Challenges in using transcriptome data to study the c-di-GMP signaling network in Pseudomonas aeruginosa clinical isolates

Abstract In the Pseudomonas aeruginosa type strain PA14, 40 genes are known to encode for diguanylate cyclases (DGCs) and/or phosphodiesterases (PDEs), which modulate the intracellular pool of the nucleotide second messenger c-di-GMP. While in general, high levels of c-di-GMP drive the switch from highly motile phenotypes towards a sessile lifestyle, the different c-di-GMP modulating enzymes are responsible for smaller and in parts nonoverlapping phenotypes. In this study, we sought to utilize previously recorded P. aeruginosa gene expression datasets on 414 clinical isolates to uncover transcriptional changes as a result of a high expression of genes encoding DGCs. This approach might provide a unique opportunity to bypass the problem that for many c-di-GMP modulating enzymes it is not known under which conditions their expression is activated. However, while we demonstrate that the selection of subgroups of clinical isolates with high versus low expression of sigma factor encoding genes served the identification of their downstream regulons, we were unable to confirm the predicted DGC regulons, because the high c-di-GMP associated phenotypes were rapidly lost in the clinical isolates,. Further studies are needed to determine the specific mechanisms underlying the loss of cyclase activity upon prolonged cultivation of clinical P. aeruginosa isolates.


Introduction
The ubiquitous second messenger bis-(3'5')-cyclic di-GMP (cdiGMP) r egulates the tr ansition between motile, planktonic and sessile, biofilm-associated lifestyles in a wide range of bacteria (Jenal andMalone 2006 , Hengge 2009 ). The intracellular le v el of cdiGMP is tightl y contr olled by the opposing activities of two classes of enzymes, diguanylate cyclases (DGCs) and phosphodiester ases (PDEs). In man y cases, the activ e GGDEF and EAL/HDGYP domains of DGCs and PDEs r espectiv el y ar e fused to tr ansmembr ane or other signal input domains thereby integr ating numer ous envir onmental and cellular stim uli into the cdiGMP signaling network (Galperin and Nikolskay a 2007, Ry an et al. 2012, Sondermann et al. 2012, Chua et al. 2017, Galperin and Schultz 2019. The group of Hengge ( 2009 ) has shown that the c-di-GMP modulating enzymes exhibit sur prisingl y distinct and specific output functions (Sar enk o et al. 2017 ). To explain the specificity, they proposed 'local' c-di-GMP signalling, which involv es dir ect inter actions between specific DGC/PDE pairs and c-di-GMP-binding effector/target systems . T hus , while the totality of c-di-GMP-modulating enzymes gener all y driv es the global switch from a highly motile and virulent bacterial pheno-type to a sessile lifestyle within biofilm structures, the various players and subsets of c-di-GMP-modulating enzymes appear to be responsible for smaller and partially nonoverlapping bacterial phenotypes (Kulesekara et al. 2006 ;D. G. Lee et al. 2006 ).
Pseudomonas aeruginosa is an opportunistic pathogen that can cause se v er e acute as well as difficult to tr eat c hr onic biofilmassociated infections . T he genome of the P. aeruginosa type strain PA14 encodes for ov er all 40 c-di-GMP modulating enzymes. For most of them, it is not known how and under what conditions they ar e activ ated, nor ar e their downstr eam effectors known. Ne w insights into how the activity of individual c-di-GMP modulating r egulons is integr ated to driv e bacterial behaviour to w ar ds biofilm formation might become the basis for de v eloping ne w str ategies to combat problematic chronic infections.
In this study, we r eanal yzed pr e viousl y r ecorded extensiv e RNA sequencing data of 414 clinical P. aeruginosa isolates (Dötsch et al. 2015a, Hornischer et al. 2019. We demonstrate that the identification of a subset of clinical isolates that expr ess thr ee differ ent sigma factors (RpoS, RpoN, and AlgU) at high le v els and another subset of clinical isolates that express the sigma factors at low levels, allo w ed for the identification of genes that were differentially regulated between those two subgr oups. Inter estingl y, the set of differ entiall y r egulated genes r eflected the r egulon of the sigma factors, as identified in a pr e vious study (Schulz et al. 2015 ).
We then tested, whether this a ppr oac h could be of use to identify genes involved in c-di-GMP signaling pathways in the opportunistic pathogen P. aeruginosa . We concentrated our analysis on two DGCs encoded by PA14_23130 (PA3177) and PA14_03790 (PA0290). The transcription of both genes was positiv el y corr elated with the transcription of the gene cluster, which is involved in alg inate biosynthesis. Alg inate is a component of the protective biofilm matrix. In addition to the pel genes, the alg genes have been shown to be regulated by ele v ated c-di-GMP le v els, and the pel and alg genes have been associated with increased biofilm formation (Hentzer et al. 2001, Valentini and Filloux 2016, Merighi et al. 2007, Whitney et al. 2012, Liang et al. 2022 ).

