OUP user menu

Serratia marcescens and its extracellular nuclease

Michael J Benedik , Ulrich Strych
DOI: http://dx.doi.org/10.1111/j.1574-6968.1998.tb13120.x 1-13 First published online: 1 August 1998


Serratia marcescens produces an endonuclease with extraordinarily high specific activity that is released into the surrounding medium. This enzyme has been the focus of studies on gene regulation, protein secretion, endonuclease action, and protein structure; it has also been found to have many applications in biotechnology. Here we briefly review these different facets of research regarding the Serratia nuclease and summarize the current state of knowledge about this enzyme.

  • Serratia marcescens
  • Nuclease
  • Extracellular secretion
  • DNA hydrolysis

1 Introduction

Students of microbiology often recall their introduction to Serratia marcescens in laboratory exercises where it was used because of its red pigment, prodigiosin. Variation in pigment formation is often used to demonstrate unstable phenotypes and the natural variations found within a population [1]. Surprisingly, the genetic and biochemical mechanisms leading to variegated production of prodigiosin remain uncharacterized to date.

A second and often cited hallmark of Serratia spp., which in fact characterizes the genus, is its extracellular nuclease. This protein, of ∼26 kDa, is the product of the nucA gene. Although many bacteria secrete extracellular proteins, and Serratia produces an abundance, extracellular nucleases are rare. They are limited to a small number of bacterial species spread broadly throughout the eubacterial kingdom. Surprisingly, these extracellular nucleases are mostly unrelated to one another by sequence or by mechanism and structure.

2 Why a nuclease?

What role might an extracellular nuclease play in the biology of a microbe? In no case can we be certain of the answer, but reasonable speculations certainly abound.

Scavenging is likely to be the predominant role these nucleases play. Unlike the luxurious life of Escherichia coli in a shaker flask of LB medium, most microbes are faced with a continual shortage of nutrients and lead an ever-changing existence of fast or famine. Extracellular degradatory enzymes are one mechanism microbes use to exploit their environment and maximize availability of nutrients. Nucleic acid may not seem a particularly suitable carbon source, however under laboratory conditions it has been demonstrated that Serratia is capable of slow growth on DNA as sole carbon source [2]. The ribose and deoxyribose are likely to be used directly to provide energy. Equally important, scavenging nucleic acids for use as nucleotides in DNA and RNA synthesis may be an efficient way to conserve on the metabolically expensive synthesis of these precursors.

The possible role of nuclease as a scavenging enzyme becomes more apparent for Serratia when one considers the other extracellular enzymes produced by this microbe. Among them exist multiple proteases [3, 4], with at least one, the metalloprotease, being of great abundance and high specific activity. Other proteases of varying specificity contribute to the net proteolytic activity. Lipolysis is mediated by two separate enzymes, a general lipase or esterase [5, 6] as well as a phospholipase [7]. Chitin degradation is mediated by at least two chitinases as well as chitobiases [8, 9]. Taken in combination, these enzymes allow Serratia to fully degrade complete organisms such as fungi and insects to simple metabolites. Other proteins are likely to be important for metabolite scavenging during an infection. The HasA protein [10] works as an iron scavenging protein, similarly the hemolysin [11] may allow release of sequestered iron from blood cells. Production of bacteriocin [12, 13] may improve the ability of Serratia to compete in a mixed microbe environment.

It has been suggested that nucleases may act as virulence factors, for example the nuclease of Vibrio cholerae has been postulated to play a role during invasion or establishment of an infection [14], however no experiments have successfully demonstrated this. In the lungs, exceptionally high concentrations of DNA comprise the viscous component of bronchial secretions. The ability to locally degrade these secretions, and allow bacteria to attach to cell surfaces hiding below, may assist in establishing a successful infection. The role of S. marcescens as an opportunistic pathogen is being increasingly recognized [15], however it is not clear whether DNA poses a problem in any of these infections nor whether the nuclease is important. One can imagine the nuclease would contribute to infectivity under any conditions where copious cell lysis is occurring.

3 An overview of the enzyme

Nuclease was first purified from culture supernatants of S. marcescens more than 30 years ago [16, 17]. Since that time it has been the target of numerous biochemical studies. The enzyme is a sugar-nonspecific hydrolase, i.e. capable of cleaving both RNA and DNA in either double or single stranded form. It requires divalent cations, preferably Mg2+, displays a broad pH range from 6 to 10 (optimal at 8–8.5) and has a wide temperature optimum between 35°C and 44°C [1619].

