A Love Affair with Bacillus subtilis

My career in science was launched when I was an undergraduate at Princeton University and reinforced by graduate training at the Massachusetts Institute of Technology. However, it was only after I moved to Harvard University as a junior fellow that my affections were captured by a seemingly mundane soil bacterium. What Bacillus subtilis offered was endless fascinating biological problems (alternative sigma factors, sporulation, swarming, biofilm formation, stochastic cell fate switching) embedded in a uniquely powerful genetic system. Along the way, my career in science became inseparably interwoven with teaching and mentoring, which proved to be as rewarding as the thrill of discovery.

My career in science was launched when I was an undergraduate at Princeton University and reinforced by graduate training at the Massachusetts Institute of Technology. However, it was only after I moved to Harvard University as a junior fellow that my affections were captured by a seemingly mundane soil bacterium. What Bacillus subtilis offered was endless fascinating biological problems (alternative sigma factors, sporulation, swarming, biofilm formation, stochastic cell fate switching) embedded in a uniquely powerful genetic system. Along the way, my career in science became inseparably interwoven with teaching and mentoring, which proved to be as rewarding as the thrill of discovery.

Early Flirtations with Escherichia coli and Salmonella
I was a junior at Princeton University (1963)(1964) when I walked into the Frick Chemical Laboratory office of a newly arrived professor of chemistry, Charles Gilvarg. I was a chemistry major and knew I wanted to be a scientist for as long as I could remember, but I understood little about what that meant. There was a lot of buzz about this brilliant biochemist from New York University, and I was hoping I could land a position on his research team. I remember how nervous I was for my interview, but fortunately, Gilvarg did accept me into his laboratory. He was an intimidating presence but, as I came to appreciate, a wise and dedicated mentor who launched me on my career in science. In what culminated in my senior thesis, I was charged with studying how ␣-acetylation affected the use of lysine oligopeptides by Escherichia coli. I learned that following ␣-acetylation, oligopeptides were poorly taken up by E. coli, derepressing the production of a reductase in the lysine biosynthetic pathway. This was by no means an earth-shattering discovery, but I was thrilled to have learned something that no one else on the planet had known before. It is a lesson I have tried to pass on to generations of undergraduates: few things in college are more rewarding than engaging in research and discovering something, however modest, that no one knew before.
I learned another lesson when I submitted my senior thesis. Gilvarg went through it with a fine-toothed comb, sparing no opportunity to criticize unjustified conclusions. In fact, I was delighted and flattered that he cared so much about what I had done and written that he took the time to give me detailed feedback. Later, he submitted a paper based on my senior thesis to the Journal of Biological Chemistry, whose legendary editor John T. Edsall would later become my colleague at Harvard University. When a review arrived, Gilvarg wrote in the margin, "This shows the referee really read, understood and appreciated the paper." It was the same lesson I had taken away from his critique of my thesis. "Effect of ␣-acetylation on utilization of lysine oligopeptides in Escherichia coli" was my first publication (1), and it appeared in the very journal that half a century later would invite me to write this Reflections article.
It was not, however, with E. coli that I had a love affair, nor was it with Salmonella (Salmonella anatum) despite the intense relationship that S. anatum and I developed during my three years as a graduate student in Phil Robbins' laboratory at the Massachusetts Institute of Technology (MIT; Gilvarg had urged me to do my Ph.D. research with Robbins). It was an exciting time to be in the Robbins laboratory; a highlight of that period was the discovery of the C 55 polyisoprenoid lipid carrier involved in O-antigen biosynthesis. (Interestingly, the same lipid carrier was simultaneously shown to be required for peptidoglycan biosynthesis by my future mentor and friend Jack Strominger.) My own project focused on an S. anatum phage (⑀ 15 ) that uses O-antigen with ␣-glycosyl linkages as a receptor (2). Cleverly, upon infection and lysogeny, ⑀ 15 converts the O-antigen to ␤-glycosyl linkages, thereby blocking its brethren from infecting the same cell. I provided evidence that the phage replaces the ␣-polymerizing enzyme of its host with a phage-borne ␤-polymerase. It turned out that another, even cleverer phage (⑀ 34 ) uses the ␤-linked O-antigen as a receptor, a cat-and-mouse tale that Robbins and I recounted in a Scientific American article (3). Robbins was an inspirational mentor who liked nothing more than doing experiments with his own hands, working side by side with members of his own laboratory (including, at the time, the future independent scientists Dennis Bray at the University of Cambridge, Andrew Wright at Tufts University Medical School, and Henry Wu at the Uniformed Services University of the Health Sciences).
