Biased inheritance of the protein PatN frees vegetative cells to initiate patterned heterocyst differentiation

Abstract

Heterocysts, cells specialized for nitrogen fixation in certain filamentous cyanobacteria, appear singly in a nonrandom spacing pattern along the chain of vegetative cells. A two-stage, biased initiation and competitive resolution model has been proposed to explain the establishment of this spacing pattern. There is substantial evidence that competitive resolution of a subset of cells initiating differentiation occurs by interactions between a self-enhancing activator protein, HetR, and a diffusible pentapeptide inhibitor PatS-5 (RGSGR). Results presented here show that the absence of a unique membrane protein, PatN, in Nostoc punctiforme strain ATCC 29133 leads to a threefold increase in heterocyst frequency and a fourfold decrease in the vegetative cell interval between heterocysts. A PatN-GFP translational fusion shows a pattern of biased inheritance in daughter vegetative cells of ammonium-grown cultures. Inactivation of another heterocyst patterning gene, patA, is epistatic to inactivation of patN, and transcription of patA increases in a patN-deletion strain, implying that patN may function by modulating levels of patA. The presence of PatN is hypothesized to decrease the competency of a vegetative cell to initiate heterocyst differentiation, and the cellular concentration of PatN is dependent on cell division that results in cells transiently depleted of PatN. We suggest that biased inheritance of cell-fate determinants is a phylogenetic domain-spanning paradigm in the development of biological patterns.

Two broad strategies used in the development of biological patterns are activator–inhibitor systems, a refinement of Turing-type reaction–diffusion systems (1, 2), and differential inheritance of cell fate determinants (3, 4). These strategies have been analyzed in multicellular eukaryotes ranging in developmental complexity from hydra to vertebrates (1, 4). Some filamentous cyanobacteria show spaced patterns of differentiated cells, such as nitrogen-fixing heterocysts and spore-like akinetes, thereby representing the simplest and oldest evolutionary model for study of the development of biological patterns and multicellular interactions. There is strong experimental evidence that an activator–inhibitor system regulates heterocyst patterning in filamentous cyanobacteria (5–7). We report here the characteristics of a gene (patN) that is required for normal heterocyst patterning, and the phenotype and protein localization of which is consistent with a role in the biased initiation of heterocyst differentiation via biased inheritance.

Heterocysts terminally differentiate from vegetative cells in response to limitation of combined nitrogen. The heterocysts appear in a nonrandom spacing of approximately 1 heterocyst to every 10–15 vegetative cells. The activator–inhibitor system governing this patterning consists of a self-enhancing activator protein, HetR (8, 9), and its antagonist, proteins containing a pentapeptide, PatS-5 (RGSGR) (10, 11). These two primary positive and negative acting regulatory elements are modulated by several other elements (5, 6, 12, 13). Deletion of patS (10) and overexpression of hetR (9) each results in clusters of two or more heterocysts [multiple contiguous heterocysts (Mch)] forming at nonrandomly spaced sites in the filaments. The Mch phenotype is also observed in mutants with alterations of expression of other positive- and negative-acting elements (6). This phenotype supports a two-stage model (6) in which a subset of cells is biased to differentiate before nitrogen limitation (stage 1, biased initiation), and the subsets are then resolved to singly spaced heterocysts via HetR and PatS-5 interactions (stage 2, competitive resolution). According to this model, biased initiation and competitive resolution occur before cells show any morphological signs of differentiation (6), the most universally distinct of which is the loss of autofluorescence from photosynthetic pigments.

