LptD depletion disrupts morphological homeostasis and upregulates carbohydrate metabolism in Escherichia coli

Abstract In a previous in silico study, we identified an essential outer membrane protein (LptD) as an attractive target for development of novel antibiotics. Here, we characterized the effects of LptD depletion on Escherichia coli physiology and morphology. An E. coli CRISPR interference (CRISPRi) strain was constructed to allow control of lptD expression. Induction of the CRISPRi system led to ∼440-fold reduction of gene expression. Dose-dependent growth inhibition was observed, where strong knockdown effectively inhibited initial growth but partial knockdown exhibited maximum overall killing after 24 h. LptD depletion led to morphological changes where cells exhibited long, filamentous cell shapes and cytoplasmic accumulation of lipopolysaccharide (LPS). Transcriptional profiling by RNA-Seq showed that LptD knockdown led to upregulation of carbohydrate metabolism, especially in the colanic acid biosynthesis pathway. This pathway was further overexpressed in the presence of sublethal concentrations of colistin, an antibiotic targeting LPS, indicating a specific transcriptional response to this synergistic envelope damage. Additionally, exposure to colistin during LptD depletion resulted in downregulation of pathways related to motility and chemotaxis, two important virulence traits. Altogether, these results show that LptD depletion (i) affects E. coli survival, (ii) upregulates carbohydrate metabolism, and (iii) synergizes with the antimicrobial activity of colistin.


Introduction
The outer membrane (OM) is a distinguishing feature in Gramnegative bacteria.It acts as a selective permeability barrier, prev enting intr acellular access to man y noxious substances found in the extracellular milieu, including antibiotics (Delcour 2009 ).This hallmark organelle plays a pivotal role in cell survival and virulence in Gr am-negativ e bacteria but is absent in Grampositiv e bacteria, including man y taxa that r eside in the gut micr obiota and whic h hav e beneficial effects on host health.The OM is ther efor e an inter esting tar get for de v eloping nov el antibiotics that selectiv el y interfer e with Gr am-negativ e opportunistic pathogens inhabiting the intestinal tract, like Esc heric hia coli , while leaving the surr ounding Gr am-positiv e bacterial community undisturbed and thus reducing the risk of dysbiosis.
OM biogenesis is carried out by three systems: the lipoprotein tr ansport system (LOL), LPS tr ansport system (LPT), and the βbarr el assembl y mac hinery (BAM) (K ono valo va et al. 2017 ).Here , we focus on the LPT pathway, which is responsible for transporting intr acellularl y pr oduced LPS acr oss the periplasmic space to the cell surface.In E. coli , the pathway consists of se v en pr oteins encoded by the essential genes lptA , lptB , lptC , lptD , lptE , lptF , and lptG .In this pathway, the LptB 2 FG ABC transporter, which is strongly associated with LptC, is responsible for extracting the LPS from the inner membrane (IM).LptC is associated with LptA, forming a periplasmic bridge, ultimately making contact with LptD, a β-barr el pr otein that, in complex with LptE, is responsible for inserting the LPS into the OM outer leaflet (Sperandeo et al. 2019 ).
Using an in silico a ppr oac h, we pr e viousl y identified LptD as a pr omising nov el antimicr obial drug tar get selectiv e for E. coli and Klebsiella pneumoniae (Sv anber g Frisinger et al. 2021 ).LptD is a large protein with extracellular and periplasmic domains.It forms a stable complex with the smaller pr otein LptE, whic h r esides inside the LptD β-barrel (Freinkman et al. 2011 ) (Fig. 1 a).Due to the LptD extracellular domains, this protein is an accessible target candidate for the de v elopment of novel antibiotics.The LptDE complex is structur all y conserv ed acr oss Gr am-negativ e species, including important pathogens such as Yersinia pestis , Shigella flexneri , K. pneumoniae , and Salmonella typhimurium (Botos et al. 2016 ).This indicates that interfering with this complex would be a good strategy for targeting multiple Gram-negative pathogens.Yet, the potential of LptD as an antibiotic target has been poorly explored.Mur epav adin is one example of an antibiotic that is able to interfere with LptD in Pseudomonas aeruginosa (Srinivas et al. 2010, Martin-Loeches et al. 2018 ).Murepavadin is thought to target the LptD N-terminus , which ma y partially explain the selectivity as this consists of around 300 residues in P. aeruginosa , compared to 180 in E. coli (Botos et al. 2016 ).Pol ymyxins ar e another example of OM targeting antimicrobials, which after binding to LPS disrupt the IM and OM (Li and Velkov 2019 ).Ho w ever, as these antibiotics are not fit for the treatment of common community-acquired infections, the need for novel antibiotics remains.(b) Schematic of the Mobile-CRISPRi system.The system consists of a c hr omosomall y integr ated dCas9 and an sgRNA.The catal yticall y inactiv e dCas9 (pink) is coupled to an sgRNA (pur ple), whic h tar gets the complex to a specific location on the tar get lptD gene and acts as a physical barrier to RNA pol ymer ase (gr een), stericall y hindering tr anscription of the gene.
Traditional molecular methods for studying gene function are not a ppr opriate for functional anal ysis of essential genes due to their constitutiv e r equir ement for cell viability.With the advent of CRISPR, new tools have been added to the microbiological toolbox.One of these is a CRISPR deri vati ve, CRISPRi, which utilizes two point mutations in the enzyme active site leading to catalytic inactivation of the Cas9 protein, generating a so-called "dead" Cas9 (dCas9) (Qi et al. 2013 ).Coupled with a gene-specific single guide RN A (sgRN A), this complex causes physical bloc ka ge of the RNA pol ymer ase, ther eby silencing the expression of the target gene (Fig. 1 b).This technological advancement has enabled the r epur posing of the CRISPR platform fr om a gene-editing to a gene-silencing tec hnology, gener all y r eferr ed to as gene knoc kdown (Larson et al. 2013 ).Today, CRISPRi provides a useful tool that can be harnessed to understand the function of essential genes and study the effects of gene silencing on cell viability and metabolism.
We applied CRISPRi to investigate the effects of lptD silencing on cell gr owth, mor phology, and tr anscriptome in E. coli .Following efficient and titratable knockdown of the gene, measurements of gene expression sho w ed that a high fold of gene r epr ession could be ac hie v ed without affecting str ain gr owth.The tr anscriptional response to lptD inhibition was investigated along with the morphological changes over time, showing that the response to knockdown is both time-and inducer concentration-dependent and that LptD depletion has profound effects on the transcriptional response with a marked upregulation in carbohydrate metabolism.

