Global Analysis of the Genetic Network Controlling a Bacterial Cell Cycle Michael T. Laub, Harley H. McAdams, Tamara Feldblyum, Claire M. Fraser, and Lucy Shapiro

Supplemental Table 1. Previously known Caulobacter cell cycle-regulated genes. For each Caulobacter gene characterized before the publication of this report, the cell type in which maximal expression is observed is listed (SW = swarmer, ST = stalk, PD = predivisional) along with the corresponding reference. Additionally, for flagellar genes, the temporal class (I, II, III, or IV) is listed. Alternative names for genes are listed in parentheses. Of the 74 genes listed, all but alkB and tacA were represented on the microarrays printed for this study. The 72 genes on the microarray and listed here as previously identified cell cycle-regulated genes were all identified in this study as cell cycle-regulated mRNAs, with expression timing closely matching that published previously.
Gene	Peak expression*	Reference
alkB	ST	(1)
ccrM	late PD	(2)
cheA	late PD	(3)
cheB	late PD	(3)
cheD	late PD	(3)
cheR	late PD	(3)
cheW	late PD	(3)
cheYI	late PD	(3)
cheYII	late PD	(3)
cheYIII	late PD	(3)
ctrA	ST/class I	(4)
divJ	SW	(5)
divK	late PD	(6)
dnaA	late PD, SW	(7)
dnaJ	late PD, SW	(8)
dnaK	late PD, SW	(8)
dnaN	ST	(9)
dnaX	ST	(10)
flaA1 (flmA)	early PD/class III	(11)
flaA2 (flmB)	early PD/class III	(11)
flaD	early PD/class III	(11)
flaF	early PD/class III	(11)
flaG (flmH)	early PD/class III	(11)
flaN (flgK)	early PD/class III	(11)
flaZ (flmE)	early PD/class III	(11)
flbA (flmG)	early PD/class III	(11)
flbD	early PD/class III	(11)
flbE	early PD/class III	(11)
flbF (flhA)	early PD/class II	(11)
flbG	early PD/class III	(11)
flbH (flgD)	early PD/class III	(11)
flbT	early PD/class III	(11)
flbY	early PD/class III	(11)
flaE	early PD/class III	(11)
fliY	early PD/class III	(11)
flgE	early PD/class III	(11)
flgF	early PD/class III	(11)
flgG	early PD/class III	(11)
flgH	early PD/class III	(11)
flgI	early PD/class III	(11)
fliF	early PD/class II	(11)
fliG	early PD/class II	(11)
fliI	early PD/class II	(11)
fliJ	early PD/class II	(11)
fliL	early PD/class II	(11)
fliM	early PD/class II	(11)
fliN	early PD/class II	(11)
fliP	early PD/class II	(11)
fliQ	early PD/class II	(11)
fliR	early PD/class II	(11)
fliX	early PD/class II	(11)
fljJ	late PD/class IV	(11)
fljK	late PD/class IV	(11)
fljL	late PD/class IV	(11)
fljM	late PD/class IV	(11)
fljN	late PD/class IV	(11)
fljO	late PD/class IV	(11)
ftsA	early PD	(12)
ftsQ	early PD	(12)
ftsZ	stalk	(13)
groEL	late PD, SW	(14)
groES	late PD, SW	(14)
gyrB	ST	(9, 15)
hemE	ST	(16)
hfaA	late PD	(17)
mcpA	late PD	(3)
parA	late PD	(18)
parB	late PD	(18)
parC	ST	(19)
parE	ST	(19)
pilA	late PD, SW	(20)
podJ	ST	(21)
rpoN	early PD	(22)
tacA	early PD	(23)
*For most genes (see individual references for details and exceptions), their identification as cell cycle-regulated is based on analysis of the gene's (or operon's) promoter fused to lacZ. Immunoprecipitation of b-galactosidase at cell cycle time points allows tracking of promoter activity during cell cycle progression. The timing ascertained by b-gal assay may differ slightly from those obtained by microarray analysis, but maximal time of expression and overall pattern matches for each gene listed above.
For flagellar genes (classes II, III, and IV), the expression patterns are summarized and individually referenced in (11).

Supplemental Figure 1. . An expanded, annotated version of Fig. 1 of the report. Clustered expression profiles for the 553 identified cell cycle-regulated transcripts are organized by time of peak expression. Profiles for genes are in rows with temporal progression through the cell cycle running from left to right. Ratios for each time point are represented using the color scale at the bottom of the figure. Clustering was done using the self-organizing map analysis of the GeneCluster software (24) and plotted using TreeView software (25). Each cluster is numbered; gene names, TIGR ORF numbers, and functional annotation information are listed on the right of the clustergram.