Bacterial strains and growth conditions
The strains used in this study are listed in Table S1 (Supporting Information). If not stated otherwise, all strains were cultured at 37 • C in l ysogen y br oth (LB) medium (10 g/l tryptone, 5 g/l yeast extract, and 7 g/l NaCl) with gentle shaking, on LB plates [supplemented with 1.5% (w/v) agar] or on sheep blood agar plates (Th. Geyer).

Deletion mutant strain construction
In order to obtain P. aeruginosa deletion m utants, the pr otocol developed by Hmelo et al. ( 2015 ) was followed. Primers for two DNA fr a gments flanking the gene of interest were used for PCR amplification (Table S3, Supporting Information). These fr a gments were then SOEing PCR-fused and cloned into pDONRPEX18Gm vector with BP Clonase (In vitrogen). T he allelic-exchange insert of both plasmids was verified by Sanger sequencing. Merodiploids with the allelic-exchange vector (Table S2, Supporting Information) were created by triparental mating of the DH5 α donor with an HB101/pRK600 helper, and the P. aeruginosa strain of interest (r ecipient). Subsequentl y, tr ans-conjugants wer e subjected to SacB-based counter selection on LB plates with 60 mg/ml gentamicin and no-NaCl-LB plates with 15% sucrose. Double cross-over tr ans-conjugants wer e identified via PCR analysis and Sanger sequencing. In this study, deletions of the gene P A14_23130 (P A3177) and P A14_03790 (P A0290) were constructed in the type strain PAO1 and clinical isolate CH3484.

RN A extr action, RN ASeq, and tr anscriptomic analysis
For RNA extraction, an overnight LB culture was diluted to an OD 600 of 0.05 in 10 ml LB medium and grown while shaking (180 rpm) to early stationary phase (OD 600 = 2). RN A w as then extracted using the RNeasy Mini Kit (Qiagen) in combination with Qiashredder columns (Qiagen) according to the manufacturer's instructions. Quality of the RN A w as c hec ked with the RNA Nano Kit (Ag ilent Technolog ies) on a fr a gment-anal yzer (Agilent Bioanalyzer 2100, Agilent Technologies). Depletion of rRN A w as done with the RiboZero Bacteria Kit (Illumina) and cDNA libraries were generated with the ScriptSeq v2 Kit (Illumina). For sequencing the RNA samples were run on an Illumina HiSeq 2500 device in single-end mode with 50 cycles, resulting in ∼37 million raw reads per sample (mean). The r eads wer e then ma pped with a stampy pipeline (Lunter and Goodson 2011 ) to the PAO1 or PA14 genome, r espectiv el y as r efer ence. After initial r emov al of genes with more than 5 counts per million and genes that were absent in at least 98% of the isolates in the dataset, differ entiall y expr essed genes were determined with the R package edgeR (v.3.20.1). Comparison between the strains was done on the basis of differential gene expression (TREAT; edgeR function glmTreat). Significance thresholds were usually set to | log2 fold-change | > 2 and FDR < 0.05, if not indicated otherwise.