Results

Inactivation of patN.

patN was identified in a transposon mutagenesis screen for altered growth of the heterocyst-forming cyanobacterium Nostoc punctiforme strain ATCC 29133, with dinitrogen as the sole nitrogen source (14). The transposon had inserted into an ORF, NpF6624, which was designated patN. The insertion mutant, subsequently constructed deletion mutants, and the wild type show identical permissive phenotypes during growth in the presence of ammonium. By 24 h, and thereafter, upon removal of combined nitrogen (N step-down), a nonrandom pattern of fully developed heterocysts was seen in the wild-type strain (Fig. 1A). However, heterocyst differentiation was delayed by more than 24 h in strain UCD 524 (ΔpatN); cells with weak autofluorescence were seen occasionally 24 h after N step-down but not in an obvious spaced pattern in the filaments. A distinct pattern of auto-nonfluorescent cells characteristic of differentiating heterocysts was present at 48 h in strain UCD 524, although many of these cells lacked the structural properties of a mature heterocyst, as seen at 120 h (Fig. 1B). The most striking morphological characteristic of strain UCD 524 is the increased heterocyst frequency of about 30% compared with 8% in the wild-type strain, and the heterocysts appear singly [multiple singular heterocysts (Msh)] in the filaments. The Msh pattern has been observed only in plant symbiotically associated Nostoc spp. and the ΔpatN mutant (6). The increased heterocyst frequency in the mutant indicates PatN functions as a negative regulator of heterocyst differentiation. Based on both morphology and loss of cellular autofluorescence to score heterocysts, the average vegetative cell interval between heterocysts for strain UCD 524 at 48 h and 120 h was 3.88 (SD = 0.28; n = 90) and 2.68 (SD = 0.16; n = 90), respectively, vs. 13.13 at 48 h (SD = 1.32; n = 90) and 12.4 at 120 h (SD = 0.1; n = 90) for the wild type. Despite the increased frequency of evenly spaced heterocysts in the patN-deletion strain, the growth rate under diazotrophic conditions and the rate of nitrogen fixation were reduced compared with those of the wild type. (Fig. S1 B and C).

Fig. 1.

Phenotype of a patN-deletion strain. (A and B) Phase-contrast (Left) and fluorescence micrographs (Right) of phycobilisome-induced autofluorescence for the wild-type (A) and ΔpatN (B) strains at 48 and 120 h (as indicated); carets indicate heterocysts or loss of phycobilisome-induced autofluorescence, indicating site of a developing heterocyst; pairs of asterisks indicate pairs of contiguous cells that have lost phycobilisome-induced autofluorescence. (Scale bar: = 10 μm.) (C and D) Phase-contrast micrographs of complementation of the ΔpatN strain with plasmid pSCR500 (C), which contains the patN promoter and coding region, and overexpression of patN from the petE promoter on plasmid pSCR570 in the wild type (D); carets indicate heterocysts. (Scale bar: 10 μm.) (E and F) Phase-contrast micrographs of overexpression of patS from the petE promoter on plasmid pSCR576 in the wild-type (E) and ΔpatN (F) strains. (Scale bar: 10 μm.)

Pairs of adjacent, usually small cells, lacking autofluorescence were often observed in strain UCD 524 after N step-down (Fig. 1B, asterisk). Because rarely, if at all, were doublet heterocysts detected, whereas small-sized heterocysts could be seen by 120 h (Fig. 1B), we infer that one or both of these small cells must have regressed to the vegetative state as the other differentiated. Adjacent nonfluorescent vegetative cells were never observed in the wild type.

