|Home | About | Journals | Submit | Contact Us | Français|
Of the known essential division proteins in Escherichia coli, FtsN is the last to join the septal ring organelle. FtsN is a bitopic membrane protein with a small cytoplasmic portion and a large periplasmic one. The latter is thought to form an α-helical juxtamembrane region, an unstructured linker, and a C-terminal, globular, murein-binding SPOR domain. We found that the essential function of FtsN is accomplished by a surprisingly small essential domain (EFtsN) of at most 35 residues that is centered about helix H2 in the periplasm. EFtsN contributed little, if any, to the accumulation of FtsN at constriction sites. However, the isolated SPOR domain (SFtsN) localized sharply to these sites, while SPOR-less FtsN derivatives localized poorly. Interestingly, localization of SFtsN depended on the ability of cells to constrict and, thus, on the activity of EFtsN. This and other results suggest that, compatible with a triggering function, FtsN joins the division apparatus in a self-enhancing fashion at the time of constriction initiation and that its SPOR domain specifically recognizes some form of septal murein that is only transiently available during the constriction process. SPOR domains are widely distributed in bacteria. The isolated SPOR domains of three additional E. coli proteins of unknown function, DamX, DedD, and RlpA, as well as that of Bacillus subtilis CwlC, also accumulated sharply at constriction sites in E. coli, suggesting that septal targeting is a common property of SPORs. Further analyses showed that DamX and, especially, DedD are genuine division proteins that contribute significantly to the cell constriction process.
Bacterial cytokinesis is mediated by a ring-shaped apparatus. Assembly of this septal ring (SR; also called the divisome or septasome) begins at the future site of fission, well before cell constriction initiates, and it remains associated with the leading edge of the invaginating cell envelope until fission is completed. The mature ring in Escherichia coli is made up of at least 10 essential division proteins (FtsA, -B, -I, -K, -L, -N, -Q, -W, and -Z and ZipA), which are each needed to prevent a lethal filamentation phenotype. The first known step in assembly of the division apparatus is polymerization of FtsZ just underneath the cytoplasmic membrane. These polymers are joined by FtsA and ZipA via direct interactions with FtsZ, resulting in an intermediate ring structure (the Z ring), onto which the remaining components assemble in a specific order to form a constriction-competent complex.
In addition to the essential SR proteins, a growing number of nonessential proteins that associate with the organelle are being identified. Some of the latter are likely to serve redundant functions, while some may be required only under particular conditions (for reviews on the topic, see references 15, 19, and 25).
FtsN belongs to the essential SR proteins and is thought to be the last of this class to join the organelle before the onset of cell constriction (1, 9, 11, 57, 59). It is a type II bitopic transmembrane species of 319 residues with a small cytoplasmic domain (residues 1 to 30), a single transmembrane domain (residues 31 to 54), and a large periplasmic domain (residues 55 to 319) (12) (Fig. (Fig.1).1). The periplasmic domain comprises three short regions with an α-helical character that are centered around residues 62 to 67 (H1), 80 to 93 (H2), and 117 to 123 (H3), an unstructured glutamine-rich linker (residues 124 to 242), and a C-terminal globular SPOR domain (residues 243 to 319) that has an affinity for peptidoglycan (55, 60) (Fig. (Fig.11).
As with most SR proteins, it is unclear what the essential role of FtsN is. The ftsN gene was first identified as a multicopy suppressor of a Ts allele in essential division gene ftsA (11). Elevated levels of FtsN were subsequently found to also suppress some Ts alleles in ftsI, ftsK, and ftsQ (11, 18), and even to allow the propagation of cells with a complete lack of FtsK (22, 26) or of FtsEX (48). Depletion of FtsN allows assembly of all the other known essential components into nonconstricting SRs, but the number of ring structures per unit of cell length in FtsN− filaments is two- to threefold lower than in wild-type (WT) cells (9). Bacterial two-hybrid studies suggest that FtsN interacts with several other SR proteins, including FtsA, FtsI (penicillin-binding protein 3 [PBP3]), FtsQ, FtsW, and MtgA (10, 16, 17, 38). Moreover, it was recently shown that the requirement for FtsN itself can be bypassed in cells producing certain mutant forms of FtsA, which are thought to stabilize the SR to a greater degree than native FtsA (5). These observations are all compatible with a general role of FtsN in stabilizing the ring structure. In addition, it was recently found that FtsN interacts directly with PBP1B, one of the major bifunctional murein synthases in E. coli, and that it can stimulate both its transglycosylase and transpeptidase activities in vitro (46). Thus, in addition to stabilizing the SR, FtsN may have a more specific role in modulating septal murein synthesis. Lastly, based on the fact that FtsN is the last known essential protein to join the SR, it is attractive to speculate the protein plays a role in triggering the constriction phase (10, 25). To what degree any of these proposed functions contribute to the essentiality of FtsN remains unclear.
