bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
1 Genetic dominance governs the evolution and spread of mobile genetic
2 elements in bacteria
3 Jerónimo Rodríguez-Beltrán1*, Vidar Sørum2, Macarena Toll-Riera3, Carmen de la
4 Vega1, Rafael Peña-Miller4, Alvaro San Millán1*.
5 1 Department of Microbiology, Hospital Universitario Ramon y Cajal (IRYCIS) and
6 CIBER for Epidemiology and Public Health, Madrid, Spain.
7 2 Department of Pharmacy, UiT-The Arctic University of Norway, Tromsø, Norway.
8 3 Department of Evolutionary Biology and Environmental Studies, University of Zurich,
9 Switzerland.
10 4 Center for Genomic Sciences, Universidad Nacional Autonóma de México,
11 Cuernavaca, Mexico.
12 Correspondence: [email protected], [email protected]
13
14 Abstract
15 Mobile genetic elements (MGEs), such as plasmids, promote bacterial evolution
16 through horizontal gene transfer (HGT). However, the rules governing the repertoire
17 of traits encoded on MGEs remain unclear. In this study, we uncovered the central
18 role of genetic dominance shaping genetic cargo in MGEs, using antibiotic resistance
19 as a model system. MGEs are typically present in more than one copy per host
20 bacterium and, as a consequence, genetic dominance favors the fixation of dominant
21 mutations over recessive ones. Moreover, genetic dominance also determines the
22 phenotypic effects of horizontally acquired MGE-encoded genes, silencing recessive
23 alleles if the recipient bacterium already carries a wild-type copy of the gene. The
24 combination of these two effects governs the catalogue of genes encoded on MGEs,
25 dictating bacterial evolution through HGT. bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
26 Introduction
27 HGT between bacteria is largely mediated by specialized MGEs, such as plasmids
28 and bacteriophages, which provide an important source of genetic diversity and play
29 a fundamental role in bacterial ecology and evolution (1). The repertoire of accessory
30 genes encoded on MGEs and their ability to be phenotypically expressed in different
31 genetic backgrounds are key aspects of MGE-mediated evolution. There are several
32 factors known to impact horizontal gene transferability in bacteria, such as the level
33 of gene expression, the degree of protein connectivity, or the biochemical properties
34 of proteins (2-5), but the specific parameters that shape the repertoire of genes
35 encoded on MGEs remain largely unknown (6, 7).
36 Genetic dominance is the relationship between alleles of the same gene in which one
37 allele (dominant) masks the phenotypic contribution of a second allele (recessive). In
38 diploid or polyploid organisms, dominant alleles stem the establishment of new traits
39 encoded by recessive mutations (an effect known as Haldane’s sieve (8, 9)). Most
40 bacteria of human interest carry a single copy of their chromosome. In haploid
41 organisms like these, new alleles are able to produce a phenotype regardless of the
42 degree of genetic dominance of the underlying mutations. Therefore, the role of
43 genetic dominance in bacterial evolution has generally been overlooked. However,
44 the bacterial genome consists of more than the single chromosome; a myriad of
45 mobile genetic elements populate bacterial cells. Many MGEs, including plasmids
46 and filamentous phages, replicate independently of the bacterial chromosome and
47 are generally present at more than one copy per cell, with copy number ranging from
48 a handful to several hundred (10, 11). Extra-chromosomal MGEs thus produce an
49 island of local polyploidy in the bacterial genome (12, 13). Moreover, HGT in bacteria
50 mostly occurs between close relatives (14, 15), and genes encoded on mobile
51 elements can therefore create allelic redundancy with chromosomal genes. In light of
52 these evidences, genetic dominance should strongly affect both the emergence of bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
53 new mutations in MGE-encoded genes and the phenotypic effects of horizontally
54 transferred alleles.
55 Results and Discussion
56 To test whether genetic dominance determines the emergence of mutations in MGE-
57 encoded genes, we used a two-gene synthetic system that can confer tetracycline
58 resistance through dominant and recessive mutations (Figure 1). This construct
59 consists of a cI gene, encoding the bacteriophage λ CI repressor, in control of the
60 expression of a contiguous tetA gene, which encodes a tetracycline efflux pump (16).
61 This system provides tetracycline resistance when tetA transcription is derepressed.
62 Derepression can be achieved through mutations that either inactivate the cI gene or
63 disrupt the CI binding site upstream of tetA (Figure 1A). The repressor gene provides
64 a large target (714 bp), but we hypothesized that mutations inactivating cI would be
65 recessive, since in-trans copies of wild type CI would repress tetA. In contrast, the CI
66 binding site is a short target (166 bp), but mutations in this region are likely to be
67 dominant because they will lead to non-repressible, constitutive TetA production.
68 We produced two otherwise isogenic Escherichia coli MG1655 clones carrying the cI-
69 tetA system either as a single chromosomal copy (mono-copy treatment) or also
70 present on a pBAD plasmid with approximately 20 copies per cell (pCT, multi-copy
71 treatment) (Figure 1A). Fluctuation assays with both clones revealed a 4.84-fold
72 lower phenotypic tetracycline mutation rate in the multi-copy treatment clone, despite
−12 73 the higher cI-tetA copy number (Likelihood ratio test statistic 55.0, P< 10 , Figure
74 1B). To determine if this effect was due to differential access to dominant or
75 recessive mutations, we first analyzed the mutations in the cI-tetA system and
76 confirmed that they were located in different regions in each treatment (Wilcoxon
77 signed-rank test, W= 313, P≈ 10-06, Figure 1C, Table S1). Next, we generated
78 homozygous and heterozygous mutant clones in order to measure the coefficient of
79 dominance (h) of a subset of mutations [h ranges from 0 (completely recessive) to 1 bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
80 (completely dominant), Figure S1-S2, Table S2]. As predicted, h was low or
81 intermediate in tetracycline resistance mutations recovered from the mono-copy
82 treatment and high in mutations from the multi-copy treatment (ANOVA effect of
83 treatment; F=70.02, df=1, P≈3x10-5; Figure 1D).
84
85 Figure 1. Genetic dominance and gene copy number modulate phenotypic mutation
86 rates. (A) Schematic representation of plasmid pCT, the experimental model and the
87 cI-tetA system (note the isogenic nature of the clones apart from the dosage of cI-
88 tetA). (B) Phenotypic tetracycline resistance mutation rates in the different clones.
89 Rifampicin mutation rates were determined to test the equal underlying mutation rate
90 of clones. Error bars represent 84% confidence intervals. (C) Location and type of
91 tetracycline resistance mutations in the mono-copy and multi-copy treatments are
92 indicated. In the cI-tetA system diagram, blue shading denotes the cI coding region,
93 and purple shading denotes the CI binding site plus the cI-tetA intergenic region. (D)
94 Coefficient of dominance (h) of 10 mutations described in panel C. Bars represent
95 the median of 8 biological replicates, error bars represent the interquartile range. bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
96 Our results indicate that interplay between genetic dominance and gene copy
97 number determines the rate at which phenotypic mutants emerge in bacteria. The
98 increased gene dosage provided by MGEs improves the chances of a beneficial
99 mutation being acquired but simultaneously masks the phenotypic contribution of the
100 newly acquired allele if it is recessive. To study the general effect of this interplay on
101 the evolution of MGE-encoded genes, we developed a computational model based
102 on the classic fluctuation assay ((17), Figure 2A and S3-S4). This model allows us to
103 simulate the acquisition and segregation of mutations located in an extra-
104 chromosomal MGE, in this case a plasmid, with a given copy number. With this
105 information, we can explore the frequency of phenotypic mutants in the bacterial
106 population at any time point; this frequency will depend on the distribution of mutated
107 and wild-type alleles in each individual cell and on the coefficient of dominance of
108 those mutations (Figure S4). The simulations showed that the frequency of
109 phenotypic mutants increases with plasmid copy number for mutations of high
110 dominance but decreases for mutations of low dominance (Figure 2A).
