Full Title: Growth substrate may influence biofilm susceptibility to antibiotics

The CDC biofilm reactor is a robust culture system with high reproducibility in which biofilms can be grown for a wide variety of analyses. Multiple material types are available as growth substrates, yet data from biofilms grown on biologically relevant materials is scarce, particularly for antibiotic efficacy against differentially supported biofilms. In this study, CDC reactor holders were modified to allow growth of biofilms on collagen, a biologically relevant substrate. Susceptibility to multiple antibiotics was compared between biofilms of varying species grown on collagen versus standard polycarbonate coupons. Data indicated that in 13/18 instances, biofilms on polycarbonate were more susceptible to antibiotics than those on collagen, suggesting that when grown on a complex substrate, biofilms may be more tolerant to antibiotics. These outcomes may influence the translatability of antibiotic susceptibility profiles that have been collected for biofilms on hard plastic materials. Data may also help to advance information on antibiotic susceptibility testing of biofilms grown on biologically relevant materials for future in vitro and in vivo applications.


Abstract Introduction 65
The CDC biofilm reactor has been validated as a robust and repeatable reactor system for 66 elucidating various aspects of biofilm physiology, morphology, growth dynamics and antibiotic 67 susceptibility profiles ( and properties, the design of the CDC reactor system is such that coupons can be placed in 70 holding rods ( Figure 1A). This allows for exposure to shear force and renewable nutrients that 71 optimize biofilm formation by simulating natural environments such as rivers, streams, oral 72 cavities, biomedical device surfaces or industrial systems (Costerton et al. 1978, Costerton  reactor was developed such that it could hold coupons made of highly porous, bioabsorbable 83 collagen ( Figure 1B). The rationale for doing so was two-fold. 84 First, there is currently a paucity of data on antibiotic efficacy against biofilms that are 85 grown on biologically relevant materials. Collagen constitutes 75% of the dry weight of human 86 skin, and is a major component of extracellular matrix and multiple tissue types that have 87 Implant Logistics (La Crosse, WI). All antibiotics used were purchased from Sigma Aldrich (St. 111 Louis, MI) or TCI America (Portland, OR). Antibiotics for each bacterium were chosen based on 112 common clinical use. E-Test strips for MIC testing of amoxicillin and erythromycin were 113 purchased from Biomérieux (Durham, NC). 114

Bacterial Isolates 115
Bacterial isolates were chosen because of their use in various applications including 116 standards assays and animal models, their known pathogenicity as well as their ability to form 117 Standard polypropylene CDC biofilm reactor holders were used to hold polycarbonate 130 coupons ( Figure 1A). Custom holders were made for the collagen plug coupons ( Figure 1B). To 131 do so, four holes of 8.5 mm diameter each were drilled in the lower portion of a blank 132 polypropylene holder ( Figure 1B). This diameter was smaller than the diameter of standard 133 reactor coupons (12.7 mm). The size difference was to ensure a tight fit of the collagen in the 134 holders. 135 To make the collagen plugs, medical grade Collaform collagen was purchased as 1 cm x 136 2 cm plugs. The collagen was aseptically removed from packaging and cut into coupons with a 137 sterile blade. Coupon size was 1 cm diameter x 0.3 cm height. Coupons were aseptically loaded 138 into modified reactor arms ( Figure 1B) that had already been autoclaved. The collagen coupons 139 remained securely in place when exposed to the shear forces in the reactor. To obtain a baseline of biofilm/coupon, two holding arms were aseptically removed from 152 a reactor and rinsed in 1x PBS. An n=6 coupons were individually placed in a test tube that 153 contained 2 mL CAMHB, vortexed for 1 min, sonicated for 10 min, then vortexed a final time 154 for approximately 10 sec. A 10-fold dilution series was used to plate bacteria in duplicate on 155 tryptic soy agar (TSA). Agar plates were incubated overnight at 37° C. CFU were counted and 156 used to calculate the CFU/coupon. 157