As the end of my Ph.D. studies approached, I decided I wanted to pursue postdoctoral research with Julius Adler at the University of Wisconsin. I was mesmerized by his experiments on chemotaxis and by the fact that it was possible to study behavior in E. coli. However, my intention to move to the University of Wisconsin was upended when Salvador Luria nominated me to be a junior fellow in the Harvard Society of Fellows. Luria was one of my heroes at MIT and an inspiring presence and teacher. (He invited graduate students to his home to discuss the writings of the French existentialist Albert Camus!) I think the fix was in as Senior Fellow Jim Watson had been Luria's first student and rightly or wrongly trusted his judgment. Watson told me that Jack Strominger was coming to Harvard and suggested that, given my experience with cell-surface biochemistry, I ask him for space in his new laboratory down the hall. Strominger agreed, and I soon found myself on the third floor of the Biological Laboratories, with Jim Watson and Wally Gilbert and their joint laboratory (with Klaus Weber) on one side and George Wald on the other. This was an extraordinary location and an intensely exciting time, as the third floor was home to nine individuals who had won or would later win a Nobel Prize. In addition to Wald, Watson, and Gilbert, future Nobel Prize-winning inhabitants of the third floor included the undergraduate Marty Chalfie; graduate students Bob Horvitz, Mario Capecchi, and Craig Mello; the postdoctoral fellow Rich Roberts; and Roger Kornberg doing a one-year stint as a junior fellow. (Another future Nobelist, Sid Altman, was on the fourth floor.) To briefly recount two stories from this era, Rich Roberts, then, like me, in the Strominger laboratory, set out to sequence a tRNA involved in bacterial cell wall synthesis, becoming one of the few individuals in the Boston area in the early 1970s who could sequence RNA. This made him attractive to Jim Watson, who invited Roberts to join him at the Cold Spring Harbor Laboratory (CSHL), where Watson was moving to become the director. At CSHL, Roberts made the startling finding that the 5Ј-ends of adenovirus (late) mRNAs were all the same; splicing had joined a common 5Ј-sequence to the coding sequence for different virion proteins (4). This extraordinary discovery earned Roberts a Nobel Prize, and it all began with his attempting to sequence a bacterial tRNA! Fast forward to the late 1980s, when graduate student Craig Mello was studying extrachromosomal DNA arrays in Caenorhabditis elegans with assistant professors Victor Ambros and Dan Stinchcomb. At the time, Ambros was discovering the first microRNA (5), a remarkable irony given that microRNAs act in the same manner as the siRNAs that Mello (and Andy Fire) would later discover and develop into a Nobel Prize-winning tool for silencing genes in the nematode (6).

Alternative Phage Sigma Factors and My Romance with Bacillus subtilis
I came to Harvard with the intention of studying the interaction of phage T4 with the E. coli membrane. I was even awarded a National Science Foundation grant to pursue this project, but geography and history altered everything! Down the hall, Dick Burgess and Andrew Travers had discovered that in E. coli, a subunit of RNA polymerase (called sigma) dictated promoter recognition. This suggested that there might be alternative sigma factors driving the expression of alternative gene sets (7). Meanwhile, my friend and dorm mate from Princeton and housemate in Cambridge, Linc Sonenshein, who was also a classmate at MIT, was studying ⌽e, a phage of the endospore-forming bacterium B. subtilis, in the Luria laboratory ( Fig. 1). Piggybacking on the robust dormancy properties of the spore, ⌽e becomes trapped in the developing spore, shielding it from harsh environmental conditions, only to emerge when its unfortunate host attempts to return to vegetative growth upon germination. Sonen-shein and I continued to interact frequently after my move to Harvard, regaling each other with stories about B. subtilis and RNA polymerase, respectively. Sonenshein was the matchmaker who started my romance with B. subtilis.