111
112 Figure 2. Interplay between genetic dominance and gene copy number. (A) Results
113 of simulations analyzing the effect of plasmid copy number and genetic dominance of
114 a mutation on the emergence of phenotypic mutants. The chart shows fold-changes
115 in phenotypic mutation rate for a plasmid-carried gene at different copy numbers
116 compared with a chromosomal copy of the same gene (black line). The insert chart
117 shows the comparison of experimental results obtained for gyrA and folA (presented bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
118 in panels B and C; bars, 84% confidence intervals) with the prediction for a plasmid
119 of 20 copies. (B) Fold change of antibiotic resistance phenotypic mutation rates in E.
120 coli, comparing multi-copy and mono-copy treatments for gyrA, rpsL and folA. Note
121 that result for rpsL is an upper bound due to the absence of phenotypic mutants in
122 the multi-copy treatment (see methods). (C) Coefficient of dominance of gyrAD87N,
123 rpsLK43T and folAL28R. Bars represent the median of 8 biological replicates, error bars
124 represent the interquartile range.
125 The results of the simulation prompted us to test our hypothesis in a more realistic
126 and meaningful experimental system. To study the effect of genetic dominance and
127 gene copy number on the emergence of phenotypic mutants, we used bacterial
128 housekeeping essential genes known to confer antibiotic resistance through
129 mutations. The genes studied were gyrA (DNA gyrase subunit A), rpsL (30S
130 ribosomal protein S12), and folA (dihydrofolate reductase). Mutations in these genes,
131 which are present as single copies in the chromosome, confer resistance to
132 quinolone, aminoglycoside and trimethoprim antibiotics, respectively (18, 19). For
133 each gene, we produced two otherwise isogenic E. coli MG1655 clones with either
134 one copy of the gene (chromosomal) or multiple copies (chromosomal + plasmid,
135 Figure S5). We then calculated phenotypic mutation rates for each clone using
136 fluctuation assays with the appropriate antibiotics and sequenced the target genes in
137 the resistant clones to confirm the presence of mutations (Figure 2B, Table S1). The
138 frequency of gyrA and rpsL mutants in the multi-copy treatment was lower than in the
−12 139 mono-copy treatment (Likelihood ratio test statistic 49.48, P< 10 ), suggesting that
140 antibiotic resistance mutations in these genes are recessive, in line with previous
141 works in the field (20, 21). Conversely, the mutation rate for folA in the multi-copy
142 treatment increased 3-fold (Likelihood ratio test statistic 5.38, P= 0.04), suggesting
143 that this gene confers trimethoprim resistance through dominant mutations. To test
144 this possibility, we determined the coefficient of dominance of a common mutation in bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
145 each gene, confirming that mutations in GyrAD87N (h=0.043) and RpsLK43T (h=0) were
146 recessive, whereas mutation in FolAL28R (h=0.748) was dominant (Figure 2C, S1 and
147 S6, for more details on the mechanistic basis of genetic dominance of resistance
148 mutations see supplementary text).
149 While these results show that genetic dominance shapes the emergence of
150 mutations in MGE-encoded genes, they leave open the question of how genetic
151 dominance affects the horizontal transferability of genes in bacteria. We
152 hypothesized that phenotypic expression of recessive alleles encoded on MGEs will
153 be masked if the recipient bacterium possesses a chromosomal copy of the dominant
154 allele, effectively hindering transfer of the recessive allele. We tested this hypothesis
155 in an experimental assay of bacterial conjugation, using the low copy-number
156 mobilizable plasmid pSEVA121 (22). The plasmid was modified by independent
157 insertion of mutated, resistance-conferring gyrAD87N (recessive) and folAL28R
158 (dominant) alleles and their wild-type counterparts (Figure S7). We transferred these
159 plasmids from E. coli ß3914 to E. coli MG1655 (Figure 3A). After conjugation,
160 resistance conferred by folAL28R was readily expressed in the recipient cells, whereas
161 resistance conferred by gyrAD87N was masked by the resident wild-type allele,
162 preventing phenotypic expression of resistance in transconjugants (Figure 3A). A
163 simple model of the general effect of genetic dominance on the transferability of
164 housekeeping genes predicted that the presence of a dominant allele in the recipient
165 cell would reduce phenotypic expression of recessive alleles over a wide range of
166 plasmid copy numbers (Figure 3B). Crucially, this effect will be particularly marked for
167 conjugative plasmids, due to their low-copy number in the host cell (typically ranging
168 from 1 to 5). bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
169
170 Figure 3. Genetic dominance limits the phenotypic contribution of horizontally
171 transferred recessive alleles. (A) Pictures of a representative replicate of the
172 conjugation assays. Overnight cultures of spots inoculated from 10-fold dilutions of
173 conjugation mixes (100 to 10-4, from left to right) on plates selecting for
174 transconjugants. Selection on carbenicillin reveals the actual number of
175 transconjugants; selection on carbenicillin plus nalidixic acid or trimethoprim reveals
176 the number of transconjugants carrying gyrA and folA alleles and expressing the
177 resistant phenotype. (B) Antibiotic resistance level conferred by a plasmid-encoded
178 resistant allele in the recipient bacterium (when a wild-type copy of the gene is
179 present in the chromosome), assuming phenotypic resistance as the product of
180 plasmid copy number and the coefficient of dominance of the allele. Experimental
181 data are presented for gyrAD87N and folAL28R in plasmid pSEVA121.
182 Our results strongly suggest that genetic dominance shapes the repertoire of genes
183 present in MGEs in bacteria. To investigate this further, we analyzed the
184 Comprehensive Antibiotic Resistance Database (CARD), which includes detailed
185 information about antibiotic resistance genes from thousands of bacterial
186 chromosomes and plasmids (23). To extend our analysis to other MGEs, we also
187 examined databases for information on integrative and conjugative elements and
188 prophages. We predicted that housekeeping alleles conferring antibiotic resistance
189 and contained in MGEs would be more frequently dominant than recessive. We
190 investigated the genes in our experimental system plus two additional housekeeping bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
191 genes that, according to results from previous works, should confer antibiotic
192 resistance through dominant (folP, dihydropteroate synthase, sulfonamide
193 resistance) and recessive (rpoB, RNA polymerase subunit beta, rifampicin
194 resistance) mutations ((24-26), see supplementary text). Results confirmed that
195 mobile resistance-conferring alleles of folA and folP were ubiquitous in naturally
196 occurring MGEs, whereas resistant alleles of rpoB, gyrA and rpsL were almost never
197 present in MGEs (Figure 4). Alternative explanations to genetic dominance, such as
198 protein connectivity or the fitness effects produced by these genes in recipient
199 bacteria, could also explain the observed bias in gene distribution. However, analysis
200 of protein connectivity and direct measurements of growth rates did not offer a
201 satisfactory explanation for our results (Figure S8-S9).
202
203 Figure 4. Genetic dominance shapes the bacterial mobilome. Distribution of
204 antibiotic-resistance conferring alleles of rpoB, rpsL, gyrA, folA and folP (indicated
205 with an asterisk) in chromosomes and MGEs across bacteria. Note that recessive
206 alleles are absent from MGEs. For a more exhaustive analysis see Figure S10 and
207 supplementary text.