Strains, media, and culture conditions
Esc heric hia coli BW25113 CRISPRi strains w ere gro wn in Luria-Bertani (LB) broth at 37 • C, 180 rpm overnight.Kanamycin plates (30 μg ml −1 , Gibco, UK) were used to select for the CRISPRi strains.CRISPRi knockdo wn w as induced using IPTG (Invitrogen Life Technologies, USA) at the desired concentration.Strains and their functions are listed in Table 1 .
Table 1.Strains used in the study.

Strain Function
Esc heric hia coli BW25113 P ar ental str ain ± CRISPRi system Esc heric hia coli MFD pir Helper strain during triparental mating

Construction of mobile-CRISPRi plasmids and strain
Streptococcus pyogenes dCas9 was expressed from the IPTGinducible P LAC promoter (#119271, Addgene).The sgRNA design was based on reported sequences (Wang et al. 2018 ).To make the final sgRNA selection, se v er al sgRNAs wer e tested, and the one showing the gr eatest knoc kdo wn efficienc y using gro wth curves was selected (Table 2 ).sgRNAs were annealed and cloned into the plasmid pJMP1339 using Golden Gate assembly, as previously described (Chen et al. 2017 ).The plasmids used to generate the CRISPRi knockdown mutant (Table 3 ) were cloned in the E. coli MFD pir str ain.Tri-par ental mating using MFD pir carrying either pJMP1339 or pJMP1039 (#119239, Addgene) together with E. coli BW25113 was performed, transferring the system to E. coli BW25113 for c hr omosomal integr ation downstr eam of glmS through conjugation without disrupting the gene, as pr e viousl y described (Peters et al. 2019 ), and thus generating the final knockdown mutant.

Cell viability after CRISPRi knockdown
The effect of the cloned CRISPRi system was assessed through gr owth curv es ov er 24 h using Bioscr een C (Labsystems Oy, Helsinki, Finland).Uninduced cells were grown overnight, diluted 1:100, and allo w ed to gro w for 2 h at 37

Light microscopy
Cells wer e gr own and pr epar ed as described abo ve .T he cultur es wer e split, IPTG was added, and the cells were grown at 37 • C, 180 rpm.Samples were taken at the specified time point and stained using Nigrosin.Cells were imaged using a Zeiss Axioplan 2 microscope with an Axiocam 702 Mono and a Plan-NEOFLUAR 100x Oil Ph 3 objective lens.Images were analyzed using Fiji/ImageJ (Rueden et al. 2017, Schindelin et al. 2019 ).