Antibiotic Treatment 158
Prior to assessing the antibiotic susceptibility of biofilms, the minimum inhibitory 159 concentration (MIC) of each antibiotic was determined against each bacterial strain (see Table  160 2). For all antibiotics, except amoxicillin and erythromycin against Streptococcus mutans, a 161 modified protocol of the Clinical and Laboratory Standards Institute (CLSI) guideline M100 was 162 used. In short, using a fresh overnight culture of bacteria, a 0.5 McFarland standard was made in 163 PBS using a nephelometer. The stock solution was diluted in PBS 1:100 to achieve a 164 concentration of ~7.5 x 10 5 CFU/mL. A 96-well plate was set up such that a final volume of 100 165 µL was present in each well. Column 1 served as the negative control of growth (antibiotic only 166 without bacteria added) and Column 11 served as the positive control of growth (bacteria only, 167 no antibiotic). 168 To accomplish this, 100 µL of CAMHB that contained antibiotic only (64 µg/mL) was 169 pipetted into each well of column 1 to serve as the negative control. Into columns 2-11, 50 µL of 170 CAMHB were added to each well. Subsequently, 50 µL of CAMHB that contained a 171 concentration of 256µg/mL antibiotic were added to each well of column 2 using a multi-channel 172 pipet. The solution was mixed, and then 50 µL were removed and added to wells of column 3. 173 This 1:2 dilution process was continued until column 10 and resulted in a range of antibiotic 174 testing from 64 µg/mL to 0.0625 µg/mL. Lastly, into each well of columns 2-11, 50 µL of the 175 bacterial solution were added, with column 11 serving as the positive control. The 96-well plate 176 was covered with adhesive film and incubated 24 h at 37° C. The concentration of antibiotic that 177 inhibited pellet formation or turbidity at the 24 h time point was considered the MIC. Once the 178 MIC was determined, biofilm analysis was performed. 179 To determine the MIC of amoxicillin and erythromycin against S. mutans, an E-Test was 180 used. In short, S. mutans was cultured on BHI agar and incubated for 48 hours under 5% CO2. A 181 1.0 McFarland standard solution was made in PBS resulting in 2.8 x 10 8 CFU/mL solution. This 182 was used to make lawns of bacteria on BHI agar by stroking back and forth with an inoculated 183 Q-tip in three directions. After drying, E-Test strips for each compound were laid (n=2 per plate, 184 2 plates per compound). The plates were incubated for 24 hours under 5% CO2 and then analyzed 185 per manufacturer's instructions. 186 Following a reactor run, n=5 polycarbonate and n=6 collagen coupons were aseptically 187 removed from a reactor, rinsed in 1x PBS, and each coupon was individually placed in a test tube 188 that contained 2 mL solution of antibiotic in CAMHB. Each of the antibiotics were tested 189 initially at 50, 100, and 200 µg/mL concentrations. In some instances, data spread needed to be 190 resolved so additional concentrations were tested down to 25 or up to 400, or 600 µg/mL. 191 Coupons were incubated for 24 h after which time each was quantified as described above. 192 Imaging 193 194 Scanning electron microscopy (SEM) imaging was used to directly observe biofilm 195 morphology and confirm growth on both coupon types. To perform SEM imaging, biofilms were 196 grown following the same protocol as above. Notably, coupons used for SEM imaging to assess 197 morphology were not the same coupons used for the antibiotic testing. This was not possible 198

Scanning Electron Microscopy
given the need to fix and process biofilms, which inherently leads to cell death and would have 199 skewed the antibiotics or baseline quantification data. 200 To process for SEM imaging, a holder was aseptically removed from a reactor, rinsed one 201 time in 1x PBS, and each coupon was individually submerged in modified Karnovsky's fixative 202 (2.5% glutaraldehyde and 4% formaldehyde in 0.2 M PBS, pH 7.4) for ~1 h. The fixed 203 specimens were dehydrated in 100% ethanol for ~1 h, and then air-dried for ~30 min. 204 Coupons were placed on an SEM stage and secured using double-sided carbon tape, then 205 coated using a Hummer 6.2 gold sputter coater (Anatech LTD). All coupons were imaged with a 206 JEOL JSM-6610 SEM in secondary electron emission mode. 207 In addition to assessing biofilm morphology, because a combination of vortexing and 208 sonication was used to remove/disrupt biofilms as part of the baseline and treatment studies, it 209 was important to confirm that the vortex and sonication process did in fact remove the large 210 majority of biofilm from the surface of the coupons. After a thorough literature review, removal 211 of biofilms from collagen using vortex and sonication does not appear to have been confirmed 212 previously. Thus, for each bacterium, biofilms were grown on both materials, samples were 213 rinsed, vortexed, and sonicated as described, and were imaged by SEM (n=3) to determine if that 214 procedure effectively removes biofilm for an accurate quantification. 215 Lastly, in order to compare surface morphologies of the collagen and polycarbonate 216 coupon materials without bacteria on them, n=3 new and unused (negative control) coupons were 217 soaked in 10% BHI for 48 hrs. Coupons were then aseptically removed and processed as 218 described above for SEM imaging. 219