208 In light of these results, we propose that genetic dominance provides a general
209 explanation for the commonly observed divergence between chromosomal-mediated
210 and MGE-mediated antibiotic resistance genes to, for example, fluoroquinolones or
211 polymyxins (27, 28). Our results suggest that chromosomal singletons are free to
212 explore their mutational landscape in a host bacterium, but only dominant alleles are
213 able to provide a selectable phenotype, reach fixation and spread successfully on
214 MGEs. MGEs are therefore enriched in the few dominant resistant alleles of bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
215 chromosomal genes or in xenogeneic genes with no resident counterpart in the host
216 bacterium.
217 In summary, here we demonstrate that genetic dominance strongly influences
218 evolution through MGEs in bacteria, revealing a new layer of complexity in the forces
219 governing HGT. The impact of genetic dominance on MGEs is twofold, affecting both
220 (i) the emergence of new mutations in MGE-encoded genes and (ii) the phenotypic
221 effects of horizontally transferred genes. The first effect is driven by the polyploid
222 nature of extra-chromosomal MGEs, which induces the filtering of recessive
223 mutations through Haldane’s sieve. However, elevated gene copy number is not an
224 exclusive property of MGEs and can arise physiologically in haploid bacteria in
225 multiple circumstances. For example, gene copy number is increased by gene
226 duplication events and by transient polyploidy during fast bacterial growth (29, 30),
227 suggesting a potential broader influence of genetic dominance on bacterial evolution.
228 The second effect is determined by the dominance relationships that emerge when a
229 new allele arrives in a bacterium that already carries a copy of that gene. Given that
230 HGT in bacteria is strongly favored between close relatives (14, 15), genetic
231 redundancy of this type must be extremely common, underlining the importance of
232 genetic dominance in determining gene transferability.
233 Materials and Methods
234 Strains, plasmids and media
235 All the strains and plasmids used in this study are detailed in Table S3. Experiments
236 were performed using BBL Muller Hinton II agar or cation adjusted broth (Becton
237 Dickinson) unless specified. Antibiotics were supplied by Sigma-Aldrich and were
238 used at the following concentrations: carbenicillin 100 µg/ml, chloramphenicol 30
239 µg/ml, nalidixic acid 30 µg/ml, streptomycin 100 µg/ml, rifampicin 100 µg/ml,
240 trimethoprim 2 µg/ml and 16 µg/ml, and tetracycline 15 µg/ml. Cultures were routinely bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
241 grown at 37ºC with shaking (225 rpm). Plasmids were extracted using a commercial
242 miniprep kit (Macherey-Nagel) and were transformed into TSS competent cells (31).
243 Cloning and site directed mutagenesis
244 The gyrA, rpsL and folA wild-type alleles were PCR amplified using HS Taq Mix
245 (PCR biosystems) polymerase, using the primers listed in table S4 and MG1655
246 chromosomal DNA as template. The cI-tetA system was amplified similarly, but using
247 IBDS1 as a template (16). Purified PCRs were subsequently cloned into the pBAD
248 TOPO TA Expression kit (Thermo-Fisher) following manufacturer instructions. We
249 selected clones in which the insert was cloned in the opposite direction from that of
250 the PBAD promoter to ensure that expression is driven only from the cognate
251 promoter. Correct clones were identified by PCR and Sanger sequenced to give rise
252 to pBAD-gene and pCT plasmids (Table S3).
253 Site-directed mutagenesis was performed on the pBAD plasmids carrying wild-type
254 alleles to construct the gyrAD87N, rpsLK43T and folAL28R mutated alleles by using the Q5
255 site-directed mutagenesis kit (New England Biolabs) and the primers listed in table
256 S4. Correct cloning was assessed by PCR and Sanger sequencing. Next, we
257 generated pBDSM plasmid variants by Gibson assembling PCR amplified (Phusion
258 Hot Start II DNA Polymerase, Thermo-Fisher; see Table S4 for primers) mutated and
259 wild-type alleles into the pBGC backbone (Figure S5 and Table S3)
260 Tetracycline resistant mutations in the pCT plasmid isolated in the fluctuation assays
261 were purified by plasmid extraction and re-transformed into MG1655, selecting in
262 carbenicillin and tetracycline plates to isolate pCT mutated plasmids from the wild-
263 type version, that could be coexisting under heteroplasmy. Isolated pCT plasmids
264 and chromosomal mutants carrying cI-tetA mutated cassette were PCR amplified
265 (Phusion Hot Start II DNA Polymerase; Thermo-Fisher) and simultaneously Gibson
266 assembled (NEBuilder HiFi DNA Assembly kit; New England Biolabs) into the pBAD
267 and pBGC backbones using the primers listed in Table S4 to give rise to pBAD and bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
268 pBDSM plasmids carrying mutated alleles of the cI-tetA system (Figure S1 and Table
269 S3).
270 Fluctuation assays for mutation rate determination
271 Briefly, independent cultures containing carbenicillin were inoculated with ~10³ cells
272 and allowed to grow during 18 hours at 37ºC with shaking (225 rpm). The following
273 day, appropriate aliquots were plated in antibiotic-containing agar plates to select for
274 spontaneous mutants and in non-selective plates to determine the number of viable
275 bacteria. After incubation at 37ºC for 24 hours (48 hours for rifampicin and
276 streptomycin plates) colonies were enumerated. The genes suspected of carrying
277 mutations were PCR amplified and Sanger sequenced using mutant colonies coming
278 from independent cultures as template. Phenotypic mutation rates, defined as the
279 rate at which mutations contributing to the phenotype emerge, 84% confidence
280 intervals and likelihood ratio tests to asses statistical significance were then
281 calculated using the newton.LD.plating, confint.LD.plating and LRT.LD.plating
282 functions of the Rsalvador package for R (32). We plotted 84% confidence intervals
283 because it has been demonstrated that they better convey statistical significance
284 than 95% confidence intervals when dealing with mutation rate data (33).
285 Mono-copy and multi-copy strains of gyrA and rpsL presented the same antibiotic
286 resistance levels, indicating that target overexpression did not increase the level of
287 resistance for these genes (Table S2). However, and as previously reported (34),
288 folA overexpression led to a 8-fold increase in the trimethoprim resistance level in the
289 multi-copy treatment. To perform the mutation rate assay, we used 4 times the IC90 of
290 trimethoprim for each clone (2 and 16 µg/ml for the mono-copy and multi-copy
291 treatment, respectively). Sequencing of folA PCR products revealed mutations in all
292 the colonies tested from the mono-copy treatment, but only for approximately 10% of
293 the ones from the multi-copy treatment. This result could be due to the lack of fixation
294 of folA mutations, which can be maintained at a low frequency in heteroplasmy and bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
295 not be evident in the chromatogram, or due to the access to different resistance
296 mutations in this treatment. To solve this problem and recover only folA mutants, we
297 introduced an extra step in the fluctuation assays for the multi-copy treatment, where
298 we restreaked the colonies from 16 to 50 µg/ml trimethoprim plates. This step
299 allowed us to recover colonies carrying folA mutations only (avoiding false positives)
300 while at the same time allowing the recovery of colonies carrying folA mutations even
301 if the frequency of folA mutated alleles in heteroplasmy would have been too low to
302 initially grow on this concentration during a fluctuation assay (avoiding false
303 negatives) (Table S1).