Electr on micr oscopy
Cells were prepared and pre-grown as previously described.The culture was grown in the presence of 1 mM IPTG for 6 h and harvested by centrifugation.Samples were prepared as previously described (Andersen et al. 2021 ).Semi-thin sections (2 μm) were cut using a glass knife and ultr amicr otome (Leica Ultracut, Leica Micr osystems, Wetzlar, German y) and stained with 1% toluidine blue (Millipore) and 1% Borax (LabChem).Ultra-thin sections (50-70 nm) were cut using a diamond knife (Jumdi, 2 mm) and the ultr amicr otome, contr asted in 2% ur an yl acetate (Polyscience) and lead citrate (Merck).The samples were examined in a Philips CM100 tr ansmission electr on micr oscope, and ima ges were obtained using an Olympus Morada camera and iTEM softwar e (Ol ympus).

RN A isola tion
Cells wer e pr e-gr own and pr epar ed as pr e viousl y described.Cells wer e harv ested at specified time points post-induction by pelleting at 10 000 rpm for 5 min at 4 • C, and supernatant was discar ded.Cell pellets w er e stor ed at −80 • C until RN A w as isolated using RNeasyMini Kit (Qiagen, Sollentuna, Sweden).The quantity and quality of isolated RN A w ere determined b y A260/280 and A260/230 ratio measurements, respectively, using a Nan-oDr op1000 spectr ophotometer (T hermoScientific , Hvido vre , Denmark).

RT-qPCR
Cells were grown as described above with 0 mM or 0.10-0.15mM IPTG and harvested at time points 0-, 2-, 4-, and 5-h post-induction.DNA digestion was done using TURBO™ DNase kit (2 U/ μl) (Ambion, Life Technologies, Naerum, Denmark) to purify RNA samples from contaminating DNA.Re-v erse tr anscription was used to conv ert RN A to cDN A using High Ca pacity cDNA Re v erse Tr anscription Kit (Life Technologies, Naerum, Denmark).R T-qPCR w as subsequently performed on the cDNA using FastStart Essential DNA Green Master (Roche, Hvido vre , Denmark) and a LightCycler 96 (Roche , Hvido vre , Denmark).Primers used to detect lptD were lptD-F: TGGT A TCGCC-CTGTA CCA GA and lptD-R: TGATTGCCA CCGCCCTTTAT, and for the control gene gapA , gapA-F: A CTGA CTGGT A TGGCGTTCC and gapA-R: GTTGC AGCTTTTTCC AGACG.Change in gene expression was calculated using the 2 − CT method.

RNA-seq
Cells were grown as previously described using seven different conditions: 0.1 mM IPTG , 1 mM IPTG , 0.04 μg/ml CST, 0.06 μg/ml CST, 0.1 mM IPTG + 0.04 μg/ml CST, 0.1 mM IPTG + 0.06 μg/ml CST, or without IPTG (contr ol).Eac h of these se v en conditions was tested in four replicates, leading to a total of 28 samples analyzed b y RN A Seq. Cells w er e harv ested at 4.5 h post-induction by centrifugation at 4 • C, 4000 rpm.Pellets wer e sna p-fr ozen in liquid nitrogen and stored at −80 • C. RN A w as isolated as described abo ve .From each sample, 400 ng of total RN A w as used to prepare the library for RN A-Seq follo wing RN Atag-Seq protocol (Shishkin et al. 2015 ), which was sequenced on Illumina NextSeq500.The bcl2fastq software (v2.19,Illumina BaseSpace) was used to conv ert r aw data to fastq files, and RNA-seq data were processed using an in-house analysis pipeline as pr e viousl y described (Lov e et al. 2014, Zhu et al. 2020 ).Finall y, gene expr ession dataset was uploaded to the BioCyc Omics Dashboard (Paley et al. 2017 ) for enric hment anal ysis.