What does seem clear is that the essential activity of FtsN takes place in the periplasm and that residues 139 to 319 are dispensable for its essential function (12, 55). In addition, as residues 1 to 45 are also dispensable for targeting of FtsN to division sites, some portion of the periplasmic domain must also be sufficient to direct the protein to the division apparatus (1).
In a genetic screen for synthetic lethality with min (slm) (6, 7), we isolated a mutant strain carrying a transposon insertion in codon 119 of ftsN. The viability of cells containing this severely truncated ftsNslm117 allele prompted us to better define the functional domains of FtsN, and we did so by studying the properties of fusions between various portions of FtsN to green fluorescent protein (GFP). To sublocalize a subset of these, we took advantage of the ability of the twin arginine transport system (Tat) to export functional and fluorescent GFP fusions into the periplasm, such that their periplasmic localization could be determined in live cells by fluorescence microscopy (6, 8, 50, 54).
We show that the essential function of FtsN can be performed by a surprisingly small periplasmic peptide of at most 35 residues that is centered around helix H2 but that this essential domain (EFtsN) itself is unlikely to contribute much, if anything, to the accumulation of FtsN at constriction sites. On the other hand, the nonessential periplasmic SPOR domain (SFtsN) localized sharply to these sites by itself, while SPOR-less FtsN derivatives localized poorly, at best. Notably, septal localization of SFtsN depended on coproduction of EFtsN, in cis or in trans, unless cells were provided with the FtsAE124A protein (5) to allow constriction to ensue in the complete absence of EFtsN. Localization of SFtsN also depended on the activity of FtsI (PBP3) and the presence of at least one of the periplasmic murein amidases, AmiA, -B, or -C. The results suggest that FtsN joins the division apparatus in a self-enhancing fashion at the time of constriction initiation, which is compatible with a role of the protein in triggering the constriction phase of the division process. In addition, the results, taken together with earlier biochemical work (44, 46, 55), suggest that SFtsN is recruited to some form of septal murein that accumulates only transiently at sites of active constriction.
In addition to FtsN, E. coli produces three proteins of unknown function that also bear a C-terminal SPOR domain (PF05036; Pfam 23) (20). Two of these, DamX and DedD, are inner membrane proteins with the same topology as FtsN, while the third, RlpA, is an outer membrane lipoprotein (43, 47, 53). We found that all three also accumulate at septal rings and that each of their SPOR domains act as autonomous septal targeting determinants. Moreover, phenotypes of the mutants indicate that both DamX and DedD contribute to the cell constriction process, leading to classification of these proteins as new nonessential division proteins.
A SPOR domain is predicted to be present in at least 1,650 (putative) proteins from over 500 bacterial species (PF05036; Pfam 23) (20), raising the question as to how far SPOR properties have been conserved. We find that the SPOR domain of CwlC, a Bacillus subtilis murein amidase that is active during late stages of sporulation (39, 44), also accumulates sharply at division sites in E. coli.
Our results predict that many other bacterial SPOR domain proteins specifically recognize the same or closely related target molecule(s) that accumulates transiently at sites of cell constriction. This is supported by a very recent study showing that SPOR domain proteins from Burkholderia thailandensis, Caulobacter crescentus, and Myxococcus xanthus accumulate at cell constriction sites as well (45).
Relevant plasmids and phages are listed in Table Table1.1. Plasmids pMLB1113 (14), pDB280 (30), pKD3, pKD13, and pKD46 (13), pAH68 and pAH83 (29), pDB391 (32), pCH157 (40), pCH201 (31), pCH151 (8), pTB6 and pTB8 (6), pTB102 (7), pMG36 and pMLB1113ΔH (24), pCH181, pCH233 and pCH268 (3), and pCH310, pFB114, pFB184, pTB98, and pTB222 (4) have been described before.
Unless indicated otherwise, MG1655 or TB28 chromosomal DNA was used as the template in amplification reactions. Sites of interest (e.g., relevant restriction sites or those allowing for targeted recombination) are underlined in the primer sequences below.
Plasmid pBL5 (attHK022 bla lacIq Plac::TTgfp) was obtained by replacing the 2,343-bp ApaI-MfeI fragment of pTB222 with the 1,374-bp ApaI-MfeI fragment of pTB6.