Inactivation of patS yields a Mch spacing pattern, whereas overexpression of it results in repression of differentiation in Anabaena sp. strain PCC 7120 (10). The following results are consistent with the presence of a functional patS-dependent pathway in wild-type and patN mutant strains of N. punctiforme. Instances of Mch, already rare in wild-type N. punctiforme, occurred at an even lower rate in ΔpatN strain UCD 524 (Fig. S1A), indicative of a PatS-like activity. Strain UCD 524 could be complemented in trans to the wild-type spacing pattern and time for heterocyst maturation with a high-copy plasmid containing the patN coding region under the control of its native promoter (Fig. 1C). Overexpression of patN from the copper-inducible petE promoter in a wild-type background had no apparent effect on the development of heterocyst spacing (Fig. 1D); the lack of a Mch pattern implies that PatS was functional in the overexpression strains as well. Both exogenous addition of a synthetic RGSGR peptide (PatS-5) at concentrations of 0.5–1.0 μM at the time of N step-down and overexpression of patS from the petE promoter (Fig. 1 E and F) completely suppressed heterocyst differentiation in the wild-type and ΔpatN strains. These results imply that other components of the PatS-dependent signaling pathway are present in these strains. Unlike PatS (Fig. 1F), greater-than-natural concentrations of PatN do not repress heterocyst differentiation (Fig. 1D); thus, it is possible that under natural conditions, PatN is saturating for its unknown mechanism in repression of the initiation of differentiation, if, in fact, the high-copy plasmid and P_petE enhanced expression resulted in more PatN protein.

Proteins highly similar to PatN were identified through BLAST (E < 10⁻⁴²) in seven heterocyst-forming cyanobacteria, the genome sequences of which have been essentially completed (Fig. S2A). In contrast, only 1 genome (Lyngbya majuscula) of the 10 filamentous non–heterocyst-forming cyanobacteria, the genomes of which are reported, has a protein with even weak similarity. No other organism has a protein with significant similarity to PatN. This limited distribution is quite different from the broad taxonomic distribution of HetR and PatS (15).

Biased Inheritance of PatN.

PatN appears to be an integral membrane protein containing both a classic signal peptide and a transmembrane α-helix (Fig. S2B). The N-terminal portion of PatN is predicted to reside in the periplasm and contains a glutamine-rich (26.2%) coiled-coil domain. No predicted features could be detected in the C-terminal cytoplasmic domain.

To confirm the membrane localization of PatN, the native allele of patN in the N. punctiforme chromosome was replaced with a patN-gfp translational fusion. The resulting strain, UCD 567, displayed an unusual pattern of fluorescence within the filaments. In cultures supplemented with ammonium, individual cells either contained little observable fluorescence, a halo of fluorescence around the entire cell, with PatN-GFP apparently localized to the cytoplasmic membrane, or a halo of fluorescence localized to only one-half of the cell (Fig. 2A). Time-lapse epifluorescence imaging confirmed a dynamic, systematic localization of PatN-GFP that progresses with each round of cell division (Fig. 2B and Fig. S3A). Before cell division, PatN-GFP localized to only one-half of the cell; thus, bisection of the cell at the division plane resulted in inheritance of detectable PatN-GFP to only one of the two daughter cells. Within 8 h following cell division, both cells had begun accumulating PatN-GFP in the half of the cell proximal to the most recently formed septum (Fig. 2B). We do not know how this cellular reorganization of the PatN-GFP proceeds; it is unlikely that the fusion protein is transferred, after septation, from the daughter cell inheriting it to the lacking daughter cell. The uneven cellular distribution of fluorescence cannot be attributed to differential transcription among cells of a filament because a P_patN-gfp transcriptional fusion showed equally distributed fluorescence in all cells (Fig. S3B). Nevertheless, this dynamic pattern of localization and asymmetrical inheritance of PatN-GFP was repeated with each subsequent round of cell division. Unfortunately, the PatN-GFP fusion was not functional, resulting in a Msh phenotype similar to that of the ΔpatN strain upon N step-down. All patN-gfp constructs, including use of an alternate, α-helical linker in strain UCD 575 and placement of the gfp coding region between the transmembrane and cytoplasmic domains in strain UCD 582, the latter of which was nonfluorescent, were dominant-negative, whether expressed from the chromosome with patN in trans on a multicopy plasmid or the converse. The cellular pattern of PatN-GFP fluorescence following N step-down was the same as ammonium-grown filaments; because the fusion protein was nonfunctional, the fluorescence pattern did not correspond to the site of heterocyst development.

Fig. 2.

