A collection of Serratia marcescens differing in their insect pathogenicity towards Manduca sexta larvae

We investigated the ability of Serratia marcescens to kill Manduca sexta (tobacco/tomato hornworm) larvae following injection of ca. 5 × 105 bacteria into the insect hemolymph. Fifteen bacterial strains were examined, including 12 non-pigmented clinical isolates from humans. They fell into 6 groups depending on the timing and rate at which they caused larval death. Relative insect toxicity was not correlated with pigmentation, colony morphology, biotype, motility, capsule formation, iron availability, surfactant production, swarming ability, antibiotic resistance, bacteriophage susceptibility, salt tolerance, nitrogen utilitization patterns, or the production of 4 exoenzymes: proteases, DNase, lipase, or phospholipase. There were marked differences in chitinase production, the types of homoserine lactone (HSL) quorum sensing molecules produced, and the blood agar hemolysis patterns observed. However, none of these differences correlated with the six insect larval virulence groups. Thus, the actual offensive or defensive virulence factors possessed by these strains remain unidentified. The availability of this set of S. marcescens strains, covering the full range from highly virulent to non-virulent, should permit future genomic comparisons to identify the precise mechanisms of larval toxicity.


Introduction
Serratia marcescens has been of interest to microbiologists for many years. Much of this 17 interest derives from the production of a bright red, blood-like pigment called prodigiosin by many 18 strains [1]. This red pigmentation led to countless deaths from "bleeding host" hysteria during the 19 Middle Ages [2] as well as providing the name for "red diaper syndrome" [3]. In early studies on the 20 feasibility of bacteriological warfare, this distinctive pigmentation was the rationale for S. marcescens

31
Eicosanoids are signaling metabolites derived from C 20 polyunsaturated fatty acids. In this work we 32 injected S. marcescens cultures into the larval hemolymph at ca 5 x 10 5 bacteria per larva and then 33 followed mortality as well as the ability or inability of larvae to clear the pathogenic bacteria from their 34 hemolymph. We observed that dexamethasone, an inhibitor of phospholipase A 2 , significantly reduced 35 the ability of larvae to clear pathogenic bacteria while this ability was restored by treatment with 36 arachidonic acid, the C 20:4 fatty acid released by phospholipase A 2 . We concluded that eicosanoids likely 37 mediate invertebrate immune responses [11]. 38 Naturally, we were concerned when another laboratory told us that they were unable to repeat 39 our findings. These difficulties were traced to the strains of S. marcescens being used. When they used 40 our strain (now called KWN) they could repeat our findings. This was the first indication we had that the 41 strain of S. marcescens used in our eicosanoid studies [11] was unusually pathogenic to insects. This 42 realization led us to compare 15 isolates of S. marcescens with regard to their pathogenicity towards M. 43 sexta larvae. The bacteria formed 6 pathogenicity groups, ranging from highly pathogenic to non-44 pathogenic. Strain KWN was in the highly pathogenic group. We now report how these isolates differ 45 from one another while providing hints regarding the biochemical and physiological factors responsible 46 for these widely differing pathogenicities. Although the precise mechanisms of their pathogenicity differences have not yet been identified, the collection should be of use because its members cover 48 stepwise gradations from being highly pathogenic to non-pathogenic.   Table 1.  [24,32], the red pigmentation by prodigiosin was only observed for 170 biotypes A1 and A2 (Table 1).

171
All strains produced large capsules when grown on the defined, high C/N Gauger's medium [16].

172
Capsules of roughly equivalent size were observed by negative staining [17] for all strains (Table 1). We 173 also compared the strains with regard to their motility on semi-solid 0.35% agar plates and their 174 swarming ability on 0.75% agar plates [18]. All strains were motile except for the non-toxic strain D1 175 and, while most of the strains were capable of swarming, there did not appear to be a correlation 176 between swarming and insect pathogenicity ( 179 marcescens for serrawettin production and wettability on four surfaces (S1 Table). Individual strains 180 produced each of the 3 serrawettins (W1, W2, and W3) identified by thin-layer chromatography by 181 Matsuyama et al [19] while strain D163 produced a unique TLC spot which we have called W4 (S1Table).