The nuclease enzyme has a high catalytic efficiency relative to other nucleases, about four times that of staphylococcal nuclease and approximately 34 times greater than DNase I. Kinetic studies have indicated that the rate limiting step for the reaction is an early one, association of the enzyme with its substrate and/or the cleavage of the resulting complex, not the dissociation of the product [20]. Recent studies have demonstrated that nuclease activity on excess substrate is not limited by substrate inhibition [20]; earlier reports to the contrary [21, 22] failed to maintain constant Mg2+ to DNA ratios needed for maximal activity [23]. The enzyme cleaves single or double stranded DNA and RNA with similar rates, so long as the substrate contains no fewer than five phosphate residues [20]. Under in vitro conditions, however, nuclease has been shown to prefer GC- over AT-rich regions, as does DNase I. When presented with a double stranded substrate, cleavage of one strand is correlated with cleavage in the corresponding position of the complementary strand yielding a double strand break [24].

4 A family of nucleases

The Serratia enzyme belongs to a family of sugar-nonspecific nucleases found in organisms ranging from bacteria to mammals. Other members of this group of nucleases have been characterized from Bos taurus[25], Synecephalastrum racemosum[26], Saccharomyces cerevisiae[27], Anabaena sp. PCC7120 [28] and Streptococcus pneumoniae[29]. They are generally homodimers with monomer molecular masses ranging from 25 to 30 kDa. Sequence analysis revealed the presence of several conserved amino acid residues within the members of this family, suggesting a common reaction mechanism [28, 30].

A comparison of Serratia nuclease to other endonucleases reveal differences in their targets on polynucleotide substrates (Fig. 1). Serratia nuclease and DNase I [31] both cleave to yield a 3′-OH and 5′-phosphoryl whereas staphylococcal nuclease cleaves at the other side of the P bond [32]. The chromophoric nuclease substrate deoxythymidine 3′,5′-bis-(p-nitrophenyl-phosphate) however is cleaved differently by DNase I and Serratia nuclease. DNase I cleaves the phosphodiester bond on the 3′ side [33], and Serratia nuclease attacks the 5′-side of the ribose sugar moiety of this particular substrate analog. This may reflect a functional difference in substrate recognition and cleavage by these enzymes or, more likely, may simply suggest some peculiarities of these enzymes' binding to this artificial substrate.


The cleavage sites for DNase I, staphylococcal nuclease and Serratia nuclease on a polynucleotide substrate.

5 Processing of the nuclease precursor

Nuclease is produced as a pre-protein of 266 amino acids with a signal peptide consisting of the first 21 residues [18, 34]. Two major isoforms are produced by S. marcescens[35, 36]. The first isoform, the 245 amino acid mature nuclease (Sm2) with a molecular mass of 26.7 kDa, is the result of typical signal sequence processing. The second isoform (Sm1) is three amino acids shorter, lacking the first three N-terminal amino acids Asp-Thr-Leu. As might be expected both isoforms share a high degree of structural similarity and possess nearly identical biochemical properties, differing only minutely in their isoelectric points [35, 3740]. However, subtle differences in their interaction with substrate, reflected in their base preferences, have recently been noted [23, 41].

In addition to these major isoforms, other isoforms present only as minor species have been identified by capillary electrophoresis [42] and electrospray mass spectrometry [4345]. These have further short deletions at the N-terminus. These isoforms are not just products from S. marcescens; the three amino acid deleted isoform Sm1 has also been detected when nuclease is produced as a recombinant protein in E. coli, albeit in smaller quantities [3840].

Several mechanisms for the generation of these isoforms have been proposed. The native molecule could be processed by various membrane or periplasmic proteases, present in both S. marcescens and E. coli, while passing through the cell envelope. Alternatively, the isoforms might be the result of heterogeneous processing of the N-terminal signal peptide by the leader peptidase. However Suh et al., [46] were unable to demonstrate that Sm2 could be processed or chased into the Sm1 form and concluded that these isoforms likely represent alternative signal peptide processing events. Their production is also temporally distinct, the Sm1 isoform is made later during the growth of a culture than is Sm2.