304 For rpsL, we were not able to recover a single resistant colony in the fluctuation
305 assays for the multi-copy treatment (Table S1). Subsequent analyses revealed that
306 the coefficient of dominance of the most common rpsL mutation isolated in the
307 chromosome is 0. The complete recesiveness of rpsL streptomycin resistance
308 mutations explains the absence of resistant mutants in the multi-copy treatment
309 because, even if the plasmid-mediated copy of rpsL mutates and reaches fixation
310 prior to plating, the wild type chromosomal copy of the gene masks the resistance
311 phenotype. Therefore, for phenotypic resistant mutants to appear, both chromosomal
312 and plasmid copies of the gene should mutate (and reach fixation in the cell), which
313 is extremely unlikely. In Figure 2B we present the fold change of streptomycin
314 resistance phenotypic mutation rate using the mutation rate calculated for the mono-
315 copy treatment and the limit of detection of the mutation rate for the multi-copy
316 treatment.
317 Rifampicin mutation rates were performed to confirm that the underlying mutation
318 rate is equal for all clones used in this study (Figure 1 and S11).
319 Coefficient of Dominance
320 The coefficient of dominance (h) of mutations ranges from 0, completely recessive, to
321 1, completely dominant. To calculate h, we developed an experimental system based bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
322 on two compatible plasmids of similar copy number (n≈ 15-20) (35, 36), where we
323 cloned the genes of interest carrying the wild type or mutant alleles under
324 investigation (Figure S1). We then transformed E. coli MG1655 with the single
325 plasmids and the combinations of plasmids to produce the two possible
326 heterozygous clones. To construct the homozygous mutant clones for rpsLK43T and
327 gyrAD87N, we used the MG1655 background carrying the same mutation in the
328 chromosomal copy of the gene under study. We then performed antibiotic
329 susceptibility assays to determine the resistance phenotypes of the different clones
330 (measured as inhibitory concentration 90, IC90, Figure S2 and S6). Using the IC90
331 values we calculated h with the following formula:
332 h= (IC90aA - IC90AA) / (IC90aa - IC90AA)
333 where IC90AA, IC90aA and IC90aa are the IC90 of the homozygous wild-type,
334 heterozygous and homozygous mutant clones, respectively. Since there are two
335 different heterozygous clones (Figure S1), we calculated h as the median value of
336 four independent replicates of each heterozygous clone (n= 8, see Table S2).
337 IC90 values were obtained following (37), with some alterations. In short, strains were
338 streaked from freeze stock onto Mueller Hinton II agar plates and incubated over
339 night at 37ºC. Single colonies were picked and suspended in liquid Mueller Hinton II
340 broth and incubated at 37ºC and 225 rpm. Overnight cultures were diluted 1:10,000
341 into Mueller Hinton II broth to a final volume of 200 µl per well in microtiter plates.
342 The wells contained increasing concentrations of the appropriate antibiotic in 1.5-fold
343 increments. Plates were incubated at 37ºC for 22 hours and 225 rpm. After
344 incubation we ensured homogenous mixture of bacteria by orbital shaking for one
345 minute (548 rpm with 2 mm diameter) before reading of OD600 in a Synergy HTX
346 (BioTek) plate reader. IC90 is defined as ≥90 % inhibition of growth which was
347 calculated using the formula: 1 – [OD600 drug / ODcontrol], where the control is the no
348 antibiotic treatment. Wells only containing media were used for background bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
349 subtraction. Appropriate antibiotic for plasmid maintenance was included in the
350 media throughout the assay.
351 Conjugation assays
352 We developed an experimental model system to determine the frequencies of
353 conjugation of plasmids pSEVA121-gyrA, pSEVA121-gyrAD87N, pSEVA121-folA and
354 pSEVA121-folAL28R between E. coli strains (Figure S7). pSEVA121 is a vector
355 carrying the IncP-type RK2 (also called RP4) replicon, with a trfA replication initiator
356 protein gene plus the oriV origin of replication, and the oriT origin of transfer.
357 pSEVA121 displays therefore low copy-number in the host cell (ca. 4 copies/cell)
358 (38), representing a great model system to reproduce the features of a natural
359 conjugative plasmid. To mobilize pSEVA121 we used strain ß3914 as donor, which is
360 auxothropic for diaminopimelic acid and carries RP4 conjugation machinery inserted
361 in the chromosome. As a recipient we used E. coli MG1655 (Table S3).
362 Pre-cultures of donor and recipient strains were incubated overnight in 2 ml of
363 Mueller Hinton broth with the appropriate antibiotics at 37ºC and 225 rpm. Next day,
364 1:100 dilution of the ON cultures in 5 ml of Mueller Hinton broth were incubated until
365 the culture reached mid-exponential phase at 37ºC and 225 rpm (2.5 hours
366 approximately). Cultures were then centrifuged for 15 minutes at 1,500 G and the
367 pellets were re-suspended in 200 µl of fresh Mueller Hinton broth. 50 µl of donor and
368 10 µl of recipient suspensions were mixed and spotted on Mueller Hinton agar plates.
369 The conjugation mix was incubated for 18 hours at 37ºC and then suspended in 2 ml
370 of 0.9% NaCl sterile solution. Dilutions of this suspension were spotted as 5 µl drops
371 on Mueller Hinton agar plates containing antibiotics selecting for donor, recipient and
372 transconjugant cells. The total number of transconjugants was determined by
373 selecting on carbenicillin, while the number of transconjugants expressing the
374 resistance phenotype conferred by gyrA and folA alleles were selected on
375 carbenicillin plus nalidixic acid or trimethoprim, respectively. We performed 4 bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
376 independent biological replicates of each conjugation assay, and we show one
377 representative replicate in Figure 3A.
378 Growth curves
379 Growth curves were performed in a Synergy HTX (BioTek) plate reader. Briefly,
380 overnight cultures were diluted 1:1,000 in of Mueller Hinton broth supplemented with
381 carbenicillin, and 200 µl of the dilutions were transferred to a 96 multi-well plate
382 (Thermo Scientific). Plates were incubated at 37ºC with strong orbital shaking before
383 reading absorbance at 600 nm every 15 minutes. Six biological replicates were
384 included per strain.
385 Bioinformatic analysis
386 CARD prevalence data containing an analysis of 82 pathogens with more than
387 89,000 resistomes and more 175,000 antimicrobial resistance allele sequences was
388 obtained from the CARD website (23) (downloaded on 18/09/2019).
389 We downloaded all available bacterial integrative and conjugative elements (ICEs)
390 from ICEberg 2.0 database (39), amounting to a total of 806 ICEs from which 662
391 were conjugative type IV secretion system (T4SS)-type ICEs, 111 were
392 chromosome-borne integrative and mobilizable elements (IMEs), and 33 were cis-
393 mobilizable elements (CIMEs). A total of 741 perfect and strict hits from 267 ICEs
394 were identified using CARD RGI (version 4.2.2, parameters used: DNA sequence,
395 perfect and strict hits only, exclude nudge, high quality/coverage, and we used
396 DIAMOND to align the CDS against the CARD database).
397 We downloaded all complete bacterial genomes from NCBI, a total of 13,169
398 genomes (Assembly level: complete, downloaded on 15/02/2019). We used the
399 algorithm phiSpy to identify prophages in the bacterial chromosomes and we
400 predicted a total of 30,445 prophages in 7,376 genomes (40). To study the presence
401 of genes conferring antibiotic resistance in prophages, we extracted the CDS bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
402 overlapping with prophages and we predicted their resistome using the command
403 line tool CARD RGI as above.