Putting two and two together, Sonenshein and I postulated that the elaborate developmental process of endospore formation was an ideal candidate for a regulatory system that could be driven by alternative sigma factors. To begin to investigate this notion, we grew vegetative and sporulating cells of B. subtilis at MIT and took the centrifuge bottles with the cell pellets to Harvard's Biological Laboratories, where we would break the cells open and isolate RNA polymerase using the methods of Burgess et al. (7). (On one occasion, Matt Meselson saw us leaving the Biological Laboratories with the empty bottles and accused us of stealing Harvard equipment. We showed him Luria's name embossed on the metal bottles.) We were thrilled to be able to demonstrate that the template specificity of RNA polymerase is altered during the transition from vegetative growth to sporulation (8). Sonenshein and I were right to focus on RNA polymerase, as became clear in the ensuing years, but our focus was premature. The discovery of alternative sigma factors had to await the invention of DNA cloning to provide templates for assaying promoter-specific transcription. Later, after joining the Harvard faculty, I was able to recruit outstanding graduate students, including Arno Greenleaf, Tom Linn, Rosalind Shorenstein, and Robert Tjian, who purified the housekeeping sigma factor A from vegetative cells, found that A was replaced by other RNA polymerase-binding proteins during sporulation, and demon-strated that sporulation involved an elaborate program of protein synthesis (9 -12).
The search for alternative sigma factors was successful because Watson's graduate student Jan Pero joined the project. (She also became my wife and life partner!) Jan had become an expert on gene regulation based on her studies of phage in the Watson laboratory. I told her about my collaboration with Sonenshein, whom we followed to the University of Paris-Sud in Orsay, France, where he was doing postdoctoral research with the pioneering sporulation geneticist Pierre Schaeffer (and where I met Patrick Stragier, who would later become a close collaborator) (Fig. 2). Upon our return to Cambridge, Pero decided to focus on the B. subtilis phage SP01, a close relative of ⌽e. Thanks to the work of Shunzo Okubo at the Osaka University Medical School, who kindly sent us key mutants from Japan, and of E. Peter Geiduschek at the University of Chicago and later at the University of California, San Diego, phage SP01 was known to exhibit a temporal pattern of gene expression under the control of phage regulatory genes (13,14).
With Jack Strominger championing my cause, I was at that point appointed an assistant professor in what was then the Department of Biology. Pero was joined on the SP01 project by my graduate students Tom Fox and Tjian (who had switched from sporulation to the phage). Pero and Fox discovered that the transcription of phage early genes was directed by the host's RNA polymerase containing the A factor and that successive expression of phage middle and late genes was driven by alternative, phageencoded sigma factors. One of the early genes (28) specified an alternative sigma factor specific for phage middle genes, two of which (33 and 34) reprogrammed core RNA polymerase to transcribe late genes. In one of the most breathtaking scientific periods of my career, Tjian and Pero reconstituted phage late transcription with the products of phage genes 33 and 34 (15). Fox proved that RNA polymerase-associated phage proteins were the products of genes 28, 33, and 34 by use of nonsense suppression to generate gene products with altered isoelectric points (16). These findings were the first direct demonstration in any organism of promoter-specific transcription directed by alternative sigma factors. Pero joined the faculty and continued the SP01 work in her own laboratory, where she discovered that phage promoters recognized by the phage-modified forms of RNA polymerase exhibited conserved sequences centered about 10 (Ϫ10) and 35 (Ϫ35) bp upstream of the transcription start site. These Ϫ10 and Ϫ35 sequences were distinct from the corresponding elements characteristic of promoters recognized by the A -containing host RNA polymerase. This led to the proposal, which proved to be correct, that sigma factors work by directly contacting and recognizing cognate sequences in promoter Ϫ10 and Ϫ35 regions (17).
In parallel with the study of SP01, graduate student Steve Clark tackled another B. subtilis phage called PBS2. Instead of modifying the host RNA polymerase, PBS2 encodes its own multisubunit RNA polymerase (18).