182
However, there was no correlation between serrawettin production and insect pathogenicity. For 183 instance, the highly pathogenic strain E223 had no detectable serrawettin or wettability (S1Table) and 184 only late developing swarming (Table 1).

185
Hemolysis and Salt Tolerance. Many bacterial pathogens release hemolysins to lyse erythrocytes and 186 other cell types so that they gain access to the nutrients in those cells, especially iron and hemoglobin.

197
The osmotic/salt tolerance of the respective strains was tested by their ability to grow on 198 mannitol-salt agar plates containing 7.5% NaCl. Twelve of the strains grew well, if slowly, on these 199 plates (Table 2) but once again there did not appear to be a correlation between salt tolerance and 200 insect pathogenicity. The three strains which could not grow on mannitol salt plates at 25°C all 201 exhibited high group 2 pathogenicity. There was also a strong temperature effect for salt tolerance.

202
For the 12 strains which grew well on mannitol-salt agar at 25°C, only two still grew at 37°C, and those 203 two grew poorly (Table 2). These results are consistent with those expected for a population of S.
204 marcescens [32] because all strains of S. marcescens ferment mannitol as the sole carbon and energy 205 source and, even though 0.5% NaCl is the optimal salt concentration for growth, > 90% of strains grew in 206 7% NaCl, 11-89% grew in 8.5% NaCl, and none grew in 10% NaCl [32]. Interestingly, the three red 207 pigmented strains (KWN, Nima, and D1) were colorless when growing on the high salt mannitol plates at 208 25°C (Table 2). It is well known that pigmentation is determined in part by cultural conditions, including 209 amino acids, carbohydrates, pH, temperatue, and inorganic ions, and that prodigiosin is not made 210 anaerobically [32]. Now we know that prodigiosin is not made on mannitol-salt agar plates.

211
Carbon and Nitrogen sources. Trehalose is the carbohydrate commonly found in insect hemolymphs. although strain 5384 used allantoin poorly. The latter 3 nitrogen sources were chosen because they 226 could be relevant for successful growth in the insect hemolymph [36]. Uric acid is a common 227 nitrogenous excretory product for terrestrial insects and it is often used as a nitrogen reserve [36] when 228 insects are grown on high nitrogen diets such as those used for tobacco hornworm larvae [14]. Urea 229 and allantoin are commonly the first products made in the degradation of uric acid.
230 Extracellular Enzymes. Serratia sp. are well known for the production of extracellular enzymes, including 231 chitinases [20,37], nucleases [22], proteases [38], lipases [39], and phospholipases [39].  Table) while lipases were detected on plates containing Tweens 20, 40, 60, or 235 80 (S4 Table). Additionally, for the proteases, lipases, and phospholipases, one set of plates was 236 incubated in air while a duplicate set was incubataed in a candle jar. The candle jar was chosen to 237 simulate the high CO 2 , microaerophilic environment of insect hemolymph [40]. However, only minimal 238 differences were observed (S3 and S4 Tables). The zones of clearing/hydrolysis for the 14 pathogenic 239 strains of S. marcescens generally agreed within + 10% while the zones for the non-pathogenic D1 were 240 often smaller because, being non-motile, their colonies were smaller. These results are consistent with 241 the impressive exoenzyme repertoire expected for all species of Serratia [32]. Our strains were also 242 tested for chitinase production on plates containing solubilized chitin [20]. Major differences were 243 observed (Table 1 and S4 Table), but it was the highly toxic strains which often did not have zones of 244 clearing. In particular, strains KWN and 9674 did not excrete chitinase (Table 1).