The secretion of nuclease may also regulate its activity indirectly. It appears that disulfide bridge formation is a necessary requirement for enzyme activity. The pre-enzyme is probably inactive in the cytoplasm and only becomes active after secretion when the protein reaches the oxidizing environment of the periplasm necessary to allow formation of disulfides. Two disulfide bridges between C9 and C13 in the N-terminal region, and C201 and C243 near the C-terminus were first identified by Biedermann et al., [18] and were later confirmed through high resolution peptide mapping [3840]. Site-directed mutagenesis of the above residues revealed their importance for nuclease stability and activity. A C9S mutant, for example, is five- to ten-fold less stable than the wild-type enzyme and displays a more than 1000-fold reduction in nuclease activity [47]. Thus, disulfide bridge formation can be considered an essential step in regulating nuclease activity, and a crucial means of restricting its potentially detrimental effect to the extracellular milieu, in combination with its removal from the cytoplasm by the secretory machinery.

6 Structure of nuclease

The primary [34], secondary [30], tertiary [48] and quaternary structures [49] of nuclease are all well characterized. The X-ray crystal structure of the free monomer was first solved to 2.1 Å resolution [48], a revised 1.7 Å resolution structure was recently published [50] and final refinement of the structure to 0.92 Å has just been completed (M. Miller and K. Krause, personal communication). This structural resolution is among the highest reported to date for proteins of this size. One characteristic of the enzyme is its unique endonuclease fold in the center of which is a β-sheet made up of six antiparallel β-strands (Fig. 2). On one side, the fold is flanked by a single dominant α-helix and a very long coiled loop, on the other side there is a helical domain composed of three α-helices and two small antiparallel β-strands. Electrostatic field calculations revealed a strongly polarized surface, which helps to attract and direct substrate to the binding pocket. Two ridges, containing positively charged amino acids like R57, R87 and R131, flank a deep cleft that contains the catalytic center of the nuclease. Modeling predicts that these ridges could interact with about one full turn of B-DNA (Fig. 3). Prominent residues within this cleft of the enzyme are H89, N119 and E127 [48].


A ribbon diagram [70, 71] depicting the nuclease dimer with N- and C-termini and the H89 residues in the active sites at the end of long α-helices.


A nuclease monomer attacking B-DNA. The α-carbon tracing [48, 72] shows that the enzymes' active site cleft and its substrate are complementary in shape and charge.

Early evidence for dimerization came from sedimentation velocity experiments [51, 52] and crosslinking studies [53]. Subsequently dynamic light scattering experiments and the analysis of X-ray crystallographic data confirmed nuclease is a homodimer that forms through complementary interactions between certain residues in the carboxy-terminal subdomain. The interface is stabilized by four symmetric salt links and multiple hydrogen bonds bridging the monomers, dominated by interactions between the two H184 residues from each monomer. In this model the active site of each of the monomers is located away from the interface, unobstructed in its ability to interact with the substrate. Computer simulations confirmed that dimerization does not affect those electrostatic properties of the surface which contribute to the correct alignment of enzyme and substrate [54].

When encountering substrate molecules, the electrostatic field at the center surface of each monomer directs the negatively charged polynucleotide towards a nucleotide binding site along two positively charged ridges that flank the active site cleft. The enzyme structure, in agreement with kinetic studies [20], suggests independent or non-cooperative action of each of the dimers' active sites. The geometry of the structure predicts that the two active sites in the dimer do not act concurrently on a single substrate molecule, except perhaps in the case of a long substrate polymer that can wrap around to interact with both catalytic sites. Despite the non-cooperativity of the two active sites, dimerization may have certain advantages for nuclease; it increases the chance for the enzyme to encounter substrate and it allows the required enzyme rotation to occur in either direction when a substrate molecule is directed from the midsection of the dipole towards the active center(s) at its ends [49].

7 Catalytic mechanism

Catalytically important amino acid residues of nuclease have been identified by structural analysis, sequence comparison, and by site-directed mutagenesis. Mutations of residues R57, R87, H89, N119 and E127 resulted in enzymes that were found to be catalytically inactive confirming that these residues constitute the active site of nuclease [30, 48, 55].