Characterizing the effects of lptD gene silencing on E. coli growth
To investigate the effects of lptD knockdown, the CRISPRi construct w as gro wn under a range of IPTG inducer concentrations (0.1-1 mM).Se v er e gr o wth inhibition w as observed for the first 9 h under high concentration of IPTG (1 mM), follo w ed b y exponential growth reaching saturation after ∼18 h (Fig. 2 a).Growth under low concentrations of IPTG (0.10-0.20 mM) exhibited a dosedependent response (Fig. 2 a).At concentrations ≤0.17 mM growth was similar to that of the uninduced control, indicating a threshold for observable knockdown (Fig. 2 a).Ov er expr ession of the CRISPRi system has pr e viousl y been shown to result in toxicity in E. coli , leading to a perturbed growth pattern (Cho et al. 2018, Cui et al. 2018 ).To rule out any non-specific effect due to the expression of the recombinant system, we de v eloped contr ol str ains containing a functioning CRISPRi system but with non-targeting sgRNAs (NC_1360, NC_1212, NC_163, and NC_118).None of the strains exhibited growth inhibition even upon strong induction of the system (Supplementary Fig. S1 ).This indicates that the gr owth r eduction of E. To investigate the degree of CRISPRi-mediated silencing on lptD tr anscription, we measur ed the r eduction in gene expression upon subinhibitory induction of the CRISPRi system (Fig. 2 b).Low levels of inducer were sufficient to generate a high-fold knockdown, which is in accordance with pr e vious findings (Peters et al. 2016 ).Although lptD knockdown was inducer concentration-dependent at the earlier time points (2 h post-induction), the highest r epr ession ( ∼440-fold reduction) was observed at 5 h post-induction using 0.10 mM IPTG, compared to the ∼128-fold r eduction ac hie v ed with 0.15 mM at the same time point.The impact of concentrations ≤0.13 mM on cell growth became visible only after 5 h (Supplementary Fig. S2 ).Together, these results show that efficient gene silencing can be ac hie v ed with low induction of the CRISPRi system, although there is a lag between gene depletion and any effect on growth as measured by turbidimetry.
To further investigate the impact of LptD depletion on cell viability, we carried out viable cell counts (Fig. 2 c) upon induction of the CRISPRi system.High le v els of inducer (0.30-1 mM) led to a decrease in viable cells up to 4 h, when the cells started to grow back to the same le v els as the control ( ∼10 12 CFU/ml) (Fig. 2 c).Whole-genome sequencing of strains isolated following growth in 1 mM IPTG sho w ed inactiv ating m utations in the CRISPRi system, thus creating suppressor mutants allowing the cells to escape the knockdown effect (Supplementary Fig. S3 ).Cultures exposed to the lo w est le v el of inducer (0.10 mM) exhibited a completely differ ent pattern: gr o wth w as identical to that of the control up until 4 h post-induction, after which a clear decline in the number of viable cells was observed (Fig. 2 c).No recovery from this do wnw ar d tr end was observ ed at the following time points, resulting in very low numbers ( ∼10 7 CFU/ml) of viable cells after 24 h.Conv ersel y, cells exposed to an intermediate IPTG concentration (0.20 mM) sho w ed a complex behavior of initial duplications, follo w ed b y a decline in numbers after 2 h and a subsequent increase after 8 h and until the experimental end-point (24 h).These results show that high le v els of induction lead to suppr essor m utants that inactivate the CRISPRi system and grow back to wild-type cell numbers.On the contrary, low levels of induction appear not to lead to suppr essor m utants under the experimental conditions tested.

LptD depletion disrupts E. coli morphology
The OM is integral for maintaining cell shape in Gr am-negativ e bacteria, and an imbalance of its composition can have detrimental effects on cell mor phology.To inv estigate the effect of LptD depletion on cell morphology, we employed light microscopy and tr ansmission electr on micr oscopy (TEM) to visualize uninduced and fully induced (1 mM) cells over time.Following growth 2 h post-induction, non-homogenous morphologies were observed with a mix of elongated, smooth cells and multi-septated, filamentous cells (Fig. 3 a).This change in morphology continued up to 8 h, with additional deterioration of cell shape .T his phenomenon was also observed under mild depletion, starting from 0.20 mM, and the effect increased with inducer concentration, although cells a ppear ed to return to normal morphology at 24 h (Supplementary Fig. S4 ), most likely due to CRISPRi inactivation.These mor phological c hanges a ppear ed to follow a different timeline than the growth experiments.For example, following 0.20 mM induction, the alter ed mor phologies wer e observ ed at the earlier time points, but cell morphology was largely restored after 24 h growth (Supplementary Fig S4 ).In contrast, the growth experiments sho w ed a reduced population size after 24 h but with an incr eased gr owth r ate between 8 and 24 h (Fig 2 c), indicating a la g between the de v elopment of suppr essor m utations inactiv ating the CRISPRi system and r estor al of the cell population density.
TEM sho w ed no effect on the uninduced control (Fig. 3 b), whereas the induced strain exhibited elongated, filamentous morphologies (Fig. 3 c).Cells a ppear ed visuall y denser and with a r educed periplasm, in line with intracellular accumulation of LPS (Sperandeo et al. 2008 ).This shows that LptD depletion causes dr astic c hanges in cell mor phology.