Localization of PatN. (A) Fluorescence micrographs of (from top to bottom) phycobilisome-induced autofluorescence (red), PatN-GFP fluorescence (cyan), and merged images of phycobilisome-induced and PatN-GFP fluorescence. (Inset, Right) Magnification (4×) of area indicated by the white square. (Scale bar: 10 μm.) (B) Time-lapse microscopy of a PatN-GFP translational fusion in strain UCD 567. Fluorescence micrographs of (from left to right) phycobilisome-induced autofluorescence, PatN-GFP fluorescence, and merged images of phycobilisome-induced and PatN-GFP fluorescence. Images of the same filament cultured with ammonium were taken at 8-h intervals for 24 h as indicated. (Scale bar: 10 μm.) Complete frames from which the enlarged images in B were extracted appear in Fig. S3A and are indicated there by arrowheads.

patN Affects Transcription of patA and hetZ.

To assess the effect of patN deletion on the transcriptional program leading to heterocyst formation, a DNA microarray comparison of the transcriptomes for the wild-type and ΔpatN strain over the course of a N step-down was performed. The transcriptomes of both strains were defined at 0, 1, 3, 6, 12, 18, 24, and 120 h after N step-down, and comparisons were made between each strain at the corresponding time point (ΔpatN vs. wild-type comparison; for the gene list, see Dataset S1), as well as between each time point and t = 0 for the same strain (wild-type and ΔpatN time-course comparisons; see Datasets S2 and S3, respectively).

Within the first 6 h of N step-down, expression of two genes known to be involved in heterocyst formation and patterning, patA and hetZ, were significantly increased in the patN-deletion strain compared with the wild type (Fig. 3A). Expression of patA increased almost immediately after N step-down and showed ∼two- and fourfold increases at 3 and 6 h respectively, whereas expression of hetZ increased by twofold at 6 h and fourfold by 24 h. Transcription of patA is very low in Anabaena sp. strain PCC 7120 (16) and N. punctiforme (0.3% of rnpB; Fig. S4). To verify the DNA microarray results, quantitative (q)PCR of patA expression was performed on wild-type and strain UCD 524 RNA samples harvested 6 h after N step-down (Fig. S4). The microarray and qPCR results showed similar 3.24- and 3.4-fold increases, respectively, in transcription in the patN mutant relative to the wild type. Previously identified genes involved in responding to nitrogen starvation and heterocyst development and patterning, including ntcA, nrrA, hetR, hetP, hetF, and patU, were not differentially transcribed between the wild-type and patN-deletion strain within the first 6 h of N step-down. Genes encoding proteins for uptake and metabolism of nitrate (nrtP, nirA, narB) and transport of urea (urtA, urtB, urtC, urtD) were highly up-regulated early in both wild-type and ΔpatN strains (Datasets S2 and S3), indicating similarities in the initial N-starvation response (17). Transcription of patN was unchanged over the course of a N step-down in the wild-type strain (Dataset S2), consistent with the results of a P_patN-gfp promoter fusion (Fig. S3B).

Fig. 3.

Microarray comparison of the wild-type and patN-deletion strain. (A and B) Heat maps showing change in transcription for DNA microarray comparison of the wild-type and the ΔpatN strain at each time point of a N step-down time course (ΔpatN vs. wild type) or the wild-type or ΔpatN strain compared with ammonium-grown cultures of the same strain (wild-type time course and ΔpatN time course, respectively). For the ΔpatN vs. wild-type comparison, asterisks indicate time points at which differential expression was considered statistically significant (B ≥ 0). Shown are a selection of heterocyst regulatory genes (A) and heterocyst structural genes (B). Color keys are included, with numerical values representing M values (log₂[experimental] − log₂[reference]). Time indicated is in hours. (C and D) Phase-contrast micrographs of the ΔpatA (C) and ΔpatN and ΔpatA double mutants (D) strains 120 h after N step-down; carets indicate heterocysts. (Scale bar: 10 μm.)