Statistical Analysis 220
Outcome measures (i.e. bacterial counts) were analyzed using an independent sample t 221 test for comparisons with alpha level set at 0.05 throughout. 222

Results 223
Biofilm Growth/SEM Imaging 224 SEM images showed that the collagen material alone was amorphous with a polymeric 225 strand network that had deep crevices throughout ( vortex/sonication, the levels of biofilm that were present initially were far more than any other 262 isolate (see Figures 3-7). Thus, the reduction was still significant. Based on surface area 263 coverage and reduction of the biofilm layers to a monolayer of cells, it was estimated that less 264 than 5% of cells remained on the surface after vortexing and sonication. 265

Baseline Quantification and Antibiotic Treatment 266
Quantification data of untreated coupons (positive controls of growth) is reported in 267  Table 3. 276 Biofilms of S. aureus ATCC 6538 grown on collagen or polycarbonate were minimally 277 affected by ciprofloxacin, but did show a statistically significant difference with biofilms on 278 collagen having lower reduction than those on polycarbonate at 200 µg/mL (p=0.001; see Table  279 3). Susceptibilities were the same when exposed to cefazolin across a range of concentrations 280 (Table 3; p>0.3 in both cases). In this data set, the hypothesis was supported in one instance with 281 ciprofloxacin, but in no cases with cefazolin. 282 In the case of P. aeruginosa ATCC 27853, there was an approximately 4-5 Log10 283 reduction following exposure to ciprofloxacin at 50, 100, and 200 µg/mL (Table 3). 284 Susceptibility of biofilms to ciprofloxacin on collagen was significantly less compared to those 285 on polycarbonate (e.g., p=0.03 at 200 µg/mL). Biofilm susceptibility to tobramycin had similar 286 outcomes. Tobramycin resulted in a 4-5 Log10 reduction at 50 µg/mL and complete kill at 100 287 µg/mL for both coupon types, with susceptibility being greater on polycarbonate coupons 288 compared to collagen (Table 3; e.g., p=0.046 at 50 µg/mL). The hypothesis was supported with 289 both antibiotics in this data set. 290 Biofilms of E. coli ATCC 9637 had the most notable differences in susceptibility profiles 291 between collagen and polycarbonate growth (Table 3). In all data sets, both ciprofloxacin and 292 ceftriaxone were more effective against biofilms on polycarbonate than those on collagen with 293 ceftriaxone having more polarized effect than ciprofloxacin (Table 3). Differences between 294 collagen and polycarbonate testing were all statistically significantly different. As a 295 representative example, p=0.001 for ciprofloxacin at 100 µg/mL. The hypothesis was most 296 strongly supported in this data set, in particular with ceftriaxone. 297 Biofilms of A. baumannii ATCC BAA 1605 were highly susceptible to colistin, with an 8 298 Log10 reduction (undetectable growth) on both collagen and polycarbonate coupons at 100 299 µg/mL, and near complete kill at 200 µg/mL (Table 3). The difference in CFU/coupon was 300 significantly different from baseline controls (p<0.005 in all cases), but there was no statistically 301 significant difference in the number of CFU/coupon between coupon types treated with colistin 302 at 50 µg/mL (p=0.371). These results indicated that efficacy of colistin against A. baumannii 303 biofilms was similar on both material types. Imipenem had minimal effect on biofilms on either 304 material type up to 200 µg/mL, and indicated that biofilm susceptibility was similar on both 305 materials (p=0.590). The hypothesis was not supported in any case for A. baumannii and the 306 antibiotics tested. 307 Biofilms of B. subtilis ATCC 19659 were found to be more susceptible to vancomycin on 308 polycarbonate than collagen (Table 3; e.g., p=0.001 at 100 µg/mL). Although there was no 309 detectable growth of biofilms on polycarbonate exposed to vancomycin at 100 µg/mL, an 310 anomaly was observed on polycarbonate coupons treated with vancomycin at 200 µg/mL; two of 311 five coupons had growth. The experiment was repeated with similar outcomes, resulting in a 312 large standard deviation at that concentration (Table 3) Table 3). At concentrations above 50 µg/mL, the hypothesis was 318 supported for both antibiotics tested in this data set. 319 In the case of MRSA USA 300, daptomycin was more effective at eradicating biofilms 320 on polycarbonate compared to collagen (p<0.001 in all cases), in particular at 400 µg/mL (Table  321 3). Vancomycin was nearly ineffective against biofilms on collagen (see Table 3) and reduced 322 biofilms on polycarbonate coupons to a much greater degree (e.g., p=0.002 at 400 µg/mL). 323 Similar to USA300 data, daptomycin was more effective at eradicating MRSA USA 400 324 biofilms on polycarbonate compared to collagen (e.g., p=0.003 at 100 µg/mL). Biofilms on 325 collagen showed no significant reduction by vancomycin at any concentration (Table 3) and on 326 polycarbonate showed a reduction of approximately 1.5 Log10 units by vancomycin at 100 µg/mL, 327 which was significant (p=0.001). Outcomes indicated that vancomycin was more effective at 328 eradicating MRSA USA 400 biofilms on polycarbonate compared to collagen. 329 Biofilms of S. mutans ATCC 25175 on collagen and polycarbonate showed no significant 330 reduction from baseline growth and no difference when exposed to 200 µg/mL erythromycin 331 (p=0.198), indicating that erythromycin was equally ineffective at eradicating S. mutans biofilm 332 on both material types. Biofilms on collagen were minimally affected by amoxicillin (Table 3. (Table 3). Biofilms on polycarbonate were reduced 343 by approximately 2.5 Log10 units with ciprofloxacin at 100 µg/mL and approximately 3 Log10 344 units at 200 µg/mL. There was a statistical difference between collagen and polycarbonate 345 outcomes with ciprofloxacin (e.g., p=0.004 at 200 µg/mL). Thus, data indicated that 346 ciprofloxacin was more effective against S. epidermidis biofilms on polycarbonate compared to 347 those on collagen. Outcomes supported the hypothesis with ciprofloxacin, but not vancomycin. 348 In the case of K. pneumoniae ATCC BAA-1705, ceftriaxone and imipenem/cilastatin had 349 minimal effect against biofilms on collagen or polycarbonate. There was about 1 Log10 more 350 bacteria that grew on collagen compared to polycarbonate. In all cases, there were more bacteria 351 on coupons in treatment groups than there were on positive controls. These data neither 352 supported nor unsupported the hypothesis as the isolate was resistant to both antibiotics, yet were 353 included to provide a comparison of outcomes when susceptibility was not present. 354 Taken together, data indicated that in 13/18 cases the hypothesis was supported. K. 355 pneumoniae data were not included in the outcome as biofilms treated with antibiotics had more 356 growth than positive controls. 357