T he tr anscriptional response to lptD knockdo wn centers around carbohydr a te metabolism
To r e v eal an y genome-wide r esponse to lptD knoc kdo wn, w e profiled and compared the transcriptomes of the induced and uninduced lptD -CRISPRi strain at 4.5 h post-induction.While only 22 genes wer e significantl y (log 2 FC > 1 and padj < 0.05) regulated in response to the presence of 0.10 mM IPTG, the expression of 752 genes c hanged significantl y due to the knockdown of lptD using 1 mM IPTG (Supplementary Table 1 , Supplementary Fig. S5 a,  b).Onl y fiv e genes, in addition to lptD , were found to be shared betw een the tw o conditions: pdxA (4-hydr oxy-L-thr eonine phosphate dehydr ogenase), prpR (pr opionate catabolism oper on r egulatory pr otein), yahO (unc har acterized pr otein), and two IS elements .T his indicates that the cell experiences mild and se v er e le v els of LptD depletion differ entl y.Additionall y, pdxA belongs to the same operon as lptD but is located one gene downstream of the CRISPRi-targeted gene and can be expressed using lptDindependent promoters (Tramonti et al. 2021 ), making it challenging to inter pr et whether the observed change is due to polar effects or a true response to LptD depletion.
Indi vidual re plicates exposed to IPTG clustered together and separated well from the 0 mM IPTG control (Supplementary Fig. S6 a, b), confirming the accuracy of performance of the experiment, processing, and analysis of RNA-Seq samples.To identify modulated metabolic and cellular processes, the gene expression dataset was further explored using the Omics Dashboard of Bio-Cyc (Paley et al. 2017 ).Multiple anabolic and catabolic processes were found to be up-or down-regulated as a response to LptD depletion.Carbohydr ate metabolism-r elated pr ocesses wer e the most significantly upregulated (Fig. 4 a and b).Remarkably, GDPsugar metabolism was highl y upr egulated in the presence of 1 mM IPTG, and most of the genes of this subsystem wer e ov er expr essed (Supplementary Fig. S7 a).In particular, the GDP-L-fucose biosynthesis pathw ay w as upr egulated, especiall y the genes fcl and gmd (Supplementary Fig. S7 b).Mor eov er, lptD knoc kdo wn w as found to lead to the downregulation of lpxK , which is involved in the synthesis of tetr aacyldisacc haride lipid (IV)A by phosphorylating the precursor disaccharide-1-phosphatedisaccharide (Supplementary Table 1 ).The lipid A core is the foundation of the LPS molecule, and the reduction of lipid (IV)A le v els likel y slo ws do wn the rate of the entire LPS biosynthesis process.