404 We then used the antibiotic resistance data from prophages, ICEs, plasmids, and
405 chromosomes to calculate the prevalence of the following Antibiotic Resistance
406 Ontology (ARO) categories: “antibiotic resistant DNA topoisomerase subunit gyrA”
407 (ARO: 3000273), “antibiotic resistant rpsL” (ARO: 3003419), “sulfonamide resistant
408 dihydropteroate synthase” (ARO: 3000558), “antibiotic resistant rpoB”
409 (ARO:3003276), and "antibiotic resistant dihydrofolate reductase" (ARO:3003425).
410 Connectivity data for all protein families was downloaded from the STRING database
411 (41), and the number of high-confidence interactions (>700 STRING confidence) was
412 determined for each protein family.
413 Stochastic model
414 We assume that plasmids replicate randomly throughout the cell cycle until reaching
415 an upper limit determined by the plasmid copy number control
416 mechanism. Therefore, the probability of plasmid replication can be modelled as
417 1 − �!(�)/�, where �!(t) represents the number of copies of a plasmid of type � at
418 time � and � the maximum number of plasmids. We explicitly consider random
419 mutations occurring during replication events, so if � > 0 denotes the probability of a
420 mutation occurring in a plasmid, the per-cell mutation rate is � ∙ � . Stochastic
421 simulations of mutation and replication dynamics were performed using a Gillespie
422 algorithm implemented in Matlab with propensities determined from the distribution of
423 plasmid copies of each allele carried by the cell (code can be downloaded from
424 GitHub) (Figure S3).
425 We also consider that plasmids segregate randomly during cell division. The
426 probability that each plasmid is inherited to one of the daughter cells is a random
427 process that follows a binomial distribution. By implementing an agent-based bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
428 extension of the replication-mutation model coupled with the segregation dynamics,
429 we simulated the intracellular plasmid dynamics of individual cells in an exponentially
430 growing population (Figure S4). The frequency of phenotypic mutants observed in
431 the population at the end of the numerical experiment was estimated from the
432 fraction of mutated/wild-type alleles in each individual cell and the coefficient of
433 dominance of those mutations (ℎ). The results presented in this study were obtained
434 after 10! simulation runs in a range of plasmid copy numbers (Figure 2A). Finally, we
435 used Rsalvador package to estimate from this synthetic data the -fold change in
436 phenotypic mutation rate with respect to chromosomally-encoded genes.
437 Statistical analysis
438 All statistical tests were performed using R (v. 3.4.2).
439 References
440 1. J. Wiedenbeck, F. M. Cohan, Origins of bacterial diversity through horizontal genetic
441 transfer and adaptation to new ecological niches. FEMS Microbiol Rev 35, 957-976
442 (2011).
443 2. R. Sorek et al., Genome-wide experimental determination of barriers to horizontal
444 gene transfer. Science 318, 1449-1452 (2007).
445 3. R. Jain, M. C. Rivera, J. A. Lake, Horizontal gene transfer among genomes: the
446 complexity hypothesis. Proc Natl Acad Sci U S A 96, 3801-3806 (1999).
447 4. O. Cohen, U. Gophna, T. Pupko, The complexity hypothesis revisited: connectivity
448 rather than function constitutes a barrier to horizontal gene transfer. Mol Biol Evol 28,
449 1481-1489 (2011).
450 5. A. Porse, T. S. Schou, C. Munck, M. M. H. Ellabaan, M. O. A. Sommer, Biochemical
451 mechanisms determine the functional compatibility of heterologous genes. Nat
452 Commun 9, 522 (2018).
453 6. D. J. Rankin, E. P. Rocha, S. P. Brown, What traits are carried on mobile genetic
454 elements, and why? Heredity (Edinb) 106, 1-10 (2011). bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
455 7. T. Nogueira et al., Horizontal gene transfer of the secretome drives the evolution of
456 bacterial cooperation and virulence. Curr Biol 19, 1683-1691 (2009).
457 8. J. B. S. Haldane, A mathematical theory of natural and artificial selection, part V:
458 selection and mutation. Pros. Cambridge Philos. Soc. 23, 838–844 (1927).
459 9. D. A. Marad, S. W. Buskirk, G. I. Lang, Altered access to beneficial mutations slows
460 adaptation and biases fixed mutations in diploids. Nat Ecol Evol 2, 882-889 (2018).
461 10. G. del Solar, M. Espinosa, Plasmid copy number control: an ever-growing story. Mol
462 Microbiol 37, 492-500 (2000).
463 11. J. Rakonjac, N. J. Bennett, J. Spagnuolo, D. Gagic, M. Russel, Filamentous
464 bacteriophage: biology, phage display and nanotechnology applications. Curr Issues
465 Mol Biol 13, 51-76 (2011).
466 12. A. San Millan, J. A. Escudero, D. R. Gifford, D. Mazel, R. C. MacLean, Multicopy
467 plasmids potentiate the evolution of antibiotic resistance in bacteria. Nature Ecology
468 & Evolution 1, 0010 (2016).
469 13. J. Rodriguez-Beltran et al., Multicopy plasmids allow bacteria to escape from fitness
470 trade-offs during evolutionary innovation. Nat Ecol Evol 2, 873-881 (2018).
471 14. J. P. Gogarten, J. P. Townsend, Horizontal gene transfer, genome innovation and
472 evolution. Nat Rev Microbiol 3, 679-687 (2005).
473 15. O. Popa, T. Dagan, Trends and barriers to lateral gene transfer in prokaryotes. Curr
474 Opin Microbiol 14, 615-623 (2011).
475 16. I. Bjedov et al., Involvement of Escherichia coli DNA polymerase IV in tolerance of
476 cytotoxic alkylating DNA lesions in vivo. Genetics 176, 1431-1440 (2007).
477 17. S. E. Luria, M. Delbruck, Mutations of Bacteria from Virus Sensitivity to Virus
478 Resistance. Genetics 28, 491-511 (1943).
479 18. E. Toprak et al., Evolutionary paths to antibiotic resistance under dynamically
480 sustained drug selection. Nat Genet 44, 101-105 (2011).
481 19. N. Woodford, M. J. Ellington, The emergence of antibiotic resistance by mutation. Clin
482 Microbiol Infect 13, 5-18 (2007).
483 20. M. W. Hane, T. H. Wood, Escherichia coli K-12 mutants resistant to nalidixic acid:
484 genetic mapping and dominance studies. J Bacteriol 99, 238-241 (1969). bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
485 21. J. Lederberg, Streptomycin resistance; a genetically recessive mutation. J Bacteriol
486 61, 549-550 (1951).
487 22. E. Martinez-Garcia, T. Aparicio, A. Goni-Moreno, S. Fraile, V. de Lorenzo, SEVA 2.0:
488 an update of the Standard European Vector Architecture for de-/re-construction of
489 bacterial functionalities. Nucleic Acids Res 43, D1183-1189 (2015).
490 23. B. Jia et al., CARD 2017: expansion and model-centric curation of the comprehensive
491 antibiotic resistance database. Nucleic Acids Res 45, D566-D573 (2017).
492 24. Hayward, DNA blockade by rifampicin-inactivated Escherichia coli RNA polymerase,
493 and its amelioration by a specific mutation. Eur J Biochem 71, 19-24 (1976).
494 25. S. Austin, J. Scaife, A new method for selecting RNA polymerase mutants. J Mol Biol
495 49, 263-267 (1970).