For pBL6 (attHK022 bla lacIq Plac::TTgfp-ftsN241-319), the 2,513-bp ApaI-HindIII fragment of pTB222 was replaced with the 1,806-bp ApaI-HindIII fragment of pMG4.
For pBL12 [cat repA(Ts) ftsAE124A], an ftsA plasmid that will be described elsewhere (pCH316), was mutagenized by the QuikChange method (Stratagene) with primers 5′-CGGTGCGTGTGCGCGATGCGCATCGTGTGCTGCATGTG-3′ and its reverse complement. This yielded pBL14, in which codon 124 of ftsA is changed from GAG to GCG, simultaneously resulting in a diagnostic FspI site (underlined). pBL12 was then obtained by replacing the 852-bp BglII-AscI fragment of pDB280 [cat repA(Ts) ftsA] with that of pBL14.
To construct pBL28 (Plac::TTgfp-damX338-428), primers 5′-AGTTGGATCCTTGAAATCGGCACCGTCCAGCC-3′ and 5′-CCCAAGCTTTTACTTCAGATCGGCCTGTACC-3′ were used to amplify a fragment of pFB269 (Plac::yfp-damX). The product was digested with BamHI and HindIII, and the 275-bp fragment was used to replace the 802-bp BamHI-HindIII fragment of pCH282 (Plac::TTgfp-ftsN55-319).
For pBL29 (Plac::TTgfp-rlpA281-362), pMG40 (Plac::rlpA) was used as a template in a PCR with primers 5′-GAACAGATCTCAAGCCGTCTCGCAAAGCGCCAGC-3′ and 5′-CGGAAGCTTTTACTGCGCGGTAGTAATAAATGACTGTAATTGGGC-3′. The product was treated with BglII and HindIII, and the resulting 254-bp fragment was used to replace the 802-bp BamHI-HindIII fragment of pCH282.
For pBL34 (Plac::TTgfp-cwlC181-255), primers 5′-GCAGGGATCCAGCTCAGGGTTATATAAGG-3′ and 5′-CCAAGCTTATGATTCTAGGATCACAATAGCGTCAAAG-3′ were used to amplify a chromosomal fragment of Bacillus subtilis strain PY79 (kindly provided by Peter Setlow). The product was digested with BamHI and HindIII, and the resulting 228-bp fragment was used to replace the 802-bp BamHI-HindIII fragment of pCH282.
To obtain pBL46 [aph repA(Ts) araC PBAD::γ β exo], the bla gene on pKD46 (13) was replaced with aph by λ-red recombineering (13, 61). The aph cassette of pKD13 was amplified with primers 5′-GTGAATGATGTAGCCGTCAAGTTGTCATAATAAATCGATGACGCTGCCGCAAGCACTCAG-3′ and 5′-TTTGGTCATGAGATTATCAAAAAGGATCTTCACCTAGATCGATCCCCTTATTAGAAGAAC-3′, resulting in a 935-bp fragment with end sequences homologous to those flanking bla on pKD46. Recombination of this fragment with the plasmid in strain TB10/pKD46 yielded TB10/pBL46.
For pBL47 (Plac::damX), the damX gene was amplified with primers 5′-GCTCTAGAGCATTAACCTTTTAGTGGGGTG-3′ and 5′-CGCTCGAGTTACTTCAGATCGGCCTGTAC-3′. The product was treated with XbaI and XhoI, and the resulting 1,318-bp fragment was used to replace the 1,099-bp XbaI-XhoI fragment of pCH310.
To obtain pBL48 (attHK022 cat lacIq Plac::zipA-gfp), the bla gene on the integrated form of pTB98 (iTB98) was replaced with cat by recombineering. The cat cassette of pKD3 was amplified with primers 5′-ATGAGACAATAACCCTGATAAATGCTTCAATAATATTGAAATGAATATCCTCCTTAGTTCC-3′ and 5′-TTAAAAATGAAGTTTTAAATCAATCTAAAGTATATATGAGGTGTAGGCTGGAGCTGCTTCG-3′, resulting in a 1,091-bp fragment with end sequences homologous to those flanking bla on pTB98. Recombination of this fragment with the integrated plasmid in strain TB28(iTB98)/pBL46 and subsequent curing of pBL46 by growth at 37°C yielded TB28(iBL48). The excised and replicative form of the plasmid was next obtained by preparation of a P1 lysate and infection of strain BW23473/pAH83, essentially as described previously (29).