Targeting LPS transport upregulates colanic acid biosynthesis
The most significantly upregulated carbohydrate biosynthesis pathw ay w as that for colanic acid biosynthesis (Fig. 4 c).All genes of this pol ysacc haride biosynthesis pathw ay w er e ov er expr essed in the presence of 1 mM IPTG but not for 0.10 mM IPTG (Fig. 4 d).
Recently, it has been proposed that colistin binds to LPS at both the IM and the OM, ultimately causing cell lysis (Sabnis et al. 2021 ).We hypothesized that knockdown of lptD increases LPS levels in the IM and potentiates colistin.Indeed, 0.10 mM IPTG was able to potentiate colistin, where the combination exposure led to a str onger gr owth r eduction compar ed to colistin alone (Supplementary Fig. S8 ).To test whether the colistin-LptD depletion interaction is linked to the upregulation of the colanic acid biosynthesis pathw ay, w e profiled the transcriptome of the lptD -CRISPRi strain grown in the presence of IPTG and colistin individually and in combination.To ca ptur e the r epr oducibility and sensitivity of the tr anscriptional response, two sublethal colistin concentrations (0.06 and 0.04 μg/ml) were selected, which exhibited ∼50% and ∼40% gr owth inhibition, r espectiv el y, compar ed to the unexposed control (Supplementary Fig. S8 ).Remarkably, 0.10 mM IPTG highly upregulated the colanic acid biosynthesis in combination with colistin regardless of the drug concentration (Fig. 4 e).The entire pathway (Fig. 4 f) was str ongl y activ ated by the combination.Conv ersel y, this low concentr ation of IPTG was not sufficient to modulate the exopol ysacc haride biosynthesis pathway, and colistin alone only moderately upregulated this pathway (Fig. 4 e).It has pr e viousl y been r eported that phosphorylation of Ugd (UDPglucose dehydrogenase) by tyrosine-kinase Wzc induces the production of colanic acid, and phosphorylation of Ugd by Etk induces capsule synthesis (Obadia et al. 2007, Lacour et al. 2008 ).Ugd, the Wzc kinase, and the Wzb phosphatase wer e all ov er expressed in the presence of colistin and highl y upr egulated when cells were grown in the presence of IPTG and colistin combination, while Etk was downregulated (Supplementary Table 2 ).

Reduced LPS transfer impairs motility
Among the processes related to the cell exterior, the LPS metabolism subsystems were upregulated when cells were grown in the presence of IPTG and colistin together (Fig. 5 a).Both flagellar and chemotaxis proteins, which determine the locomotion of the cell, were downregulated in the presence of 0.10 mM IPTG but upregulated in the presence of colistin (Fig. 5 a).The combination of IPTG and 0.04 or 0.06 μg/ml colistin str ongl y downr egulated almost all the genes of these two processes (Fig. 5 b and c).c-di-GMP plays a central role in the regulation of flagella biogenesis by modulating the function of regulator YcgR (Guttenplan and Kearns 2013 ), where the c-di-GMP binding form of YcgR inhibits the production of flagella.The intracellular levels of c-di-GMP are in turn regulated by the PdeH (c-di-GMP phosphodiesterase).While exposure to colistin alone increased the expression of pdeH , the combination of IPTG and colistin str ongl y downr egulated the pdeH le vels (Fig. 5 c).This suggests that the combination of colistin and lptD knoc kdown incr eases le v els of c-di-GMP, whic h in turn inhibits the biogenesis of flagella by binding to and inactivating YcgR.