Alternative Bacterial Sigma Factors
The discovery of phage-encoded alternative sigma factors was a powerful incentive to return to the question of how B. subtilis turns on expression of different genes at different times during sporulation. The discovery of a family of stage-specific and compartment-specific sigma factors became the core of the genetic network controlling sporulation that my laboratory was to decipher. The development of recombinant DNA methods in the early to mid-1970s enabled my students Jacqueline Segall, Michelle Igo, and Frank Ollington and postdoctoral fellows Peter Zuber and Mike Stephens to clone DNA templates with individual bacterial promoters and use the cloned DNAs to study the regulation of individual genes (19 -23). This made possible the discovery of the first alternative bacterial sigma factor (then called 37 and later B ) in 1979 by my postdoctoral fellow Bill Haldenwang (24,25). The B factor turned out not to be a sporulation sigma factor but rather a stress response sigma factor, as demonstrated in my laboratory by Craig Binnie, Mary Lampe, and Igo (26 -28) and by Price and co-workers (29). Shortly after the discovery of B in 1981, Haldenwang and Naomi Lang-Unnasch uncovered the first sporulation-specific sigma factor, then called 29 and later called E (30 -32). The ensuing decade saw the discovery of what turned out to be the four remaining sporulation sigma factors through work from my laboratory (33) as well as from the laboratories of my former postdoctoral fellow Charles Moran, my collaborator Stragier, Peter Setlow, and Issar Smith (34 -36). Thus, five sigma factors in total ( E , F , G , H , and K ) were responsible for the program of gene expression that transformed a growing cell into a dormant spore (and with the surprise that the gene for K was a composite of two partial coding sequences that were fused by excision of a large segment of the chromosome during sporulation (37)).
Just as the phage sigma factors fit neatly into a simple hierarchical regulatory cascade ( A 3 SP01-28 3 SP01-33,34 ), so too the sporulation regulatory proteins fell into a linear dependent sequence ( H 3 F 3 E 3 G 3 K ) but with an exciting spatial twist. Spore formation takes place not in a single cell, but rather in a sporangium. After a process of asymmetric division, the sporangium consists of two cellular compartments known as the forespore and the mother cell. Thus, we had to consider not only the time of appearance of each of the five sigma factors but also their location over the course of development. It turned out that H is present in the predivisional sporangium, but the remaining four are compartment-specific transcription factors: F and G are active in the forespore, and E and K are active in the mother cell (as demonstrated by Adam Driks, Liz Harry, Peter Margolis, and Kit Pogliano) (38 -40). Moreover, E activity in the mother cell is dependent on F in the forespore; G activity in the forespore is tied to activation of E in the mother cell; and activation of K in the mother cell is dependent on G . Stragier and I called this pattern of intercompartmental control "crisscross regulation" (Fig. 3) (41). How did crisscross regulation work? It turned out that both E and K are initially produced as inactive proproteins (pro-E and pro-K ) and that conversion to the active sigma factors is controlled by analogous (but, remarkably, non-homologous!) intercompartmental signal transduction pathways under the control of F and G , respectively (42)(43)(44). Astonishingly, the pro-K -processing enzyme (SpoIVFB) would later turn out to be the founding member of a family of membraneembedded metalloproteases that includes site-2 proteases (45), which, as shown by Michael Brown and Joseph Goldstein at the University of Texas Southwestern Medical Center (46), are proprotein-processing enzymes for mam-malian transcription factors (sterol regulatory elementbinding proteins) involved in cholesterol biosynthesis. The K pathway was worked out by postdoctoral fellows Simon Cutting, Lee Kroos, Valerie Oke, Orna Resnekov, and David Rudner (43,(47)(48)(49)(50). Kroos and Rudner continue to make outstanding contributions to our understanding of the signaling pathways linking the forespore to the mother cell in their own laboratories in Michigan State University and Harvard Medical School, respectively. Meanwhile, Amy Camp in my laboratory (51,52) and Moran in his own laboratory at Emory University (53) determined that G is linked to E by a pathway involving a multiprotein channel.
Adding more intricacy to the sporulation circuitry was the further discovery of sporulation-specific DNA-binding proteins (33, 54 -58). Thus, the mother cell line of gene expression involves the action of the DNA-binding proteins SpoIIID and GerE, creating a hierarchical regulatory cascade of the form E 3 SpoIIID 3 K 3 GerE.