496 26. G. Vedantam, G. G. Guay, N. E. Austria, S. Z. Doktor, B. P. Nichols, Characterization
497 of mutations contributing to sulfathiazole resistance in Escherichia coli. Antimicrob
498 Agents Chemother 42, 88-93 (1998).
499 27. L. S. Redgrave, S. B. Sutton, M. A. Webber, L. J. Piddock, Fluoroquinolone
500 resistance: mechanisms, impact on bacteria, and role in evolutionary success. Trends
501 Microbiol 22, 438-445 (2014).
502 28. Y. Y. Liu et al., Emergence of plasmid-mediated colistin resistance mechanism MCR-
503 1 in animals and human beings in China: a microbiological and molecular biological
504 study. Lancet Infect Dis 16, 161-168 (2016).
505 29. J. Nasvall, L. Sun, J. R. Roth, D. I. Andersson, Real-time evolution of new genes by
506 innovation, amplification, and divergence. Science 338, 384-387 (2012).
507 30. L. Sun et al., Effective polyploidy causes phenotypic delay and influences bacterial
508 evolvability. PLoS Biol 16, e2004644 (2018).
509 31. C. T. Chung, S. L. Niemela, R. H. Miller, One-step preparation of competent
510 Escherichia coli: transformation and storage of bacterial cells in the same solution.
511 Proc Natl Acad Sci U S A 86, 2172-2175 (1989).
512 32. Q. Zheng, rSalvador: An R Package for the Fluctuation Experiment. G3 (Bethesda) 7,
513 3849-3856 (2017). bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
514 33. Q. Zheng, Methods for comparing mutation rates using fluctuation assay data. Mutat
515 Res 777, 20-22 (2015).
516 34. A. C. Palmer, R. Kishony, Opposing effects of target overexpression reveal drug
517 mechanisms. Nat Commun 5, 4296 (2014).
518 35. C. Lee, J. Kim, S. G. Shin, S. Hwang, Absolute and relative QPCR quantification of
519 plasmid copy number in Escherichia coli. J Biotechnol 123, 273-280 (2006).
520 36. E. Hiszczynska-Sawicka, J. Kur, Effect of Escherichia coli IHF mutations on plasmid
521 p15A copy number. Plasmid 38, 174-179 (1997).
522 37. L. Imamovic, M. O. Sommer, Use of collateral sensitivity networks to design drug
523 cycling protocols that avoid resistance development. Sci Transl Med 5, 204ra132
524 (2013).
525 38. D. H. Figurski, D. R. Helinski, Replication of an origin-containing derivative of plasmid
526 RK2 dependent on a plasmid function provided in trans. Proc Natl Acad Sci U S A 76,
527 1648-1652 (1979).
528 39. M. Liu et al., ICEberg 2.0: an updated database of bacterial integrative and
529 conjugative elements. Nucleic Acids Res 47, D660-D665 (2019).
530 40. S. Akhter, R. K. Aziz, R. A. Edwards, PhiSpy: a novel algorithm for finding prophages
531 in bacterial genomes that combines similarity- and composition-based strategies.
532 Nucleic Acids Res 40, e126 (2012).
533 41. D. Szklarczyk et al., STRING v11: protein-protein association networks with
534 increased coverage, supporting functional discovery in genome-wide experimental
535 datasets. Nucleic Acids Res 47, D607-D613 (2019).
536 42. R. T. Cirz et al., Inhibition of mutation and combating the evolution of antibiotic
537 resistance. PLoS Biol 3, e176 (2005).
538 43. R. Nagel, A. Chan, Mistranslation and genetic variability: the effect of streptomycin.
539 Mutat Res 601, 162-170 (2006).
540 44. M. Watson, J. W. Liu, D. Ollis, Directed evolution of trimethoprim resistance in
541 Escherichia coli. FEBS J 274, 2661-2671 (2007).
542 45. G. M. Brown, The biosynthesis of folic acid. II. Inhibition by sulfonamides. J Biol
543 Chem 237, 536-540 (1962). bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
544 46. S. Roland, R. Ferone, R. J. Harvey, V. L. Styles, R. W. Morrison, The characteristics
545 and significance of sulfonamides as substrates for Escherichia coli dihydropteroate
546 synthase. J Biol Chem 254, 10337-10345 (1979).
547 47. G. Swedberg, S. Castensson, O. Skold, Characterization of mutationally altered
548 dihydropteroate synthase and its ability to form a sulfonamide-containing
549 dihydrofolate analog. J Bacteriol 137, 129-136 (1979).
550 48. G. Swedberg, O. Skold, Characterization of different plasmid-borne dihydropteroate
551 synthases mediating bacterial resistance to sulfonamides. J Bacteriol 142, 1-7 (1980).
552 49. J. Ilhan et al., Segregational Drift and the Interplay between Plasmid Copy Number
553 and Evolvability. Mol Biol Evol 36, 472-486 (2019).
554 Acknowledgments: We would like to thank R. Craig MacLean and José R. Penadés
555 for helpful discussion. We are grateful to Ivan Matic, Jesús Blázquez, Jose A
556 Escudero, Didier Mazel and the SEVA collection for generous gifts of strains and
557 plasmids. Funding: This work was supported by the European Research Council
558 under the European Union’s Horizon 2020 research and innovation programme (ERC
559 grant agreement no. 757440-PLASREVOLUTION) and by the Instituto de Salud
560 Carlos III (grant PI16-00860). ASM is supported by a Miguel Servet Fellowship
561 (MS15-00012). JRB is a recipient of a Juan de la Cierva Fellowship (FJCI-2016-
562 30019). MTR acknowledges support from the Swiss National Science Foundation
563 (Ambizione grant, PZ00P3_161545). RPM was funded by CONACYT Ciencia Básica
564 (grant A1-S-32164). VS is supported by Northern Norway Regional Health Authority
565 (HNF1494-19) and The National Graduate School in Infection Biology and
566 Antimicrobials (NFR grant no. 249062). Author contribution: Conceptualization,
567 JRB, ASM. Formal analysis, JRB, VS, MTR, RPM. Funding acquisition, ASM.
568 Investigation, JRB, VS, CdlV. Project administration and supervision, ASM. Software,
569 JRB, MTR, RPM. Writing, JRB, ASM. Competing interests: Authors declare no
570 competing interests. Data and materials availability: All data is available in the
571 main text or the supplementary materials. bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
572 Supplementary Materials
573 Supplementary text
574 Mechanistic basis of genetic dominance of antibiotic resistance mutations analyzed
575 in this work.
576 We experimentally tested the degree of genetic dominance of mutations conferring
577 antibiotic resistance in three housekeeping genes: gyrA (GyrAD87N), rpsL (RpsLK43T)
L28R 578 and folA (FolA ). Mutations in gyrA (DNA gyrase subunit A) and rpsL (30S
579 ribosomal protein S12) were recessive, while mutation in folA (dihydrofolate
580 reductase) was dominant. The recessive nature of mutations in gyrA and rpsL can be
581 explained by the toxic effect produced by the combination of antibiotic and wild type
582 alleles. In the case of gyrA, when the quinolone antibiotic binds to the
583 gyrase, double-strand DNA breaks occur, leading to cell death (42). In the case of
584 streptomycin and rpsL, the binding of the antibiotic to RpsL in the ribosome leads to
585 mistranslation events producing a toxic effect in the cell (43). Therefore, even if
586 resistance mutations reduce or avoid the binding of the antibiotic to the mutated
587 target, if there are wild type enzymes present in the cell the antibiotic will still produce
588 a toxic effect, explaining the recessive nature of these mutations.