Discussion
The present study describes the response of E. coli to LptD depletion using a CRISPRi system that allows modulation of lptD gene expression.The CRISPRi construct exhibited severe growth inhibition upon strong induction of the CRISPRi system, whereas a sublethal and tuneable response was observed at lo w er inducer concentrations.
We were able to show an efficient knockdown at sub-inhibitory inducer concentrations, suggesting either a protein surplus or slow pr otein turnov er, allowing surviv al during LptD depletion.LptD has pr e viousl y been r eported to be a low-abundance protein with an estimated 60 molecules per cell (Link et al. 1997 ), suggesting that the latter scenario may be the case.In support of this hypothesis, it has been pr e viousl y shown that LptD depletion is slow and can take four to five generations before the associated phenotypic effects are observable (Braun and Silhavy 2002 ).Our results indicate that this lag time depends on the le v el of depletion and that cells can survive under r epr essiv e conditions for up to 4 h before a change in growth can be observed.The LptDE complex forms an extr emel y stable macromolecule in the OM (Chng et al. 2010 ).T hus , the growth inhibitory effects under mild inhibition may be related to LptD dilution through cell division, rather than protein degradation leading to an LptD shortage.
LptD depletion was found to be associated with morphological c hanges.Pr e vious studies on LptD depletion hav e found this to be associated with filamentation, loss of membrane integrity, and cell lysis (Braun and Silhavy 2002 ), or membranous bodies protruding into the periplasm (Sperandeo et al. 2008 ).Abnormal periplasmic membrane structures were not observed in the present study.Instead, cells were found to become highly elongated, displaying either a smooth or filamentous morphology.This discrepancy in phenotype observed between different studies may be related to the LptD depletion levels, which in turn may be affected by methodology making direct comparisons difficult.Re-gardless, it is clear that the cells were heavily impacted as a result of LptD depletion, as the controls containing function CRISPRisystem with non-targeting sgRNAs induced at the same le v el did not exhibit an altered morphology.
The tr anscriptional r esponse to LptD depletion a ppear ed to mainl y r e volv e ar ound carbohydr ate metabolism.Although v arious biosynthesis and degradation processes were found to be upor down-regulated, it is interesting to note that the most prevalent responses were related to an upregulation of carbohydrate metabolism.Ad ditionally, LptD de pletion in combination with colistin exposure was found to reduce the expression of se v er al genes involved in motility.Motility is essential for bacterial pathogens in certain growth phases and is intimately linked with virulence (Josenhans and Suerbaum 2002 ).Increased expression of motilityrelated genes has also been associated with str ains ca pable of causing persistent bovine mastitis when compared to strains causing transient infection (Lippolis et al. 2018 ).The reduced expression of motility genes may indicate that the cells are less likely to establish persistent infection, and thus targeting LptD may be an attr activ e way to contr ol infection.
Expression of all genes involved in the colanic acid biosynthesis pathway incr eased str ongl y in the pr esence of IPTG and colistin in combination, suggesting that LptD depletion potentiates colistin and cells increase production of colanic acid exopol ysacc haride in response to the envelope damage caused by the antibiotic.Conditional LptD depletion has pr e viousl y been shown to lead to an accumulation of LPS decorated with colanic acid in the IM in E. coli (Sperandeo et al. 2008 ).Furthermore, biosynthesis of colanic acid in E. coli has been shown to be dependent on LPS structure (Ren et al. 2016, Wang et al. 2020 ).Studies of the proteomic response to LptC depletion in E. coli found this to be linked to an upregulation of proteins involved in envelope biogenesis, including three proteins involved colanic acid biosynthesis (Martorana et al. 2014 ).Using tr anscriptomic anal ysis of lptD knoc kdo wn, w e successfully sho w ed an upregulation of all 11 genes of the colanic acid biosynthesis pathway.Our results indicate that the abundance of LPS in the OM and the condition of the cell envelope ultimately modulate colanic acid production.The extreme upregulation of colanic acid biosynthesis observed only under the combination treatment indicates a synergistic action between colistin and lptD knockdown.Colanic acid has been reported to play an important role in biofilm formation (Danese et al. 2000 ), which is tightly coupled with motility (Guttenplan and K earns 2013 ).Tr ansiting fr om planktonic to biofilm growth might be a strategy to overcome the str ess of env elope dama ge.Her e, w e sho w that the ov er expr ession of colanic acid exopol ysacc haride is a potential strategy of E. coli to impede the potency of an env elope-dama ging a gent like colistin or to maintain envelope integrity.Our results also indicate that targeting of LPS transport induces a complex transcriptional r esponse, r e v ealing an intricate connection between colanic acid biosynthesis, motility, and biofilm growth.It would be interesting to follow up on these results by investigating the effects of lptD depletion on the Rcs system, which is important for sensing OM damage and is a positive regulator of colanic acid biosynthesis (Meng et al. 2021 ).
Protein depletion using CRISPRi is associated with dr awbac ks, including off-target and the bad-seed effect, which are both linked to the sgRNA design.To avoid any off-target effects, the sgRNAs used in the present study were based on previously published designs following strict rules to avoid off-target effects (Wang et al. 2018 ).Pr e vious findings have shown that certain "seed sequences" within sgRNAs are associated with unexpected toxicity, which is unrelated to the gene expression modification, known as the "bad-seed" effect (Cui et al. 2018 ).To avoid such unwanted and non-specific toxicity, the selected sgRNAs did not contain seed sequences associated with the bad-seed effect.Another dr awbac k of CRISPRi mediated expr ession r egulation is potential polar effects, which arise when the CRISPRi system pr e v ents tr anscriptional r eadthr ough of genes downstream of the intended target.In E. coli , lptD is part of a complex, m ultifunctional oper on comprising the genes lptD , surA , pdxA , rmsA , and apaGH , containing at least eight differ ent pr omoters ca pable of giving rise to a range of differ ent tr anscripts (Tr amonti et al. 2021 ).An y suc h unintentional regulation of these genes may result in additional effects on the cell that could be misinter pr eted as being due to LptD depletion, especiall y surA , whic h assists in the folding of certain OM pr oteins (Lazar and Kolter 1996 ).Ho w e v er, none of the genes present in this operon exhibited significant changes in gene expression a part fr om pdxA , indicating little or no polar effect on the remaining genes.
LPS is essential in the majority of Gr am-negativ e bacteria, where the tight packaging and negative charge of the LPS molecules act to exclude extracellular toxic stressors, conferring intrinsic resistance to w ar d many noxious substances (Bertani and Ruiz 2018 ).Disruption of the OM has been shown to overcome intrinsic, acquired, and spontaneous antibiotic resistance b y allo wing intracellular access or b y over coming many antibiotic inactivation determinants while reducing the rate of spontaneous r esistance de v elopment (MacNair and Br own 2020 ).Furthermore , LPS pla ys an important role in Gram-negative virulence and pathogenicity.In a recent study, a non-pathogenic strain of E. coli exhibiting 500-fold higher virulence than the par ental str ain was isolated from a silkworm infection.This was partially attributed to LptD and LptE mutations, leading to increased secretion of LPS containing OM vesicles and endowing the strain with heightened resistance to w ar d various antibiotics, antimicrobial peptides, and host complement (Kaito et al. 2020 ).Reducing the surface LPS by inhibiting the Lpt pathwa y ma y ther efor e aid in r educing the virulence of a pathogen.T hus , targeting LptD ma y be a good strategy for reducing membrane stability and increasing permeability, thereby potentiating otherwise inactive antibiotics while additionall y decr easing virulence, whic h would aid in the clearance of infection.
Despite its merits, the study presented here has limitations, and some findings could be verified by further r esearc h.It would be interesting to experimentally measure whether CRISPRimediated LptD depletion causes intr acellular accum ulation of LPS.In addition, RT-qPCR could be used to confirm the transcriptional changes in genes critical to the discussion ( surA , pdxA , rmsA , and apaGH ).