Medium version | Full size version

Supplemental Figure 2a. Expanded, annotated versions of Fig. 2, A to C, of the report, with expression profiles for functionally related sets of genes. For each cell cycle-regulated function, the expression profiles of associated genes are shown as color bars, coded as in Web figure 1. The class I-IV flagellar genes shown in (C) are expressed in temporal order consistent with their known ordering in flagella construction. The column labeled ctrA^ts shows the change in expression level for each gene in response to loss of CtrA function where blue, black, and yellow indicate decrease, no change, and increase, respectively. For each gene, the corresponding TIGR ORF number and one-line annotation information are listed to the right of its profile.

Medium version | Full size version

Supplemental Figure 2b.

Medium version | Full size version

Supplemental Figure 2c.

Medium version | Full size version

Supplemental Figure 3. An expanded, detailed version of Fig. 3 of the report. The CtrA regulatory network governing Caulobacter cell cycle progression with newly identified pathways shown in red. Phosphorylated CtrA, binding to a conserved motif, autoregulates its own transcription (26), and activates the transcription of sets of genes required for flagellar biogenesis, DNA methylation, pili biogenesis, cell division, and six other operons of unknown function. CtrA also acts as a repressor of the newly identified G₁-S sigma factors sigT and sigU and, as previously suggested, of ftsZ (13). Binding of CtrA to sites in the origin of replication inhibits replication initiation (26). In addition, our results suggest there are at least 113 genes in a wide range of functional categories that appear to be indirectly regulated by CtrA; these include members of the chemotaxis machinery, ribosomal subunits, RNA polymerase subunits, and NADH dehydrogenase categories. CtrA is proteolyzed at the G¹-S transition by the protease ClpPX (27). The histidine kinases CckA and DivL mediate CtrA phosphorylation. DivK, DivJ, and PleC are all thought to be involved in a cell cycle phosphorelay [reviewed in (28, 29)].

Medium version | Full size version

Supplemental Figure 4. Expanded, annotated version of Fig. 4 of the report. Regulatory genes expressed in a cell cycle-dependent pattern. Genes encoding histidine kinases, response regulators, and sigma factors that are transcribed as a function of the cell cycle are shown at the time of their peak expression levels during the cell cycle. Newly identified regulatory genes are denoted by their predicted type and numbered based on their order of expression in the cell cycle: black, response regulators with an identifiable output domain; green, single-domain response regulators; red, histidine kinases; blue, histidine kinases with a fused response regulator domain; and orange (, sigma factors. Underlined genes have been characterized previously. The genes in black boxes encode proteins that are known to be dynamically localized during the cell cycle. TIGR ORF numbers are listed for each gene shown.

Medium version | Full size version

Supplemental Figure 5. Distribution of cell cycle-regulated genes within functional categories. The number of genes in each category is in parentheses. Genes were assigned to a functional category based on the classification scheme defined by Clusters of Orthologous Genes (COGs) (30).

Medium version | Full size version

METHODS

Wild-Type Time Course Methods
A culture of wild-type Caulobacter (CB15N) was grown in minimal media supplemented with glucose (M2G) at 28°C to an OD₆₆₀ of 0.3 and then synchronized as described (4). After resuspension of isolated swarmer cells in fresh M2G, cells were harvested at 15-min intervals for the duration of the 150-min cell cycle.

RNA from cells harvested at each time point was reverse-transcribed into Cy3-labeled cDNA, and competitively hybridized on a microarray with a Cy5-labeled reference cDNA from a mid-log phase asynchronous culture. Thus, subsequent log ratios (Cy3:Cy5) greater or less than zero indicate higher or lower RNA abundance, respectively, relative to the averaged RNA levels in the unsynchronized reference population. Details of RNA preparation, cDNA probe preparation, and microarray hybridization are at (31).

Identification of Cell Cycle-Regulated Transcripts / Discrete Cosine Transform Analysis
To identify cell cycle-dependent transcripts, the discrete cosine transform (DCT) (see devbio1.stanford.edu/usr/hm/Mathematica/Discrete Transforms/index.html) was calculated for each of the 2966 expression profiles with valid data. Three parameters were calculated for each profile: (i) the maximum DCT coefficient, (ii) the percentage of signal power contained in the first four DCT coefficients, and (iii) the peak-to-trough ratio of the profile. A cell cycle-dependent profile is expected to have a large maximum DCT coefficient, a significant percentage of total signal power in the lower frequency (e.g., the first four) DCT coefficients, and a high peak-to-trough ratio. We selected profiles for which parameters i, ii, and iii (above) were (a) 0.0017, 0.8, and 1.8, or (b) 0.0017, 0.6, 3.0, respectively. These parameters were chosen to minimize the false positive rate and to ensure inclusion of previously known cell cycle-regulated genes as described below. This procedure selected 579 candidate profiles.