Isolate gene-expression based gene-regulon calculation
Regulons based on gene expression of isolates were calculated for each gene individually. Isolates that did not show expression of the r espectiv e gene wer e excluded. Furthermor e, onl y the top 4% of isolates with the highest and lo w est expression w ere selected for differential expression analyses. Exceptions were made if more than five isolates were significant outliers with higher or lo w er expr ession. In this manner, the r egulon was calculated between groups of 5-16 isolates (algorithm-based isolate selection described in Tables S4 and S5, Supporting Information). Outliers were calculated using the grDevices R package and the boxplot.stats function with default settings. After calculating differential expression (edgeR with glmTreat), the resulting lists of differ entiall y expr essed genes that had a | log2 fold-change | > 2 and a P -value < .05 were selected. When a set of differentially expressed genes r eac hed a certain size, the cut-off was tightened to focus on more significant and specific r esults: If mor e than 10 hits were obtained with FDR < 0.1, the significance was set to FDR < 0.1. If more than 20 hits were obtained for FDR < 0.05, the significance was set to FDR < 0.05. If more than 200 hits for FDR < 0.05 and or more than 50 hits for FDR < 0.01 were found, the significance was set to FDR < 0.01. The applied thresholds for each gene regulon are highlighted in red in Tables S4 and S5 (Supporting Information).

96-w ell pla te based gro wth assay
Growth behavior was monitored in a 96-well plate and OD was monitored in a BioTek Synergy plate reader at 37 • C with doubleorbital shaking at 537 rpm for a total of 18 h and an OD 600 measur ement interv al of 20 min. Pr ecultur es wer e gr own in 3 ml LB under standard conditions and diluted to an OD 600 of 0.05, before 200 μl was transferred into the wells of the 96-well plate. The outer w ells w ere filled with sterile water and each isolate was tested in three technical replicate wells, which were randomly distributed on the plate. In total, three individual experiments (biological replicates) were carried out.

Swimming motility assay
Swimming motility assays were conducted following the protocol published by Ha et al. ( 2014a ). Briefly, plates containing 0.3% swim a gar wer e allo w ed to dry at r oom temper atur e (RT) for ∼4 h prior to inoculation. All swim plates were incubated upright at 37 • C for 16-18 h prior to analyses using Ima geJ softwar e (NIH). Since P. aeruginosa motility can be visualized as a sharp ring of growth that forms as the bacteria move outw ar d from the original inoculation point, swim zones of a strain were determined as the colony diameter in cm and av er a ged fr om thr ee tec hnical r eplicates per plate. As a negativ e contr ol, PA14 fliC with a known swimming deficienc y w as used.

Swarming motility assay
Swarming motility assays were performed as pr e viousl y described (Ha et al. 2014b ). Briefly, fr eshl y pr epar ed swarm a gar (0.5%) plates were left to dry at RT for a ppr oximatel y 1 h prior to inoculation. All swarm agar plates were incubated upright at 37 • C for 16-18 h prior to analysis using Ima geJ softwar e (NIH). Since P. aeruginosa mov ement r esults the formation of tendrils or fr actals r adiating from the inoculation center, swarm zones of a str ain wer e determined as the colony area in cm 2 and averaged from two technical replicates per plate. As a negative control, PA14 fliC with a known sw arming deficienc y w as used.

Twitching motility assay
Twitching motility assays were performed as previously described (Turnbull and Whitc hurc h 2014 ). Briefly, LB a gar (1.5%) plates that wer e pr epar ed the day before use were left to adjust to RT for appr oximatel y 1 h prior to stabbing holes into the agar with sterile shortened y ello w tips. All twitc hing a gar plates wer e incubated upright at 37 • C for 24 h prior to removing the agar and staining the twitc hing ar ea with 0.1% crystal violet (Sigma). The twitching area was analyzed using ImageJ software (NIH) and defined as the colony area in cm 2 . Average twitching areas from three technical replicates per plate were determined. As a negative control, either a PA14-like or PAO1-like clinical isolate with known twitchdeficienc y w as used (B ACTOME).

Biofilm quantification by crystal violet attachment assay
A crystal violet attachment assay was utilized to quantify the biofilms, with slight modifications from the method described by O'Toole ( 2011 ) including some minor modification. Overnight cultur es wer e adjusted to an OD 600 of 0.02 and 100 μl of the inoculum were aliquoted to one row of eight wells of a nontreated Ubottom PVC 96-well plate (Corning Inc., NY, USA). For incubation at 37 • C, plates were sealed with an air-permeable membrane and placed in a humid incubator for 24 h or 48 h, r espectiv el y. The supernatant was r emov ed and the wells were washed with water before adding 150 μl of a 0.1% w/v crystal violet staining. After 30 min of incubation, the staining solution was r emov ed and the w ells w ere w ashed and air-dried. 200 μl of 96% ethanol w as added to each well for destaining and incubated for 30 min at R T. A v olume of 125 μl of the solution was tr ansferr ed to a fresh 96-well flat bottom plate and the absorbance was measured at 540 nm. Eac h str ain was tested in two independent experiments with eight tec hnical r eplicates eac h time.