Discussion 358
Biofilms can be grown on a wide variety of surfaces, materials and exposed to myriad 359 environmental conditions. One of the most common reactor systems to grow biofilms is the CDC 360 biofilm reactor, which is beneficial in that it is a robust system that can be modified to hold a 361 conformed to the polymeric collagen network. As noted in Figure 2, deep valleys and ravines 563 were consistent throughout the amorphous structure. (D-F) Surface of a polycarbonate coupon on 564 which biofilms of P. aeruginosa ATCC 27853 grew. Biofilm structure was estimated to be 565 greater than 20 cell layers thick. Growth followed the contour of the surface, even displaying the 566 coupon machine marks (100x). Arrow (panel F) indicates extracellular matrix components that 567 were observed. 568 569 Images of residual cells on collagen after vortex and sonication. Data indicated that biofilms 587 were effectively disrupted/removed with >5% of cells remaining on the surface. 588 589 coupons with deep ravines. Although these ravines may have formed during the dehydration 595 procedure, it is likely they were sites of water channels that provided fracture points within the 596 biofilm communities. (E-F) Images showed that there was still a fair amount of surface coverage 597 by biofilms post-vortex/sonication. However, it was estimated that there were >5% of cells 598 remaining on the surface similar to other bacterial isolates examined. 599 600