245
Autoinducers and Quorum Sensing Molecules. We next examined our collection of S. marcescens strains 246 with regard to their production of acylhomoserine lactone (HSL) autoinducers. N-acyl homoserine 247 lactone-based quorum sensing commonly regulates surfactant production [33], swarming [34,39,41], 248 adhesion and biofilm formation [42] as well as the release of exoenzymes [39] and exopolysaccharides 249 [42]. In addition, based on the precedent of other Gram negative bacteria, it could regulate an as yet 250 unidentified insecticidal toxin. We used the thin layer chromatography overlay method described by 251 Shaw et al [13] to identify HSLs based on color production by three reporter strains, Agrobacterium 252 tumefaciens, Chromobacterium violaceum, and a strain of Escherichia coli responsive to the C 12 3-oxo 253 HSL autoinducer made by Pseudomonas aeruginosa. Each of the reporter strains can respond to an 254 exogenous autoinducer but does not produce its own autoinducer. The Chromobacterium reporter 255 responds best to the C 4 , C 6 , and C 8 3-unsubstituted HSLs while the Agrobacterium reporter responds 256 best to the C 6 , C 8 , and C 10 3-oxo or 3-hydroxy HSLs [13]. Our results are shown in Table 3. There were 257 marked differences among the strains. Eight moderately insecticidal strains of clinical origin gave 1 or 2 258 purple spots with the C. violaceum detection system (Table 3). Based on their R f values these spots are 259 likely due to C 4 and C 6 3-unsubstituted molecules, butanoyl and hexanoyl HSL, identified by previous 260 researchers [39,41,42]. However, the highly insecticidal strains KWN and 9674 did not give any spots 261 with the C. violaceum detection system. Instead, they gave a single spot with the A. tumefaciens 262 detection system which is likely the 3-oxo or 3-hydroxyl C 8 HSL while strains 968A and 2698B gave spots 263 which are likely the 3-oxo or 3-hydroxyl C 10 molecules ( Table 3). None of the S. marcescens strains 264 tested produced any molecules which reacted with the 3-oxo C 12 specific reporter. These HSL 265 identifications are based on comparison of their R f values with known compounds and must still be 266 considered as tentative.  [13] of 0.82, 0.68, 0.41, 0.18, and 0.07 for the C 4 , C 6 , C 8 , C 10 , and C 12 3-oxo HSLs, respectively, and 0.77, 0.47, 0.23, 0.09, and 0.02 for the C 4 , C 6 , C 8 , C 10 , and C 12 3-unsubstituted HSLs, respectively. ND = not determined. Note that the R f values reflect migration on the C 18 reversed phase TLC plates and thus should apply equally for all three detection systems.
B/ For cells grown on trypticase-soy broth, none detected for cells grown on minimal ccy medium [25].
C/ None detected for cells grown either on trypticase-soy broth, or on minimal ccy medium [25].
D/ For cells grown on minimal ccy medium, only one AI (R f = 0.46) was detected.

267
Antibiotic Sensitivity/Resistance. Antibiotic resistance profiles were determined for 16 strains of S.
268 marcescens using 24 antibacterials for which antibiotic discs were commercially available (   a collection of seven broad host range   284  bacteriophage (SN-1, SN-2, SN-X, SN-T, BHR1, BHR2, and D 3 C 3 ), known to be active versus multiple Gram   285 negative bacteria [28]. The SN and BHR bacteriophage had originally been isolated from Sphaerotilus 286 natans and Pseudomonas aeruginosa, respectively [28]. We previously showed [28] that these broad 287 host range bacteriophage, as produced on either P. aeruginosa or S. natans, were unable to infect S.

288
marcescens KWN [28]. We now observed that they were also unable to infect or propagate on any of 289 the 15 strains listed in Table 1. No plaques were produced on agar plates and no increases in phage titer 290 were observed in liquid culture (data not shown). These findings are consistent with the generalization 291 of Grimont and Grimont [32] that bacteriophages isolated on genera other than Serratia rarely multiply 292 on Serratia. These phage sensitivity screens were conducted in part in the hope of identifying effective 293 biocontrol mechanisms for S. marcescens but also as an indirect method for determining the presence 294 and importance of Type IV pili in insect pathogenicity. The Type IV secretion system is a well known 295 virulence mechanism allowing bacteria to inject protein toxins into other cell types while phage SN-T 296 was shown to be broad host range because it attached to various bacteria by means of their Type IV pili 297 [43]. Thus, the absence of phage infectivity provides no evidence for Type IV pili under laboratory 298 conditions but does not rule out a role for Type IV secretion in insect pathogenicity.