Sta tistical anal ysis
An analysis of variance was performed to reveal significant differences in values between each isolate and their respective wildtypes using a pairwise two-tailed Student t -test in the Gr a phP ad Prism software (GraphPad, La J olla, C A).

Impact of the expression level of global regulators on the overall transcriptional profiles of clinical isolates
We have previously recorded transcriptional profiles of 414 P. aeruginosa clinical isolates grown under standard laboratory conditions (Dötsch et al. 2015a, Hornischer et al. 2019. Analyzing gene expr ession le v els of individual genes acr oss a m ultitude of clinical isolates allows for the selection of those clinical isolates that expr ess an y gene of interest at a high and low le v el, r espectiv el y. As shown in Fig. 1 (A), we exemplarily selected clinical isolates with the highest expr ession le v el of the genes encoding for the three sigma factors RpoS, RpoN, and AlgU. The r espectiv e gr oups of high-and low-expressing isolates clustered separately in a principal component analysis (PCA) based on the expression of genes present in the majority of the clinical isolates ( Fig. 1 B). T hus , despite the fact that the various clinical isolates do not share the same genetic bac kgr ound, high or low expr ession of global r egulators exhibited a common impact on the transcription of a large set of genes, so that the transcriptional profiles clustered in two distinct groups.

Using the di v ersity of transcriptional profiles of clinical isolates to identify gene regulons
Since we found that differential expression of genes encoding global regulators affects the overall transcriptional profile of clinical isolates, we wanted to determine whether the RNA-seq data could be used to identify the regulons of global regulators , i.e . the genes that ar e differ entiall y r egulated in the pr esence or absence of the r espectiv e r egulator. As depicted in Fig. 1 (C), we found 568 genes that were differentially expressed between the two groups of clinical isolates that expressed rpoS at high versus low le v els. The same number of genes were found to be differ entiall y r egulated between clinical isolates expressing rpoN at high versus low le v els, wher eas onl y 110 genes were found to be significantly different in the algU high versus low expressing clinical isolates. A total of 26.4% (42 genes) of the genes that wer e pr e viousl y found to be differ entiall y expr essed in an rpoS deletion m utant as compared to its respective wild-type grown under rpoS inducing envir onmental conditions wer e also found to be differ entiall y expressed in the rpoS high versus low expressing clinical isolates (Schulz et al. 2015 ) (Fig. 1 D; and Table S5, Supporting Information). Similarly, 28.9% (28 genes) of the genes that were previously found to be differ entiall y expr essed in an algU deletion mutant as compared to its respective wild-type grown under algU inducing envir onmental conditions wer e found to be differ entiall y expr essed in the algU high versus low expressing clinical isolates. Ho w ever, no big ov erla p was found for the RpoN regulon (2.2%, seven genes), despite the finding that the transcriptomes of the high versus low rpoN expressing clinical isolates were distinct from another. In conclusion, our data indicate that analyzing clinical isolates, which exhibit altered expression of global regulators under standard laboratory growth conditions , might pro vide a unique opportunity to overcome the challenge that for many c-di-GMP modulating enzymes it is not known under which conditions their expression is activated.