The final and arguably the most important challenge was and is to understand how crisscross regulation gets started, i.e. how F is activated selectively in the forespore. Unlike other microbes (such as Caulobacter) that pass on asymmetry from generation to generation, B. subtilis generates asymmetry de novo when it switches from binary fission to forming a septum near one pole of the cell (which pole is chosen is stochastic) at the start of sporulation (59). Graduate students Ruth Schmidt, Len Duncan, Danielle Garsin, and Scott Alper discovered that an anti-sigma factor (one of the first examples) and an anti-anti-sigma factor controlled the activity of F (60 -63, 104). Meanwhile, Michael Yudkin discovered that the anti-sigma factor is also a serine kinase that inactivates the anti-anti-sigma factor (64).
The key to the story proved to be a third protein, SpoIIE. Graduate student Peter Margolis discovered that SpoIIE heads the regulatory cascade that triggers the activation of F (38). How it does this became the subject of an exciting collaboration between Patrick Stragier and his student Fabrizio Arigoni, who came to my laboratory, and my student Len Duncan and postdoctoral fellow Kit Pogliano. Together, we found that SpoIIE dephosphorylates the anti-anti-sigma factor, allowing F to escape from the anti-sigma factor (65). We also discovered (see below for the story of how we accomplished this feat) that SpoIIE localizes to the septum after asymmetric division (66). Thus, a stochastic process of asymmetric division delivers SpoIIE near the pole that will house the forespore and in which F will become activated. In subsequent work, graduate students Karen Carniol and Nicole King dissected SpoIIE genetically, providing additional valuable insights (67)(68)(69). The final story on just how SpoIIE brings about cell-specific activation of F is still to be written! Finally, the full picture of how a cascade of sigma factors drives the program of sporulation gene expression would not have been possible without the parallel efforts of others in my laboratory who developed powerful molecular genetic tools and approaches to identify and clone developmentally regulated genes in the 1980s. Phil Youngman, John Perkins, and Kathleen Sandman invented transposon-based vectors for identifying and studying the regulation of sporulation genes (70 -73), and Bill Donovan and Biao Zheng used reverse genetic approaches to discover and clone the genes for coat proteins that envelop the dormant spore (74,75). These genes became the tools that they and others used to decipher the program of late sporulation gene expression (74 -78).

"Uncolis" and the Birth of Microbial Development
When I entered the sporulation field, my only goal was to discover alternative sigma factors involved in spore formation, but sporulation turned out to be far more intricate and fascinating than I had anticipated. During this period, I became good friends with Lucy Shapiro (then at Albert Einstein College of Medicine and later at Stanford University), who, in parallel, was tackling the differentiation cycle of the dimorphic bacterium Caulobacter crescentus. In a scientific world dominated by E. coli, we were the uncolis (a name inspired by the then popular 7UP Uncola commercial). Lucy and I (with Amar Klar, then at CSHL) organized a meeting on microbial development at CSHL during the summer of 1983, bringing together colleagues, including Dale Kaiser at Stanford University and Keith Chater at the John Innes Centre in the United Kingdom (Fig. 4), working on Myxococcus xanthus, Streptomyces coelicolor, and other uncolis. This helped transform the subject into a field and led to a monograph on microbial development (79). Here is why: not only does B. subtilis offer the best system for doing experimental genetics due to its natural ability to be transformed by DNA (genetic competence), but, as detailed below, it is remarkably rich in its biology, much more so than the dominant experimental organism of the day. Over the years, colleagues and departments have come and gone (biology became cellular and developmental biology and, later, molecular and cellular biology), but I have not budged from the third floor of the Biological Laboratories nor waned in my fascination with the same bacterium.