As the synchronization procedure involved a shift from 4°C to 28°C at the zero time point, some genes responding to this temperature shift were anticipated in the 579 candidate profiles. These potential heat shock genes were identified by two criteria: (i) their profiles showed maximal expression at the zero time point followed by an immediate drop to minimal expression for the remainder of the time course, and (ii) their expression levels in an asynchronous population subjected to the same temperature shift were similar to the levels seen in the synchronized, swarmer cell population at the zero time point (data not shown). Genes that are expressed at a higher level in the synchronized population relative to the unsynchronized population after the same temperature shift are thus swarmer-specific and not heat shock-responsive. The original group of 579 candidate cell cycle-regulated genes contained 26 genes that responded solely to the synchronization method and were discarded; this led to a final group of 553 apparently cell cycle-regulated genes.

To estimate the number of false positives included in this set of 553 genes, the entire data set was randomly shuffled (2356 genes with 11 time points each = 25,916 randomly shuffled data points) and subjected to the same DCT analysis procedure. Only 46 of the 2356 randomized expression profiles exceeded the chosen parameters; accordingly, the false positive rate is estimated to be less than 9% (46/553). All genes shown by biochemical methods to be cell cycle-regulated were present in the set of genes identified as cell cycle-regulated. Visual inspection of the 553 individual profiles also confirmed the cyclical expression patterns of these genes.

Analysis of ctrA Mutant Strains
Wild-type and ctrA401^ts Caulobacter strains were grown at the permissive temperature of 28°C in peptone-yeast extract (PYE) media to an OD660 of 0.05 and then shifted to the restrictive temperature of 37°C. Cells were collected at 0 and 4 hours after the shift with total RNA harvested and treated as described in (31). RNA levels on microarrays were compared in three ways, where "mutant" represents the ctrA401^ts strain and "control" is the isogenic, wild-type strain: (a) control at 0 hours vs. control at 4 hours, (b) control at 0 hours vs. mutant at 0 hours, (c) mutant at 0 hours vs. mutant at 4 hours. Each of the three comparisons was done twice, using independent wild-type and ctrA401^ts colonies each time, with results averaged. The correlation coefficient for the duplicate experiments was greater than 95%. To identify genes whose expression changes significantly in response to the ctrA mutation, we multiplied the ratios of mutant at 4 hours/mutant at 0 hours (c) and mutant at 0 hours/control at 0 hours (b) to yield a ratio of mutant at 4 hours/control at 0 hours, which we termed the ctrA^ts ratio. This ratio should reflect changes in expression of a gene in the mutant strain relative to the wild type at either the permissive (b) or restrictive (c) temperature, or both. Genes whose ctrA^ts ratio showed at least a 1.75-fold change in either direction were considered as significant. However, those genes whose expression changed significantly in the control at 0 hours vs. control at 4 hours hybridization (a) were eliminated, since these are genes responding to the temperature shift alone and not to loss of CtrA function. The threshold of 1.75 was chosen because microarray-based comparison of RNA prepared from independent colonies of the same strain, grown under identical conditions, indicated that 100% of the 2966 spots analyzed showed less than twofold variability and 96% of spots varied less than 1.75-fold.

Strains LS3326 and LS3327 were grown overnight in PYE + 0.05% glucose, diluted back to an OD660 of 0.025 in PYE, grown to an OD₆₆₀ of 0.05, and then induced by addition of xylose to a final concentration of 0.05%. LS3327 contains the plasmid pJS71 with a xylose-inducible promoter driving expression of ctrAD51ED3W. This allele encodes a CtrA derivative with a C-terminal three-amino acid truncation rendering it resistant to normal, cell cycle-dependent proteolysis. In addition, a critical aspartate is replaced with a glutamate, D51E, that partially mimics constitutive phosphorylation. LS3326 contains the plasmid pJS71 with no insert as a control. RNA for microarray hybridization was harvested as described in (31) from cells collected at 0 and 4 hours after addition of xylose. As with ctrA401^ts, three comparisons were made with microarrays (each comparison was done twice, independently), where "mutant" represents the inducible ctrAD51ED3W strain and "control" is the isogenic parent strain: (d) control at 0 hours vs. control at 4 hours, (e) control at 0 hours vs. mutant at 0 hours, (f) mutant at 0 hours vs. mutant at 4 hours. Identification of significantly affected genes was done in the same manner as the ctrA401^ts experiment, but with a ratio of mutant at 4 hours/control at 0 hours calculated by multiplying ratios from hybridizations (e) and (f). This ratio summarizes changes resulting from induction of the CtrA allele (f) or leaky expression in repressive conditions (e). Genes that responded only to xylose addition to the media (d) were eliminated.