Identification of clinical isolates exhibiting high versus low expression of c-di-GMP modulating enzymes
We selected clinical isolates that exhibited a high versus a low expression of genes encoding for c-di-GMP modulating enzymes under standard LB growth conditions . T he higher transcriptional activity of genes encoding c-di-GMP modulating enzymes despite growth under noninducing conditions may allow uncovering details on the regulon of the respective c-di-GMP modulating enzyme. Figure 2 shows the divergence of the gene expression levels of a selection of 10 c-di-GMP modulating enzymes in the 414 clinical isolates (profiles for all c-di-GMP modulating enzymes can be found in Figure S1, Supporting Information). Most of the clinical isolates express the genes at a comparable level. Ho w ever, the dataset on 414 clinical isolates seems large enough to also identify at least a few isolates that exhibit gene expression levels that are isolates ( x -axis) sorted according to their sigma factor (RpoS, RpoN, and AlgU) gene expression levels ( y -axis). The 4% of the isolates (amounting to 15-16 isolates) that express the genes at the highest (red) or lo w est levels (blue) are colour-coded. (B) Expression profiles of all genes present in 98% of isolates (softcore genome) were used for a PC A. T he ellipses r epr esent the 95% confidence interval for the clinical isolate groups exhibiting a high (red) versus low (blue) expression of the r espectiv e gene. (C) Differ entiall y expr essed genes (log2 fold-c hange > 2; P -v alue of at least < .01) between isolates with high v ersus low expr ession of the three respective sigma factor genes are depicted. Numbers indicate the number of differ entiall y expr essed genes, whic h ar e either upr egulated in high-expr essing (r ed) or low-expr essing isolates (blue). (D) Comparison of the genes that wer e differ entiall y r egulated between gr oups of clinical isolates expressing the respective sigma factor genes at high versus low levels (C) with genes that were previously described to belong to the respective sigma factor regulon (Schulz et al. 2015 ). clearl y differ ent fr om the median for most of the genes encoding c-di-GMP modulating enzymes.

Identification of the PA14_23130 and PA14_03790 c-di-GMP cyclase regulons
We concentrated on the differ entiall y expr essed genes in gr oups of clinical isolates that expressed two genes encoding c-di-GMP cyclases (PA14_23130 and PA14_03790) at high versus low le v els. A total of 38 genes were differentially regulated in groups of clinical isolates that expressed PA14_23130 at high versus low le v els, whereas 98 genes were differentially regulated in groups of clinical isolates that expressed PA14_03790 at high versus low levels (Table S4, Supporting Information). As shown in Fig. 3 , the expression of both cyclase genes was positiv el y corr elated to the expr ession of the alginate biosynthesis genes across the transcriptional profiles of the 414 clinical isolates. A positiv e corr elation of the expression of these two cyclases and the alginate biosynthesis genes was also found in pr e vious RNA-seq studies of P. aeruginosa isolates grown in artificial sputum medium (Huse et al. 2010, Damron et al. 2012. In order to study the downstream effects of the PA14_23130 and PA14_03790 cyclase activity further, we deleted the two genes in the two P. aeruginosa type strains PAO1 and PA14 and recorded their motility phenotypes, growth behaviour and ability to produce biofilms. As depicted in Fig. 4 , swimming, swarming and twitching motility was not altered in the cyclase gene mutants as compared to their respective wild-types . Furthermore , no