Figure 1 .
Figure 1.Gene silencing of LptD in E. coli using CRISPRi.(a) Schematic of the Lpt system in E. coli .LPS is extracted across the IM by LptB2FG together with MsbA and transported through the periplasm by the transenvelope bridge formed by LptA to finally be exported to the OM outer leaflet by LptDE.(b)Schematic of the Mobile-CRISPRi system.The system consists of a c hr omosomall y integr ated dCas9 and an sgRNA.The catal yticall y inactiv e dCas9 (pink) is coupled to an sgRNA (pur ple), whic h tar gets the complex to a specific location on the tar get lptD gene and acts as a physical barrier to RNA pol ymer ase (gr een), stericall y hindering tr anscription of the gene.

F
igure 2. (a) Gro wth curv es of E. coli lptD knoc kdown m utant gr own in LB br oth with differ ent concentr ations of IPTG.Data r epr esent the av er a ge of three biological replicates.(b) Expression of lptD over time quantified by RT-qPCR using increasing concentrations of inducer relative to the housek ee ping gene gapA .Data r epr esent the av er a ge of three biological replicates.(c) Colony-forming units at various concentrations of IPTG.Y -axis shows log-transformation of cell counts.Error bars are SDs of three technical replicates.coli lptD -CRISPRi strain observed in the previous experiments was due to silencing of lptD expression.

Figure 4 .
Figure 4. Expr ession c hange of biosynthesis and colanic acid-r elated genes.(a) Av er a ge expr ession c hanges of individual subsystems involv ed in biosynthesis processes are presented in large dots.Small dots represent the fold-change value of individual genes of the subsystem.(b) Enrichment scores of individual biosynthesis processes are presented as −log 10 of the P -value of enrichment.(c) Average expression changes of the various carbohydrate biosynthesis pathwa ys .(d) Fold change of expression of the different genes found in the colanic acid biosynthesis pathway.(e) Average expr ession c hanges of the v arious carbohydr ate biosynthesis pathways under LptD depletion and/or colistin exposur e .(f) T he colanic acid biosynthesis pathway in E. coli .

Figure 5 .
Figure 5. Expr ession c hanges as a r esult of LptD depletion and/or colistin exposur e.(a) Expr ession c hanges in the cell exterior pathwa ys .T he larger dots r epr esent the av er a ge expr ession c hanges of the individual subsystems, and the smaller dots r epr esent individual genes within the subsystem.(b) Expr ession c hanges in the fla gellum-r elated genes.(c) Expr ession c hanges in the c hemotaxis-r elated genes.

Table 3 .
Plasmids used to generate the CRISPRi knockdown mutant.