300
Fifteen strains of S. marcescens were examined for their insect pathogenicity towards M. sexta 301 larvae. They fell into six groups, ranging from the highly toxic Group 1 to non-toxic Group 6 ( Table 1).

302
The need for comparatively high inocula (5 x 10 5 bacteria per larva) to achieve these variable LD 50  conditions, four of the six highly pathogenic (Groups 1 and 2) strains were chitinase negative (Table 1).

319
The relative importance of extracellular chitinase would likely have been different if the bacterial cells 320 had been provided in the diet or sprayed on the larval surface where they would have had to penetrate 321 the peritrophic membrane or the cuticle to exert their pathogenicity. We note that Ruiz-Sanchez et al 322 [44] compared 102 strains of S. marcescens for their chitinolytic activity. They found that S. marcescens 323 Nima, which exhibited very little pathogenicity in our assays (Fig. 1), had ca. 43 times higher chitinolytic 324 activity than most other S. marcescens strains [44].

325
Pathogenicity did, however, correlate with the type of autoinducer produced (Table 3). S.

326
marcescens, like a great many Gram negative bacteria [45], is known to produce HSL autoinducers which 327 act in a quorum sensing manner [41,46]. Bainton et al [46] used an autoinducer-dependent 328 bioluminescence system to detect 3-oxo-hexanoyl HSL activity in S. marcescens supernatants. The 329 identity of this molecule was confirmed by infrared, mass spectrometric, and NMR analysis [46]. Later, Eberl et al [41] showed that Serratia liquefaciens, now S. marcescens [42], produced butanoyl and 331 hexanoyl HSLs in a ratio of 10:1 and these autoinducers controlled the differentiation to swarming 332 motility. In this case, the autoinducers without a side chain oxygen at position 3 were more effective in 333 causing swarming than those with a 3-oxo side chain [41]. The distinction is relevant because: A/ the 334 strain of S. liquefaciens used [41] has been reclassified as S. marcescens based on its 16S rRNA sequence 335 [42]; and B/ the Agrobacterium based TLC detection system strongly prefers the 3-oxo homoserine 336 lactones whereas the Chromobacterium based system strongly prefers the 3-unsubstituted molecules 337 [13]. Based on a comparison with R f values of known HSLs [13], the hightly insecticidal strains KWN and 338 9674 (Group 1) produced 3-oxo or 3-hydroxy C 8 HSL while the moderately insecticidal clinical isolates, 339 1682 through 968A in Table 3, produced one or both of the 3-unsubstituted C 4 and C 6 HSLs (Table 3), the 340 same HSLs as found by Eberl et al [41]. Finally, the poorly insecticidal strains 968A and 2698B produced 341 3-oxo hydroxy C 10 HSL. We believe that the 3-oxo or hydroxy C 8 and C 10 HSLs are autoinducers not 342 previously reported from Serratia sp.

343
Unfortunately we have as yet little evidence regarding which genes or virulence factors are 344 being regulated by the respective autoinducers. The extracellular nuclease of S. marcescens is regulated 345 in a growth-phase and cell-density dependent manner [22] as are all the exoenzymes produced by S.

346
liquefaciens [39], now S. marcescens [42], and the C 4 and C 6 HSL autoinducers also regulate surfactants 347 [19,33], swarming [34,39,41], adhesion, biofilms, and exopolysaccharide production [42]. Some strains 348 of S. marcescens also produce a heat-labile enterotoxin as a virulence factor [47] which could also be 349 regulated in a quorum sensing manner. An additional complication is that all of our studies were 350 conducted with typical bacterial growth media under aerobic growth conditions. However, since the 351 HSLs produced can vary depending on the growth media and growth conditions [13 and