difference in growth under standard rich medium conditions was observ ed and inactiv ation of the PA14_23130 (PA3177) or PA14_03790 (PA0290) gene did not impact biofilm formation in the PAO1 or PA14 str ain bac kgr ound. We also r ecorded the tr anscriptional profile of the respective deletion mutants. As compared to their PA14 and PAO1 wild-types, 148 genes were differentially regulated in the PA14_23130 deletion mutant and 13 genes in the corresponding P AO1P AO1 P A3177 deletion mutant (Table S6, Supporting Information). The PA14_03790 deletion mutant exhibited two differ entiall y r egulated genes in the PA14 strain background and the corresponding PA0290 deletion mutant 25 genes in the PAO1 str ain bac kgr ound. Ther e wer e no genes that wer e differ entiall y regulated in both deletion mutant strain backgrounds.
We hypothesized that inactivation of the two genes encoding the DGCs in a strain expressing the corresponding genes at ele v ated le v els might pr ovide mor e insight into the downstr eam r egulon. We ther efor e gener ated CH3484 PA3177 and CH3484 PA0290 deletion mutants in the PAO1-like clinical isolate CH3483, which expressed both cyclase encoding genes at high le v els. Again, inactiv ation of the tw o c yclase encoding genes did not have an impact on swimming, twitching and swarming activity. Furthermor e, gr o wth w as not affected nor w as biofilm formation (Fig. 4 ). We also recorded the transcriptional profile of the cyclase deletion mutants. We found 30 and 8 differ entiall y r egulated genes in the CH3484 PA3177 and CH3484 PA0290 deletion mutants as opposed to their corresponding wild-type, respectiv el y. Of note, the tr anscriptional pr ofiles of the CH3484 PA3177 and CH3484 PA0290 deletion mutants were almost identical and no gene was differ entiall y r egulated at a significant le v el between the two deletion mutants. We also found no gene that was differentiall y r egulated in mor e than one P A0290/ P A14_03790 deletion m utant str ain bac kgr ound (P AO1, P A14 or CH3484) and only two genes (PA4888 and PA4889) that wer e differ entiall y r egulated in the PA3177/ PA14_23130 deletion mutants in both the PAO1 and the CH4484 clinical strain background.
Since we did not detect any gene involved in the production of alginate to be differ entiall y r egulated in either of the two cyclase mutants, we manually looked into the alginate biosynthesis gene cluster. In both CH3484 PA3177 and CH3484 PA0290 deletion mutants, the alginate biosynthesis genes were not expressed at a high le v el. Ho w e v er, a r er ecording of the RNA-seq profile of the clinical isolate CH3484 r e v ealed that the alginate gene cluster was also not expressed at an elevated level in CH3484. Instead and as opposed to pr e viousl y r ecor ded RN A-seq pr ofile (Dötsc h et al. 2015b , Hornischer et al. 2019 ; Fig. 3 ), the alginate gene cluster expr ession le v el of CH3484 compar ed well to that of the PA14 and PAO1 strains, indicating that the clinical isolate has lost its high alginate gene expression level (Fig. 5 A).
We, ther efor e, selected a second clinical isolate (M70564993), which had been demonstrated to express the alginate genes at ele v ated le v els (Fig. 3 ), and passa ged the str ain for ov er all 18 days by transferring 100 μl culture to 5 ml fresh LB medium twice per day. Figure 5 (B) shows that while one replicate of a freshly thawed sample of M70564993 still expressed the alginate genes at elev ated le v els, in thr ee other fr eshl y thawed samples (and both of the passaged isolate samples) the high alginate gene expression le v els wer e lost and compar able to that of the PA14 and PAO1 type strains.