Between 6 and 24 h, the number of differentially transcribed genes in the wild-type relative to the ΔpatN strain greatly increased and appears to correspond to the delay in heterocyst development in the ΔpatN strain (Fig. 3 A and B). A number of heterocyst structural genes, including heterocyst envelope polysaccharide and glycolipid genes and genes comprising the nitrogenase enzyme complex, have decreased expression in the ΔpatN strain compared with the wild type between 6 and 24 h but reach wild-type levels by 120 h (Fig. 3B). A pair of previously uncharacterized genes, NpR1607 and NpR1606, also displayed differential transcription in the ΔpatN strain, but inactivation of either did not result in any obvious defects in heterocyst patterning (Fig. S5 A and B).

patA Is Epistatic to patN.

Results from the microarray comparison, confirmed by qPCR, implied that patN may modulate heterocyst patterning by affecting the expression of patA, a gene known to be involved in heterocyst patterning. In contrast to patN, inactivation of patA results in heterocysts primarily at the filament termini, with occasional intercalary heterocysts (16). Thus, PatN and PatA appear to have an antagonistic relationship; PatN increases and PatA decreases the number of vegetative cells between heterocysts. To test the hypothesis that PatN functions by modulating the level of PatA, epistasis analysis was performed with null alleles of patA and patN. A patA-deletion strain of N. punctiforme, strain UCD 561, formed heterocysts primarily at the filament termini, with occasional intercalary heterocysts (Fig. 3C), a phenotype similar to that observed for inactivation of patA in Anabaena sp. strain PCC 7120 (16). In strain UCD 558, which contains deletions of both patN and patA, the phenotype was similar to that of the patA-deletion strain, indicating that PatA is epistatic to PatN (Fig. 3D). This observation supports the hypothesis that PatN affects heterocyst patterning by limiting the expression of patA.

Discussion

We hypothesize that the concentration of PatN affects the competency of vegetative cells to initiate heterocyst differentiation. This scenario posits that before, or at, the signal of nitrogen starvation, certain cells in a filament are transiently more likely than others to initiate differentiation; it is only after this biased initiation that lateral inhibition is used to resolve differentiating cells in close proximity (Fig. 4 A and B). If PatN is involved in determining the differentiation competency, then the absence of PatN would result in all cells in a filament being equally competent to initiate differentiation at the time of N starvation. The biased inheritance of PatN provides a mechanism for determining this competency and implies that recently divided cells that have not inherited PatN are poised to differentiate. The model in Fig. 4, with the outermost immediately arising granddaughter cells devoid of PatN and poised to initiate differentiation, is compatible with the smaller daughter cell–division rule for heterocyst differentiation proposed some 40 y ago from astute observations in uniquely asymmetrically dividing Anabaena catenula by Mitchison and Wilcox (18).

Fig. 4.

Working model of PatN function. (A) With each round of cell division, PatN (blue lines) is partitioned in vegetative cells (green) to the half of the cell containing the most recently formed septum. Upon cell division, PatN is inherited primarily in one of the two daughter cells. This process is repeated with each subsequent round of cell division. The model depicts synchronous division to simplify presentation, although division is not normally synchronous in all cells of a filament. (B) Recently divided vegetative cells that have not inherited PatN (here depicted as arising from nonsynchronous division) are transiently poised to undergo heterocyst differentiation upon N step-down, establishing a rudimentary pattern that is resolved to a single heterocyst (yellow cell) at a site via lateral inhibition by PatS-5. (C) Mechanism for regulation of heterocyst differentiation pathway by PatN. PatN negatively regulates expression of patA, which has previously been demonstrated to be involved in a positive-feedback loop with hetR (19). Arrows indicate positive regulation; lines with bar indicate negative regulation.