Green Fluorescent Protein and Bacterial Cell Biology
One day in the early 1990s, I had a visit from Marty Chalfie, a former Harvard undergraduate who was giving a departmental seminar. He mentioned his work with green fluorescent protein (GFP), which he was not going to talk about in his seminar. I almost fell out of my seat, realizing almost immediately that just as DNA cloning had changed my scientific life, so too would this remarkable tool lead me to redesign how we studied bacteria. Just a year after Chalfie's classic 1994 publication with Prasher in Science (80), we (Fabrizio Arigoni, Kit Pogliano, Stragier, Aurelio Teleman, and Chris Webb) reported visualizing forespore-and mother cell-specific gene expression in living cells and, by creating a protein-GFP fusion, visualizing the location of specific proteins within the bacterial cell (66,(81)(82)(83). This was significant because the traditional view of the bacterial cell was that of an amorphous vessel in which proteins freely diffused, but we could show that individual proteins had their own distinctive subcellular addresses. The use of GFP enabled us to discover that the protein phosphatase SpoIIE (see above) at the head of the pathway that governs the activation of F in the forespore localizes to the asymmetrically positioned division septum that creates the forespore compartment (66). GFP also enabled us to visualize the location within the cell of specific sites on the bacterial chromosome, such as the origin and the terminus (Chris Webb and Aurelio Teleman) (82), and to make the discovery that, during entry into sporulation, a protein (RacA) collapses the two chromosomes into a filament and tethers their origins to the extreme opposite poles of the cell (Sigal Ben-Yehuda) (83). GFP helped bring bacteria into the world of cell biology, and I was lucky enough to have been there at the beginning.

Back to the Wild
B. subtilis continued to surprise me with its seemingly unending biological virtuosity. I assumed that sporulation was a unicellular process, but postdoctoral fellow Alan Grossman discovered that spore formation is stimulated at high cell population densities by a self-produced extracellular factor (84). The discovery of genes encoding spore coat proteins opened up the new (to me) realm of morphogenesis, which in turn led to the discoveries by postdoctoral fellow Kumaran Ramamurthi of protein localization by recognition of a geometric cue (positive curvature) and of a process of self-assembly driven irreversibly by ATP hydrolysis (85,86).
Also unexpected was the discovery that B. subtilis is both fratricidal and cannibalistic! While searching for genes needed for spore formation, José Eduardo González-Pastor, now at the Instituto Nacional de Técnica Aeroespacial, and Errett Hobbs, now at the Lawrence Livermore National Laboratory, unexpectedly encountered genes that, when mutated, accelerate (rather than impair) sporulation (87). What emerged is that cells entering the sporulation process produce toxins that kill nonsporulating siblings and then feed on the nutrients thereby released. Spore formation is a stress response of last resort, and B. subtilis delays committing itself to this elaborate and energetically costly process for as long as possible by cannibalizing its siblings.
More surprises emerged when we realized that many decades of propagation in the laboratory had domesticated B. subtilis, causing it to shed some of its most alluring features. One example is swarming. Laboratory strains of B. subtilis are capable of swimming in liquid medium or on soft agar plates, but on plates with concentrations of agar too high to allow swimming, wild (but not laboratory) strains are able to scoot over the surface and do so by forming large hyperflagellated cells that cluster together in raft-like motile assemblages (88). Once again, new biology led to finding new genes. Postdoctoral fellow Dan Kearns found that laboratory strains harbor a frameshift mutation in a previously uncharacterized gene called swrA, needed for swarming motility but not swimming (89).
Even more striking, wild (but not most laboratory) strains are capable of forming architecturally complex communities (biofilms) on solid and liquid surfaces, a discovery made in close collaboration with my good friend Roberto Kolter, also at Harvard. Indeed, the architecture of these communities is so beautiful and alluring that it was simply irresistible to ask how B. subtilis was able to organize itself in this way. We found that the cells in these communities are in the form of long chains that are held together in parallel by an extracellular matrix consisting of exopolysaccharide and an amyloid-like protein. Uncovering this new biology enabled our joint team of students and postdoctoral fellows (Steve Branda, Win Chai, Frances Chu, Dani Lopez, Dan Kearns, Anna McLoon, Hera Vlamakis, and others) to discover a plethora of previously uncharacterized or misunderstood genes dedicated to biofilm formation (90 -98). Among these are the regulatory genes sinI, sinR, and slrR (about which I will have more to say); the three-gene operon for amyloid fiber production; and the 15-gene operon for exopolysaccharide production.
Interestingly, wild strains of B. subtilis and related spore-forming bacteria are used in the United States, China, and other countries as plant protectants, forming biofilms on the roots of plants. These biofilms promote plant growth and protect plants from various pathogens. Armed with our mutants, we showed, together with collaborators in Nanjing (who hosted me for a riveting trip to China), that the ability to adhere to roots and protect plants from a bacterial pathogen depends on biofilm formation (99).