589 Our experimental results showed that mutations in FolA were strongly dominant.
590 FolA is a key enzyme in folate metabolism. Specifically, it catalyzes the reduction of
591 dihydrofolate to tetrahydrofolate via hydride transfer from NADPH to C6 of the
592 pteridine ring. Trimethoprim binds to FolA and inhibits its activity, blocking folate
593 metabolism and producing a bacteriostatic effect. Mutations in FolA confer
594 trimethoprim resistance through reduction of the binding affinity of FolA for the drug
595 and/or by increasing the activity of this enzyme (both by increasing expression or by
596 specific mutations that increase enzymatic activity (18, 44)). However, in this case,
597 the binding between FolA and trimethoprim does not produce a toxic effect per se. bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
598 Therefore, in the presence of trimethoprim a cell carrying an antibiotic resistant
599 version of the enzyme will be able to grown even if there are wild-type copies of FolA,
600 explaining the genetic dominance of the mutation.
601 We included two more housekeeping genes known to confer resistance to antibiotics
602 through mutations in our bioinformatic analysis: rpoB (RNA polymerase subunit beta,
603 rifampicin resistance) and folP (dihydropteroate synthase, sulfonamides resistance,
604 Figure 4). We chose these two genes because the genetic dominance of their
605 resistance mutations could be inferred from previous works. For rpoB, the complexes
606 formed between rifampicin and the wild type RNA polymerase block DNA
607 transcription, even if there is a resistant version of RpoB (24). Therefore, rifampicin-
608 resistance mutations in rpoB are recessive (24, 25).
609 FolP is a dihydropteroate synthase (DPHS) in the folate synthesis pathway and it
610 catalyzes the condensation of para-aminobenzoic acid (PABA) with pteridine
611 diphosphate (PDP) to form dihydropteroate (DHP). Sulfonamides compete with
612 PABA for binding to the enzyme DHPS, resulting in their covalent attachment to PDP
613 in place of PABA (45). The resulting product, dihydropterin–sulfonamide, is not toxic
614 to E. coli (46), and the bacteriostatic activity of these antibiotics comes from the
615 depletion of the essential metabolite PDP. The fact that dihydropterin–sulfonamide is
616 not toxic suggests that folP mutations can be dominant (in contrast to rpsL,
617 gyrA and rpoB mutations). Palmer and Kishony recently described that the fraction of
618 PDP that is converted to DHP (the correct product) or to dihydropterin–sulfonamide
619 (the incorrect one) depends on the ratio of drug to drug-competing substrate
620 (sulfonamide to PABA ratio), weighted by binding affinity of the DHPS enzyme to
621 each substrate (31). Resistance mutations in folP decrease the binding affinity to
622 sulfonamides drastically (47, 48). As a consequence, in this scenario the degree of
623 dominance of mutations is going to depend on the rate of synthesis of DHP by the
624 mutant allele, compared to the rate of synthesis of dihydropterin–sulfonamide by the bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
625 wild type allele (in other words what fraction of PDP is used in the correct or the
626 incorrect reaction). The rate of synthesis of DHP by the mutant allele depends, in
627 turn, on its binding affinity to PABA. Therefore, the higher the binding affinity of the
628 mutant FolP to PABA the higher coefficient of dominance of the mutation will be. In
629 summary, DHPS resistant alleles will range from mildly dominant (point mutations
630 in folP with reduced affinity to PABA (47)) to strongly dominant (mobile resistant
631 alleles of DHPS, such as sul1 and sul2, with high binding affinity to PABA (48)).
632 bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
633 Supplementary Figures
634 Figure S1. Model system developed to calculate coefficient of dominance of
635 mutations.
636
637
638 Schematic representation of plasmids pBAD and pBDSM (Table S1). The gene or
639 genes under study were cloned into these plasmids using Gibson assembly. We
640 cloned both wild type and mutant alleles of the gene or genes under study in both
641 plasmids. We constructed E. coli MG1655 strains carrying plasmids creating
642 homozygous wild type and mutant clones plus heterozygous clones (2 different
643 heterozygous combinations per gene or genes under study). To construct the
644 homozygous mutant clones for recessive alleles, we used the MG1655 background
645 carrying the same mutation in the chromosomal copy of the gene under study. bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
646 Figure S2. Tetracycline resistance levels of homozygous and heterozygous clones
647 constructed to measure the coefficient of dominance of mutations.
648
649
650 Tetracycline resistance phenotypes of the different clones constructed to measure
651 the coefficient of dominance of tetracycline resistance mutations detailed in Figure
652 1C. The tetracycline inhibitory concentration 90 (IC90, in mg/L) of the homozygous
653 mutant clones (Mut), heterozygous mutant clones (HT) and homozygous wild type
654 clones (WT) are represented by boxes. The line inside the box marks the median.
655 The upper and lower hinges correspond to the 25th and 75th percentiles. The letters
656 in the panels correspond to the mutations indicated in Figure 1C. Note that for the
657 mutations isolated in the multicopy treatment the resistance level of HT clones is
658 similar to that of the Mut clones (dominant mutations). bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
659 Figure S3. Numerical simulations of the stochastic model of plasmid dynamics in the
660 absence of selection.
20 WT MUT
10 PCN
0 0 1 2 3 4 5 6 7 8 9 10 661 Time (divisions)
662 Replication events increase the number of copies of plasmids carried in a cell until
663 the cell divides and plasmids are segregated randomly between both daughter cells.
664 In this example, we consider that a mutation occurred in a copy of a WT plasmid at
665 t=0 (multiple stochastic realizations of this experiment are shown with light colors):
666 blue lines correspond to the number of copies of the WT plasmid and red lines the
667 number of mutant plasmids as a function of time. Note that in some cases the mutant
668 allele is lost through segregational drift (49), while in others (for instance the
669 simulation highlighted with the thick lines) the stochastic nature of the plasmid
670 dynamics produces a gradual increase in mutant frequency. bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