Discussion
Rising c-di-GMP le v els driv e the switc h to w ar ds biofilm phenotypes. Ho w e v er, in the opportunistic pathogen P. aeruginosa there ar e v arious c-di-GMP modulating enzymes, which are responsible for different and in parts nonov erla pping bacterial phenotypes (Kulesekara et al. 2006, Römling et al. 2013, Ha et al. 2014c, Valentini and Filloux 2016. Only knowledge on the activity of the individual c-di-GMP signaling pathwa ys , their regulons and how their activities are integrated to adjust bacterial beha viour will pro vide the basis to understand bacterial adaptation to complex environmental challenges, such as those found during an acute versus chronic infection process. In this study, we aimed at making use of available data on the transcriptional profiles of a plethora of clinical P. aeruginosa isolates (Dötsch et al. 2015a, Hornischer et al. 2019. We selected clinical isolates that express genes encoding for distinct c-di-GMP modulating enzymes at high le v els e v en under conditions that normall y would not induce the system. We expected that deletion of the r espectiv e c-di-GMP modulating enzymes in these clinical isolates could provide insight into the role of the enzymes and the regulated pathwa ys , e v en if the environmental conditions that trigger the system are unkno wn. We could sho w that this a ppr oac h w orked w ell for the identification of the regulon of two out of three major sigma factors (AlgU, RpoS).
In this study, we concentrated on two not well c har acterized cdi-GMP c yclases (Poudy al and Sauer 2018 , Wei et al. 2019, Bhasme et al. 2020, whose expression was found to be coregulated with genes involved in alginate biosynthesis in many of our clinical Figure 3. Alginate biosynthesis and PA14_23130 and PA14_03790 gene expression is coregulated in P. aeruginosa clinical isolates . T he heat map depicts the expression level of all genes ( y -axis) that were shown to be differentially expressed in those 58 clinical isolates ( x -axis) that express either PA14_23130 or PA14_03790 at high or low le v els (the r espectiv e r egulons ar e indicated in the two v ertical columns on the far left). High alginate gene expr ession corr elates positiv el y with expr ession of the GGDEF domain containing cyclases P A14_23130 and P A14_03790 (indicated by r ed arr ows on the right side of the heatma p), negativ el y with swimming and positiv el y with survival in G. mellonella larvae and a mucoid colony surface (orange = mucoid, blue = nonmucoid) (Hornischer et al. 2019 ). The clinical isolates CH3484 and M70564993 that were further analyzed in this study are marked by red arrows at the bottom of the heatmap. PA3177 and PAO1 PA0290 or the clinical isolate CH3484 PA3177 and CH3484 PA0290 deletion mutants compared to their respective wild-types. All experiments were performed in three replicates. A fliC deletion mutant served as a negative control (A) . The capability to form biofilm was measured using the crystal violet assay and was also not significantly different in the deletion mutants compared to their respective wild-types both after 24 (filled circles) and 48 h (empty circles). All experiments were performed in three replicates. A mucA (hypermucoid) deletion mutant served as contr ol (B) . Gr o wth w as monitored in a 96-w ell plate in ric h medium at 37 • C in biological triplicates in thr ee independent experiments. No differ ences in growth, motility, and biofilm formation w ere detected betw een the deletion m utants compar ed to their r espectiv e wild-types (C) . * * * * , P < .0005; * * , P < .005.  isolates, as well as under environmental conditions that mimic the cystic fibrosis lung environment (Alvarez-Ortega and Harwood 2007 , Son et al. 2007, Tralau et al. 2007, Anderson et al. 2008, Fung et al. 2010, Bobadilla Fazzini et al. 2013. In order to unr av el whether there is a casual relation between the high expression of the cyclases and the expression of alginate biosynthesis genes, we sought to inactivate the respective cyclases in the high cyclase expressing clinical isolate (CH3484) and characterize the downstream effects (Liang et al. 2022 ). As expected, the inactivation of the two cyclases did not elicit any motility, growth, biofilm, or alginate gene expression phenotype in the two P. aeruginosa type strains PA14 and PAO1, which do not express the two genes at ele v ated le v els under standard labor atory gr o wth conditions. Ho we v er, we also did not detect a correlation between the expression of either cyclase or a motility, growth, biofilm, or alginate gene expression phenotype in the clinical isolate CH3484. Nevertheless, our results do not exclude a causal relationship either, because we found that the increased alginate gene expression was quickly lost in r estr eaked isolates of the CH3484 wild-type . T he same observation was made for another clinical isolate, which following r estr eaking and passa ging also lost the high expression of the alg genes. It is thus v ery likel y that while constructing the corresponding cyclase gene deletion mutations in the clinical isolate the unstable phenotype of high alg gene expression was lost independently of the deletion of the cyclase genes.
Although our ov er all a ppr oac h was not successful, our finding that high gene expression levels of the two cyclases, at least in the two strains analyzed in this study (CH3484 and M70564993) are not stable is inter esting. These r esults fit well with a recent study published by us, where we show that ele v ated c-di-GMP le vels in a clinical strain is gradually lost upon sub culturing (Koska et al. 2022 ). Inter estingl y, the gr adual decline in ele v ated c-di-GMP concentr ations a ppear ed to be independent of genetic ada ptation, but was reminiscent of a memory response in which the switch to low c-di-GMP concentrations in the studied P. aeruginosa isolate was delayed when the bacteria were cultured under non c-di-GMP-inducing conditions (Kordes et al. 2019 ).
In conclusion, it appears that in the future one cannot avoid to construct strains that overexpress the cyclase of interest or e v en better, to identify the environmental conditions under which cyclase gene expression is activated and then determining the downstream effects in the mutant versus wild type under these conditions. A systematic survey of the expression of the various P. aeruginosa c-di-GMP modulating enzymes under different environmental conditions as depicted in Fig. 6 could be a good starting point.

Supplementary data
Supplementary data is available at FEMSMC Journal online.

Funding
This w ork w as funded b y the Deutsc he Forsc hungsgemeinsc haft (DFG, German Research Foundation) under Germany's Excellence Str ategy-EXC 2155-pr oject number 390874280 and the SPP1879). S.H. was furthermore supported by the European Union (EU, ERC Consolidator grant COMBAT 724290) and the Novo Nordisk Foundation (NNF 18OC0033946). This work and M.G. was supported by the Cystic Fibrosis Trust Foundation grant SRC017.

Da ta av ailability
Transcriptional data used in this study have been previously uploaded to the Gene Expression Omnibus (GEO) database with the accession numbers GSE123544, GSE159698, and GSE217100. Transcriptional data of the deletion mutants that w ere recor ded in this study were uploaded with the accession number GSE233207. The genomic data of the 414 clinical isolates wer e pr e viousl y uploaded to the Sequence Read Arc hiv e (SRA) with the accession number PRJNA526797.