In view of their prominent location in the center of the active site of the nuclease model (Fig. 4), the amino acid residues H89 and E127 were primary targets for mutational and biochemical investigations [48]. Mutant H89A is affected exclusively in its kcat, but not in its Km, suggesting H89's immediate involvement in the actual hydrolysis of the phosphodiester bond [55]. One proposed reaction mechanism (Fig. 5A) predicts that H89 acts like a general base, which abstracts a proton from a water molecule, activating it for a nucleophilic attack on the phosphorus atom adjacent to the scissile bond [33]. In this model N119, along with residue R57, stabilizes the resulting phosphorane intermediate. Interestingly a recent structure determination of nuclease identified N119 as being bound to Mg2+ (Miller and Krause, personal communication). Therefore the stabilizing role of N119 is likely mediated by this Mg2+. This residue could have an additional role in positioning the attacking water molecule relative to the phosphorus atom. A similar function has been discussed for R57, acting through its guanidinium group [33, 55].


α-Carbon tracing [72] of the Serratia nuclease active center. A magnesium cation in the center (red) is bound to the amide oxygen of Asn119. Five water molecules (orange) are located in the sphere of the magnesium.


Two proposed models for the catalytic mechanism nuclease attack. H89 in the active center of the enzyme is perceived to act either as a general base (A) or a general acid (B) in a Mg2+-dependent hydrolysis reaction.

An alternative model (Fig. 5B) suggesting that H89 might instead function as the general acid, protonating the leaving group, and E127 being the general base [48, 50, 54] has also been proposed. However the ability of H89A to cleave the artificial chromophoric substrate deoxythymidine 3′,5′-bis-(p-nitrophenyl-phosphate), which does not require protonation of the leaving group, appears to contradict this model. Mutant H89A is inactive with this nucleotide analogue, whereas an E127A mutant still cleaves. This argues that E127 is dispensable for the initial water activation [33], and therefore by default H89 is the opposite, at least for this artificial substrate.

When the role of those amino acid residues that are essential for nuclease activity, but are not directly involved in the catalysis reaction, was studied by site-directed mutagenesis, three additional residues outside of the active center were identified. R87, R131 and possibly also D86 mutants are mainly affected in their ability to bind nucleic acid substrate but not in their catalytic activity [55]. More specifically, R87 is thought to interact with the phosphate group at the 3′-end of the ribose sugar [33], which is essential for cleavage near the 5′-end [20]. With regard to the nuclease model, all three residues are located in the putative substrate binding site of the enzyme, suitably positioned to assist in positioning diverse nucleotide substrates into a conformation that is accepted by the enzyme [48].

8 Control of nuclease expression

Not surprisingly, nuclease production is regulated. However the parameters of its regulation are not initially obvious. Nuclease expression is not substrate regulated, the addition of nucleic acid does not induce its expression nor does the addition of free nucleotides repress it. It also is not catabolite regulated. Instead environmental signals control nuclease expression. First, transcription of nucA is growth phase regulated [56]. Transcription increases as growing cultures increase in density and approach saturation. Preliminary results (Y. Suh and M. Benedik, unpublished) suggest this is modulated by a factor released by the bacteria into the growth media, likely to be related to the HSL signaling molecules [57]. How this signal mediates its effect on nucA transcription has yet to be elucidated.

A transcriptional regulator which acts at the nucA promoter as an activator of transcription is the NucC protein. This protein is a member of the P2 Ogr family of phage transcriptional activators [58] and likely interacts with the α subunit of RNA polymerase. NucC binds to a region between −82 and −51 upstream of the transcriptional start of nucA as determined by footprinting analysis and this region includes a copy of the TGT-N12-ACA activator recognition motif. Deletion of the upstream TGT obliterates activation (G. Christie, personal communication). Transcription of nucC appears also to be growth phase regulated. The simplest model would have extracellular density signals modulate production of NucC which in turn activates nucA transcription, however this suggestion awaits experimental demonstration.

Interestingly, nucC lies in an operon with two other genes, nucD and nucE[58]. These bear a strong resemblance to phage proteins, specifically NucE resembles holin proteins involved in releasing lysozyme to the peptidoglycan of Gram-negative bacteria and NucD is such a lysozyme. These proteins appear to play no significant role in nuclease secretion; their deletion from the Serratia genome does not affect nuclease secretion (U. Strych, W. Dai and M. Benedik, in preparation). The likely viral origin of these genes makes it easy to speculate that they originated as part of a cryptic prophage.