Promoter Analysis
A promoter database containing the 600 base pairs upstream of the predicted start codon of each predicted ORF was constructed. A profile of known CtrA binding sites was built using MEME (Multiple Expectation Maximization Motif Elicitation), available through the GCG software package (Genetics Computer Group, Madison, WI). This profile was then used with MotifSearch (GCG software) to identify candidate CtrA binding sites in the promoter database. CtrA binding sites were selected by requiring a combined MotifSearch P value of < 0.01 and a minimum 7 out of 9 match to the consensus CtrA binding site of TTAA-n7-TTAAC.

References

1. D. Colombi, S. L. Gomes, J. Bacteriol. 179, 3139 (1997).

2. G. Zweiger, G. Marczynski, L. Shapiro, J. Mol. Biol. 235, 472 (1994).

3. M. R. Alley, S. L. Gomes, W. Alexander, L. Shapiro, Genetics 129, 333 (1991).

4. K. C. Quon, G. T. Marczynski, L. Shapiro, Cell 84, 83 (1996).

5. N. Ohta, T. Lane, E. G. Ninfa, J. M. Sommer, A. Newton, Proc. Natl. Acad. Sci. U.S.A. 89, 10297 (1992).

6. G. B. Hecht, T. Lane, N. Ohta, J. M. Sommer, A. Newton, EMBO J. 14, 3915 (1995).

7. G. Zweiger, L. Shapiro, J. Bacteriol. 176, 401 (1994).

8. S. L. Gomes, J. W. Gober, L. Shapiro, J. Bacteriol. 172, 3051 (1990).

9. R. C. Roberts, L. Shapiro, J. Bacteriol. 179, 2319 (1997).

10. E. Winzeler, L. Shapiro, J. Mol. Biol. 264, 412 (1996).

11. J. W. Gober, J. C. England, in Prokaryotic Development, Y. V. Brun, L. J. Shimkets, Eds. (ASM Press, Washington, DC, 2000), pp. 319-339.

12. M. J. Sackett, A. J. Kelly, Y. V. Brun, Mol. Microbiol. 28, 421 (1998).

13. A. J. Kelly, M. J. Sackett, N. Din, E. Quardokus, Y. V. Brun, Genes Dev. 12, 880 (1998).

14. M. Avedissian, S. Lopes Gomes, Mol. Microbiol. 19, 79 (1996).

15. M. F. Rizzo, L. Shapiro, J. Gober, J. Bacteriol. 175, 6970 (1993).

16. G. T. Marczynski, L. Shapiro, J. Mol. Biol. 226, 959 (1992).

17. R. S. Janakiraman, Y. V. Brun, J. Bacteriol. 181, 1118 (1999).

18. D. A. Mohl, J. W. Gober, Cell 88, 675 (1997).

19. D. V. Ward, A. Newton, J. Bacteriol. 181, 3321 (1999).

20. J. M. Skerker, L. Shapiro, EMBO J. 19, 3223 (2000).

21. W. B. Crymes Jr., D. Zhang, B. Ely, J. Bacteriol. 181, 3967 (1999).

22. Y. V. Brun, L. Shapiro, Genes Dev. 6, 2395 (1992).

23. M. V. Marques, S. L. Gomes, J. W. Gober, J. Bacteriol. 179, 5502 (1997).

24. P. Tamayo et al., Proc. Natl. Acad. Sci. U.S.A. 96, 2907 (1999).

25. M. B. Eisen, P. T. Spellman, P. O. Brown, D. Botstein, Proc. Natl. Acad. Sci. U.S.A. 95, 14863 (1998).

26. I. J. Domian, A. Reisenauer, L. Shapiro, Proc. Natl. Acad. Sci. U.S.A. 96, 6648 (1999).

27. U. Jenal, T. Fuchs, EMBO J. 17, 5658 (1998).

28. D. Hung, H. McAdams, L. Shapiro, in Prokaryotic Development, Y. V. Brun, L. J. Shimkets, Eds. (ASM Press, Washington, DC, 2000), pp. 361-378.

29. N. Ohta, T. W. Grebe, A. Newton, in Prokaryotic Development, Y. V. Brun, L. J. Shimkets, Eds. (ASM Press, Washington, DC, 2000), pp. 341-359.

30. R. L. Tatusov, E. V. Koonin, D. J. Lipman, Science 278, 631 (1997).

31. For complete protocols, see http://caulobacter.stanford.edu/CellCycle.