671 Figure S4. In silico fluctuation assays.
672
A Plasmid fraction B Plasmid fraction =2 0 .5 100%1 =15 00 .5 100%1 0.0625 0.33333
Mutant alleles Mutant alleles 0.26667 0.66667 0 0 0.33333 0.33333 0.5 1 0 0 0 0 0.4375 0 0.125 0.6875 0.33333 1 0.5 0 1 1 0 0 0.25 0 0 0 0.4 0.125 0.3125 0 0.66667 0.25 0.73333 0.33333 0.92857 1 0.53333 0 1 0 1 1 0 0 1 0 0 1 0 1 0 1 0 0.4 1 0 0.25 0 0.26667 0.1875 0.5625 0.25 0 0 0.066667 0.3125 1 0.9375 0.66667 0.6875 0.46667 0.625 0.26667 0.6875 1 1 0 0 0.33333 0 0 0 0 0 0 0 1 0.5 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 1 1 0 0.6 1 1 0.5 0 1 1 1 0.25 1 0.25
1 1
0.5 0.5 Dominance Dominance 0 0
0 0.5 1 0 0.5 1 673 Mutant frequency Mutant frequency
Supplementary Figure S5. In silico fluctuation assay in a population of cells with (A) maximum two plasmids 674 per-cell and (B) 15 plasmids The tree illustrates numerical simulations of the segregation-replication dynamics In silico fluctuationmodel in an exponentially assays growing ofpopulation populations of cells in the absence of cells of selection. with We considera maximum that mutations of two plasmids can occur at any point during plasmid replication, so we track the fraction of mutant alleles in each cell (WT alleles represented in grey, with the fraction of mutant plasmids in each cell is represented with a gradient of red). 675 (left) and At 15the end plasmids of the computational per experiment-cell (right). (in the case illustrated Trees here illustrate 5 generations, but numerical in the simulations simulations of the discussed in the manuscript 25 generations), we simulate a fluctuation assay in order to estimate the phenotypic mutant frequency from the fraction of mutated/wild-type alleles in each individual and the degree of dominance of the mutation. Red circles represent cells that survived the fluctuation assay and grey cells those that did not. 676 segregationNote- howreplication the mutant frequency dynamics decreases with modelthe degree of i ndominance an exponentiallyof the mutation (bars). growing population of
677 cells in the absence of selection. We consider that mutations can occur randomly
678 during plasmid replication, so we track the fraction of mutant alleles in each cell as a
679 function of time (WT alleles represented in grey, with the fraction of mutant plasmids
680 in each cell is represented with a gradient of red). At the end of the computational
681 experiment, we simulate a fluctuation assay to estimate the phenotypic mutant
682 frequency from the fraction of mutated/wild-type alleles in each individual and the
683 degree of dominance of the mutation. Red circles represent cells that survived the
684 fluctuation assay (bottom). In the case illustrated here, simulations are performed for
685 five generations, but in the synthetic data presented in the manuscript, the
686 experiment was run for 25 generations. Note how, in both cases, the frequency of
687 phenotypic mutants (survivors of the fluctuation assay) decreases as the degree of
688 dominance of the mutation decreases (grey bars). bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
689 Figure S5. Construction of mono-copy and multi-copy experimental systems of
690 housekeeping genes.
691
692 (A) Schematic representation of plasmid pBAD carrying a housekeeping gene. (B)
693 Construction of experimental system using E. coli MG1655 as recipient strain. To
694 obtain completely isogenic backgrounds we transformed the empty pBAD vector in
695 the mono-copy treatment strain. pBAD presents approximately 20 copies in the host
696 bacterial cell. bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
697 Figure S6. Antibiotic resistance levels of homozygous and heterozygous clones
698 constructed to measure the coefficient of dominance of mutations.
699
700
701 Antibiotic resistance phenotypes of the clones constructed to measure the coefficient
702 of dominance of folAL28R (trimethoprim), gyrAD87N (nalidixic acid) and rpsLK43T
703 (streptomycin) mutations. The inhibitory concentration 90 (IC90, in mg/L) of the
704 homozygous mutant clones (Mut), heterozygous mutant clones (HT) and
705 homozygous wild type clones (WT) are represented by boxes. The line inside the box
706 marks the median. The upper and lower hinges correspond to the 25th and 75th
707 percentiles. Note that for folAL28R mutation the resistance level of HT clones is similar
708 to that of the Mut clones (dominant mutations), while for gyrAD87N and rpsLK43T
709 mutations the resistance level of HT clones is similar to that of the WT clones
710 (recessive mutations). bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
711 Figure S7. Construction of a mobilizable plasmid carrying gyrA or folA genes.
712
713
714
715 Schematic representation of mobilizable plasmid pSEVA121. We cloned two different
716 alleles of gyrA (WT and D87N) and folA (WT and L28R) into pSEVA121. pSEVA121
717 presents approximately 4 copies in the host bacterial cell. bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
718 Figure S8. Growth curves of MG1655 with plasmid- and chromosomal-encoded wild
719 type and mutated alleles of the genes under study.
720
721 Growth curves measuring change in bacterial optical density at 600 nm (OD600, log
722 scale) over time. Solid lines present the average of 6 biological replicates and the
723 shaded region surrounding the curve represents standard deviation. Wild type E. coli
724 MG1655 strain carrying an empty pBAD plasmid (red), MG1655 carrying the
725 chromosomally encoded resistance-conferring mutant allele and an empty pBAD
726 plasmid (blue), MG1655 carrying a pBAD plasmid encoding the wild type allele of the
727 gene of interest (purple) and MG1655 carrying pBAD plasmid encoding a resistance-
728 conferring mutant allele of the gene of interest (green). Note that none of the
729 recessive resistance-conferring alleles (gyrAD87N and rpsLK43T) present severe growth
730 defects compared to the chromosomal mutants. Given that the frequency of wild-type
731 strains carrying chromosomal mutant alleles is high in nature (Figure 4), our results
732 suggest that the absence of mobile versions of these alleles is not due to the fitness
733 costs imposed by the MGE-encoded alleles. bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
734 Figure S9. Connectivity of the proteins under study.
735
736
737 Number of protein-protein interactions of the proteins under study according to
738 STRING database (41). RpsL and RpoB are involved a high number of protein-
739 protein interactions, which could constitute a barrier to their horizontal gene transfer
740 (4). However, gyrA, which also confers antibiotic resistance through recessive
741 mutations, encodes a protein with similar number of interactions as those conferring
742 resistance through dominant mutations (e.g. folA), which are commonly present in
743 MGEs (Figure 4). bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
744 Figure S10. Distribution of antibiotic resistance genes in plasmids and chromosomes.
745
746 Frequency of genes belonging to the different categories of the Antibiotic Resistance
747 Ontology group “mechanism of antibiotic resistance” in plasmids and chromosomes.
748 We tried to explore the potential effect of genetic dominance on the general
749 distribution of all resistance genes on plasmid/chromosomes according to their
750 Antibiotic Resistance Ontology (ARO, grouped by “mechanism of antibiotic
751 resistance”, data obtained from CARD). These groups include genes that are not
752 necessary present on the chromosomes of recipient bacteria; therefore genetic
753 dominance relationships between alleles may not impact their transferability. Actually,
754 we hypothesized that those horizontally transferred genes with no pre-existing alleles
755 on the recipient cells would be frequent in MGEs, since their phenotypic effect would
756 not be masked regardless of the dominance of the allele. In line with this idea, we
757 observed that the “antibiotic inactivation” category, which is mainly comprised of
758 dedicated antibiotic degrading enzymes of xenogeneic origin, is extremely prevalent
759 on plasmids. Conversely, categories including housekeeping genes vary in frequency
760 on plasmids. Interestingly, a category enriched in housekeeping genes carrying loss
761 of function (i.e. recessive) mutations is absent on plasmids. This is the category
762 “reduced permeability to antibiotic”, which includes mostly inactivated outer
763 membrane porins. bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
764 Figure S11. Rifampicin resistance mutation rates of the clones under study.
765
766 Rifampicin resistance phenotypic mutation rates were performed to confirm the
767 normo-mutator status of the clones used in this study. There were no significant
768 differences in rifampicin mutation rates among strains (Likelihood Ratio Test P>0.13
769 in all cases). bioRxiv preprint doi: https://doi.org/10.1101/863472; this version posted December 3, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
770 Supplementary Tables
771 Table S1: mutations described in the fluctuation assays.
772 Spreadsheet containing the mutations found after sequencing one clone of each
773 biological replicate of the fluctuation assays, either from the mono-copy or the multi-
774 copy treatments (Figures 1C, 1D and 2C).
775 Table S2: Antibiotic susceptibility tests (IC90 results).
776 Spreadsheet containing the results of the IC90 assays used to calculate the
777 coefficient of dominance in Figures 1D and 2C.
778 Table S3: Strains and plasmids used in this study.
779 Spreadsheet containing a short description of the strains and plasmids used in this
780 study.
781 Table S4: Primers used in this study.
782 Spreadsheet containing the information about the oligonucleotides used in this study
783 including their sequence, and a short description of their objective.