A second environmental signal known to regulate nuclease expression is the bacterial SOS system [59]. Nuclease production is increased strongly by agents which induce SOS controlled genes. A LexA binding site lies near the transcription start site of nucA[56], a second site lies upstream of the nucC operon [58] and controls its transcription. So nuclease appears to be regulated dually by SOS through LexA repression of transcription both at nucA itself as well as its activator gene nucC. Overexpression of LexA reduces nuclease levels as do mutations in recA. Mutations in the LexA binding site of the nucA promoter increase nuclease expression dramatically.

Pleiotropic regulatory mutants have also been described which increase nuclease expression [56, 60]. Recent work in our laboratory has shown that most of these act indirectly by partially inducing the SOS system, therefore their role in regulating nuclease expression is indirect (L. Guynn and M. Benedik, submitted).

There exist two potential start codons for the nuclease open reading frame. Only the second is positioned well with respect to a consensus ribosome binding site. Site-directed mutations at these start codons demonstrate that both can be used in vivo, however the first Met start is used with less than 10% the efficiency of the second. No translational regulation has been identified [56].

Sequence analysis of the nucA gene from five different Serratia strains, not surprisingly, reveals a high degree of conservation (M. Benedik, unpublished). Although each strain has at least a single amino acid change, no strain differs by more than two amino acids from the consensus and all these changes are neutral. There is of course more variation found at the DNA level, but the genes are still about 96% identical. None of the base substitutions would be expected to alter gene expression or regulation.

9 Extracellular secretion of nuclease

Nuclease is found as an extracellular enzyme in cultures of S. marcescens, as are many other proteins. Remarkably, Serratia uses a variety of secretory systems by which it exports proteins to the growth medium. The mechanism employed to export nuclease remains obscure and has features suggesting it may be unique. The ability of Serratia to secrete nuclease appears to be regulated, probably by host cell physiology. Bacterial cultures at differing cell densities display different kinetics and efficiencies of nuclease secretion, these parameters can also be modulated by growth medium, growth conditions, and host cell mutations [61].

Nuclease is initially synthesized as a precursor protein carrying an N-terminal signal peptide [18]. This signal peptide mediates the export of nuclease to the bacterial periplasm using the standard general secretory system; inhibitors of E. coli envelope secretion also block the export of nuclease to the periplasm [61].

Is nuclease intrinsically ‘greasy’ leading to its extracellular secretion? Early reports suggested this may be the case. If nuclease is overexpressed in E. coli one is able to find significant enzymatic activity in the growth medium [34, 62]. This appears to be especially true when cultures are grown to high densities under conditions of very high expression such as in a well aerated fermenter [62]. For production purposes this makes it possible to easily produce and purify nuclease without lysing the bacterial cells. In fact, this property may be true for many typically periplasmic enzymes and has been observed both for bacterial alkaline phosphatase and β-lactamase. It is likely that this is an artifact of overproduction and does not represent the normal ‘biological’ situation.

In a series of experiments performed at ‘normal’ levels of nuclease expression, very different conclusions were formed. When nuclease is expressed in E. coli it is found predominantly in the bacterial periplasm whereas under similar conditions the enzyme is mostly extracellular when produced in S. marcescens[61]. The periplasmic species is an intermediate in the secretion of nuclease from the cytoplasm to the periplasm via the general secretory path, and then extracellularly via an unknown mechanism. This second step is relatively slow and, depending upon growth conditions, may take 5–60 min. This difference between E. coli and Serratia points to a specific mechanism existing for the extracellular step of nuclease secretion by Serratia under these conditions.

Two surprising observations have been made regarding nuclease secretion. The first suggests a promoter specificity for extracellular secretion. When nuclease is produced from a plasmid where transcription initiates from the lac promoter, the enzyme is found predominantly in the periplasm, whereas if expressed from the native nucA promoter even on plasmid, the enzyme can be found extracellularly [63]. Careful monitoring of pulse-labeled protein confirms this localization remains true for hours after labeling. The simplest explanation for this data suggests that precise temporal control of nuclease production may be important to allow efficient secretion. Expression at the wrong phase of growth results in protein trapped solely in the periplasm. The nucA promoter may be coordinately regulated with some facet of the secretion machinery and protein production and secretion must be coupled for extracellular secretion to occur [61, 63].