The proposed two-stage, biased initiation–competitive resolution model for the initiation of heterocyst differentiation holds that all cells sense the depletion of combined nitrogen, most likely through NtcA but that only a subset of cells can response to the signal (6). Identical profiles of the early and continuously up-regulated transcription of genes for nitrate assimilation and urea transport in the wild-type and ΔpatN strains after N step-down (Datasets S2 and S3) confirm that the patN mutant senses and responds to the signal of N limitation, similar to the wild type (17). Excepting for NtcA, all other proteins currently identified as involved in the initiation of heterocyst differentiation (noted earlier) affect either competitive resolution or maintenance of pattern and not biased initiation. PatN is a protein possibly involved in the transient determination of which cells may or may not differentiate in cyanobacteria. We do not know the precise mechanism by which the biased inheritance of PatN frees or inhibits a cell from initiating differentiation. The evidence presented here implies that PatN affects the competency of a vegetative cell to differentiate into a heterocyst, at least in part, by limiting expression of patA (Fig. 4C). This is supported by the fact that deletion of patN results in increased transcription of patA, as well as the epistasis analysis indicating that the phenotype of a patN deletion strain requires a functional patA gene. Additional evidence in support of this model comes from a recent study in Anabaena sp. strain PCC 7120 demonstrating that overexpression of patA decreases the vegetative cell interval between heterocysts (19), resembling the phenotype of the ΔpatN strain. The effect of patN on transcription of patA is not likely to be direct, however, given that PatN is an integral membrane protein and contains no obvious DNA-binding domains.

The Msh pattern of the ΔpatN strain most likely results from competitive resolution. Competitive resolution in the context of lateral inhibition is proposed to be attributable to the production of HetR and PatS in a differentiating heterocyst, resistance of that cell to the effects of PatS, and diffusion of PatS-5 to adjacent vegetative cells (7). Because time is also a factor, such a model may account for the delay in heterocyst formation in the ΔpatN strain. Complementation analysis (Fig. 1 C and D) established that the lack of PatN was responsible for both the increased heterocyst frequency and the temporal delay in differentiation in the ΔpatN strain. Resolution of a cluster of differentiating cells to evenly spaced heterocysts may take longer when all cells in the filament are competing to differentiate and are in an unstable developmental state. This could additionally explain the late resolution of pairs of small auto-nonfluorescent cells into a heterocyst and a vegetative cell seen in the ΔpatN strain (Fig. 1), because these pairs likely arise from a mother cell simultaneously initiating heterocyst differentiation and cell division. In the wild type, cells close to division would be excluded from initiating differentiation by the presence of PatN. This role of PatN is also consistent with, and provides a mechanistic foundation for, studies implicating cell division and/or the cell cycle in regulating heterocyst differentiation (6, 20); cell division is required for the biased inheritance of PatN. Thus, it appears that biased inheritance of cell fate determinants is a paradigm in the development of multicellular patterns, spanning bacterial 1D patterns that evolved billions of years ago to multidimensional patterns in more recent plant and animal systems.

Experimental Procedures

Strains and Culture Conditions.

For a detailed description of the plasmids, strains, and oligonucleotides used in this study, and real-time qPCR analysis, refer to SI Experimental Procedures, Tables S1 and S2, and Fig. S4. N. punctiforme strain UCD 153 and its derivatives were cultured in Allan and Arnon medium diluted fourfold (AA/4), as described previously (21), with the exception that cultures were supplemented with 30 mM fructose to suppress hormogonium differentiation (22). For nitrogen step-down, cultures supplemented with 2.5 mM NH₄Cl at a chlorophyll (Chl) a concentration of 2–3 μg/mL were harvested at 1,000 × g for 5 min, washed three times with AA/4, and suspended in AA/4 (21). For selective growth, the medium was supplemented with 25 μg/mL neomycin. For induction of the copper-inducible petE promoter, cultures were supplemented with 2.5 μM CuSO₄. Escherichia coli cultures were grown in Luria–Bertani (LB) broth for liquid cultures or LB supplemented with 1.5% (wt/vol) agar for plates. Selective growth media were supplemented with 50 μg/mL kanamycin, 50 μg/mL ampicillin, and 10 μg/mL chloramphenicol.