Nothing is more satisfying to me in science than discovering and uncovering functions for genes involved in a previously unknown or unexplored biological process. The discoveries of cannibalism, swarming, and biofilm formation opened unexpected windows into the genome of B. subtilis, revealing previously unrecognized functions for scores of genes.

Chance and Determinism
Something odd stood out about B. subtilis from my earliest days of looking at it under the microscope. Unlike uniform-looking E. coli cells, B. subtilis cells were a mixture of solitary motile bacteria and long chains of sessile cells. I could have been accused of not having mastered pure culture techniques, but without question, B. subtilis grew as two cell types and did so when examined at the mid-exponential phase of growth. It was as though, even when maintained under seemingly optimal environmental conditions, B. subtilis was capable of switching spontaneously between alternative states. Fast-forward to the advent of fluorescent reporters for visualizing gene expression in individual living cells and a great deal of genetics and molecular biology carried out by Chai, Kearns and Tom Norman, it emerged that switching between the two states was largely governed by a simple circuit consisting of SinI, SinR, and SlrR (100,101). Chaining results from repression of genes for cell wall-degrading enzymes (autolysins) that release daughter cells from each other following cytokinesis. Motile cells express autolysin genes and genes for flagellum biosynthesis, whereas chains are repressed for autolysin and motility genes but instead express biofilm matrix genes. We therefore surmised that the chains represented a trial period for biofilm formation, whereas the motile cells were nomads on the prowl for new favorable niches. B. subtilis was hedging its bets by continuously producing two cell types! I began in science at a time when gene regulation in biology was assumed to be deterministic, i.e. genes were turned on or off in a rigid fashion as a direct response to environmental or developmental signals. Indeed, Einstein had famously written that "I, at any rate, am convinced that He does not throw dice," referring to his dismissal of Heisenberg's Uncertainty Principle. Now, it is well accepted that stochastic events sometimes play an important role in regulatory and developmental biology (witness the calico cat). Thus, it seemed that the motile/chaining switch might be just such a stochastic process even when freed of all environmental constraints. To test this, I joined forces with systems biologist Johan Paulsson at Harvard University, his student Nate Lord, and our shared student, Tom Norman. Norman and Lord built a microfluidic device that made it possible to visualize switching (using fluorescent reporters) under exceptionally constant conditions for remarkably long periods of time. Indeed, motile/chaining switching is noise-driven, and we were able to understand how it works in terms of simple mathematical principles (102). We soon convinced ourselves that sometimes, at least, She does throw dice!

Being a Teacher of Science
Just the other day and out of the blue, I heard from someone who had worked in my laboratory twenty-five years ago during her undergraduate years. She wrote, "The time I spent in your lab was truly the most rewarding years I had at Harvard. The skills I learned about presenting my research have helped me every time I need to give a lec-ture . . . Still a part of the academic world, I spend a lot of time teaching the medical students and residents. And, I realize now how special the way you ran your lab so that all students felt welcome and part of your team. Every day I continue to strive to make my interactions with students the same as you did." Being a teacher and mentor has been every bit as rewarding to me as being a scientist. Notes like this remind me why I do what I do, and, to come full circle, how important it was fifty years ago that I marched into the office of my undergraduate teacher and mentor, Charles Gilvarg.
I grew up in an era in which teaching science and doing science were often seen as being in competition with each other (happily, this is much less so the case now) (103). On the contrary, not only am I a better teacher for being a scientist, but I believe I am a better scientist for being a teacher. Armed with generous and sustained funding from the Professors Program of the Howard Hughes Medical Institute and from Harvard and together with like-minded colleagues Rob Lue, Tom Torello, and Alain Viel, I was able to launch hands-on teaching facilities in which students get to experience "ownership and discovery" in science and a program that fulfils my passion for helping to diversify science, a payback to my inspiring high school science teacher, Casper Hill, himself an underrepresented minority group member. As important as the research findings of my students and postdoctoral fellows have been, the greatest long-term impact I have had as a scientist is as a teacher and mentor, thereby instilling in my students and trainees a love of scientific thinking and discovery.