A second and perhaps more useful observation, nevertheless, yields a similar interpretation. Recall that there are two predominant isoforms of the processed mature nuclease. The Sm2 isoform is the native enzyme after removal of the signal peptide. The slightly shorter Sm1 isoform lacks the first three N-terminal amino acids after processing of the signal peptide. Suh et al., [61] were unable to demonstrate that Sm2 could be chased to the Sm1 form and suggest these species represent alternate signal peptide processing events. The appearance of these isoforms are not contemporaneous, rather the Sm2 form appears predominantly during exponential growth whereas a switch to the Sm1 isoform occurs as the cells enter stationary phase. In itself this would be unique for secreted proteins if it truly represents a change in signal peptide cleavage events. More striking is the fate of these isoforms. The Sm2 isoform is secreted extracellularly whereas the Sm1 isoform remains trapped in the periplasm until much later, such as after overnight growth.

What does this all mean? The obvious conclusion is that for nuclease there is no simple signal present in the protein which triggers its extracellular secretion. Expressing identical proteins from different promoters, or virtually identical isoforms from the same promoter yield species with very differing fates, either trapped in the periplasm or secreted extracellularly. Rather some facet of spatial or temporal localization may have an effect on the final destination of nuclease.

10 Nuclease and its many uses

With its high intrinsic activity and broad substrate tolerance, nuclease has found many uses in a variety of biotechnological applications. Purified nuclease (under the tradename of Benzonase by A. Benzon Pharma) is being sold commercially. Its primary market is for use in downstream processing. Typical uses are to eliminate nucleic acid contamination from purified proteins, commonly from recombinant DNA products, or to reduce viscosity for subsequent processing steps. The lysis of any significant quantity of bacteria leads to the release of nucleic acid contaminants which greatly increases the viscosity of the sample and can lead to complications in subsequent purification steps. The addition of nuclease is an inexpensive method to remove this contaminating nucleic acid. It has the added advantage of destroying any recombinant DNA molecules that might otherwise contaminate the target protein of choice.

The ability to destroy DNA has led to use of the S. marcescens nuclease as a killer gene for the self-destruction of microorganisms released into the environment. In general, nuclease requires secretion out of the cytoplasmic compartment to gain activity, however if expressed in the absence of a signal peptide, enough residual activity remains to kill cells by hydrolyzing their nucleic acids. Expressing the gene for mature nuclease, thereby preventing its export, results in a selective system for the destruction of any bacterium carrying such a gene. Therefore the release of genetically engineered organisms can be controlled, environmental signals triggering the expression of such a toxic cassette results in suicide. There are of course numerous such suicide systems, however the nuclease system has the added advantage of destroying any recombinant DNA plasmids as well, thereby alleviating fears concerning the release of such DNA [64].

There has been much interest in exploring novel uses for the Serratia nuclease in the former Soviet Union. Studies have shown nuclease can act as an effective anti-viral agent and inhibit the replication of both DNA and RNA viruses [65]. In fact it is extensively used to prevent viral paralysis in honeybee hives. Along the same lines, Boeke and colleagues [66] have developed an elegant antiretroviral therapy based on delivery of nuclease by a modified retroviral genome to block subsequent heterologous viral replication in the cell. Serratia nuclease was found to be too active and toxic to the host cell; RNase H1 was found useful instead. However other applications for such a toxic agent could be envisioned using a viral delivery system when the death of a target cell is the desired outcome.

Nuclease has also been shown to have significant [67, 68] anti-tumor properties, presumably by interfering with replication of dividing cells. Obviously a suitable targeted delivery system is the limiting factor for an effective treatment, something also lacking for many other molecules with similar anti-tumor properties. Of significant importance to the use of nuclease in any therapeutic regime is its ability to be tolerated by the mammalian immune system ([69], M. Filimonova, personal communication).

11 Epilogue

Although under investigation for decades, interest in the Serratia nuclease has been resurgent for the past 10 years. It remains an active subject of research by many on secretion, mechanisms of nuclease action and evolution, as well as for its unique properties useful in many biotechnological applications with undoubtedly many more to come.


For the dozen years the Benedik lab has labored on this topic many wonderful and stimulating students and post-doctoral fellows, past and present have made important contributions. To them we give our sincere thanks. Drs. Mitch Miller and Kurt Krause are gratefully acknowledged for their assistance with figures for this article. The Serratia community at large, albeit small, has been helpful, encouraging, and always open with their ideas, which has made the time pass swiftly. Work on Serratia has been supported by the NIH, NSF, Texas Advanced Research Program and Welch Foundation, we thank them all.


View Abstract