Microscopy.

All images were taken using a 100× lens on a Nikon Eclipse 80i microscope equipped with a QImaging Micropublisher 3.3 RTV camera using Q-Capture pro 5.0 software (QImaging). For fluorescence images, an X-cite 120 Fluorescence Illumination System (Lumen Dynamics) was used with the filter set G-2A (excitation, 510–560 nm; emission, 590 nm long pass) (Nikon) for phycobilisome-induced autofluorescence and a custom filter set [excitation, D380/30x; dichromatic mirror, 420 dichroic long pass (dclp); emission, D510/40m] (Chroma) for GFP.

For time-lapse microscopy, a grid was first hand drawn onto a coverslip to facilitate repeated viewing of the same field. The coverslip was inverted, and 100 μL of growth media containing 0.5% melted agar was spotted onto the coverslip and allowed to solidify. Five microliters of a liquid culture of the appropriate strain were spotted on this agar pad. Once standing liquid on the pad had absorbed into the agar or evaporated, the coverslip was inverted and placed in a glass-bottom culture dish (MatTek; part no. P35G-0-14-C), so that the bacteria were sandwiched between the glass bottom of the culture dish and the agar pad on the coverslip. Melted agar was then spotted on the side of the coverslip to anchor it to the culture dish, so it would not move during the course of the experiment. One milliliter of water was added to the lid to prevent the agar from drying out, and the culture dish was subsequently closed, sealed with parafilm, and incubated under standard growth conditions, with images of the same filament taken every 8 h.

Microarray Experiments.

For DNA microarray studies, two 50-mL cultures of strains UCD 153 or UCD 524 at 2–3 μg of Chl a/mL were harvested at t = 0 (supplemented with 2.5 mM NH₄Cl) and 1, 3, 6, 12, 18, 24, and 120 h after N step-down (n = 3). RNA extraction was performed as described previously (21). Ten micrograms of total RNA were used for cDNA synthesis using SuperScript Double Stranded cDNA Synthesis Kit (Invitrogen). cDNA (0.5 μg) was labeled with Cy3 dye using the NimbleGen One-Color DNA Labeling Kit. Two micrograms of Cy3-labeled cDNA were prepared for hybridization using the NimbleGen Hybridization Kit and Sample Tracking Controls. Labeled cDNA was hybridized to a NimbleGen 12 × 135k slide fitted with a HX12 mixer at 42 °C on a NimbleGen Hybridization System 4 hybridizer for 16–20 h. Following hybridization, slides were washed with NimbleGen Wash Buffer and scanned with a GenePix 4000B scanner using GenePix Pro-6.0 software (Molecular Devices).

Scanned images (.tif) were imported into NimbleScan software (NimbleGen), where the image containing all 12 arrays on each slide was separated into 12 individual image files. An automatic alignment was performed for each individual array, and X-Y Signal (.xys) files were generated. Normalization of .xys files was performed using the R package oligo (23), and the R package oneChannelGUI (24) was used to compute a linear model fit and contrasts to generate top tables containing the change in expression between experimental and reference data points as M values (log₂[experimental] − log₂[reference]), along with statistical information such as P, adjusted P, and B values. All data and statistical analyses were computed from three independent biological replicates. Contrasts were computed for the patN-deletion strain using the wild-type strain as a reference at each time point (i.e., the ΔpatN strain: t = 3 h; wild type: t = 3 h) (ΔpatN vs. wild-type contrast), as well as for each time point using t = 0 of the same strain as a reference (i.e., the ΔpatN strain t = 3 h vs. the ΔpatN strain t = 0 h) (ΔpatN or wild-type time-course contrast). Differential expression of genes with B values ≥0 was considered statistically significant, as before (21). Heat maps were generated using Genesis (25), with the M values from top tables for each time point.