y-axis of a 96-well 433 microtiter plate, leaving the last column as a blank (no added drug or bacteria). We evaluated all