Plasmids pBL49 (attHK022 cat lacIq Plac::TTgfp) and pBL50 (attHK022 cat lacIq Plac::TTgfp-ftsN241-319) were obtained as described for pBL48 above, except that the cat fragment was recombined with strains TB28(iBL5)/pBL46 and TB28(iBL6)/pBL46, respectively.
For pBL57 (attHK022 cat lacIq Plac::TTgfp-cwlC181-255), the 228-bp BamHI-HindIII fragment of pBL34 was used to replace the 243-bp BamHI-HindIII fragment of pBL50.
To create pCH276 (Plac::gfp-ftsN1-123), primers 5′-CGCAGCTAGCAAAGGAGAAGAACTTTTCACTGG-3′ and 5′-CGAACTCGAGAAGCTGACGTTGTTCTGGTGTCAG-3′ were used to amplify a 1,139-bp fragment of pCH201 (Plac::gfp-ftsN). This fragment was digested with NheI (internal to fragment) and XhoI (underlined), and the 411-bp product was used to replace the 1,119-bp NheI-XhoI fragment of pCH181 (Plac::gfp-minD minE-le).
Plasmid pCH282 (Plac::TTgfp-ftsN55-319) was created in several steps. Primers 5′-CTGTACGGATCCACGCATCACAAGAAAGAAGAGTCC-3′ and 5′-TGAGAAGCTTAACCCCCGGCGGCGAG-3′ were used to amplify a 989-bp fragment of pCH201, which was digested with BamHI and HindIII. The 802-bp product was used to replace the 25-bp BamHI-HindIII fragment of pET21A, yielding pCH231 (PT7::ftsN55-319). The 835-bp NheI-HindIII fragment of pCH231 was subsequently used to replace the 66-bp NheI-HindIII fragment of pTB6 (Plac::TTgfp), yielding pCH282.
For pCH288 (Plac::TTgfp-ftsN55-123), primers 5′-CTGTACGGATCCACGCATCACAAGAAAGAAGAGTCC-3′ and 5′-GACGTTGTAAAACGACGGCCAGTG-3′ were used to amplify a 268-bp fragment of pCH276, and this fragment was treated with BamHI and HindIII (internal to fragment). The 230-bp product was used to replace the 802-bp BamHI-HindIII fragment of pCH282, yielding pCH288.
Plasmid pCH305 (Plac::yfp-mreD-le) encodes an E. coli codon-optimized yellow fluorescent protein (YFP) and was used as a cloning vehicle. To create it, yfp was amplified from pNR14 (kindly supplied by Nick Renzette and Steven Sandler) using primers 5′-TTAACCATGGCCGACTCTGGTGACTACCTTTGGTTATGGTCTG-3′ and 5′-TCATGCTAGCTTTGTAGAGCTCATCCATGCCGTGCGTGATACC-3′. The 563-bp product was treated with NcoI and NheI, and the 549-bp fragment was used to replace the 549-bp NcoI-NheI fragment of pCH303 (Plac::gfp-mreD-le). The latter was obtained by removal of the SacI site in pCH233 (3) by digestion with the enzyme, treatment with T4 DNA polymerase, and ligation of the blunted ends.
For pCH354 (Plac::gfp-ftsN1-243), we used primers 5′-ATGACTGGTGGACAGCAAATGG-3′ and 5′-GGCGCTCGAGTTTTTTCTCCGCCGTCGGTTTTGG-3′ to amplify a 781-bp fragment of pCH201. The fragment was treated with BamHI (internal to fragment) and XhoI (underlined), and the 744-bp product was used to replace the 330-bp BamHI-XhoI fragment of pMG12.
For pFB236 (Plac::gfp-dedD), dedD was amplified with primers 5′-CGCGGATCCGTGGCAAGTAAGTTTCAGAATCGGTTAGTGGGC-3′ and 5′-CCCAAGCTTTTAATTCGGCGTATAGCCCATTACCACGCC-3′. The product was treated with BamHI and HindIII, and the resulting 669-bp fragment was used to replace the 1,039-bp BamHI-HindIII fragment of pFB114 (Plac::gfp-mreB5-347).
For pFB239 (Plac::dedD), dedD was amplified with primers 5′-GCTCTAGAGCACATGTCATGGAAGTGATTGACGCG-3′ and 5′-CCCAAGCTTTTAATTCGGCGTATAGCCCATTACCACGCC-3′. The product was treated with XbaI and HindIII, and the resulting 706-bp fragment was used to replace the 1,974-bp XbaI-HindIII fragment of pFB211 (Plac::mreC-rfp), a pMLB1113 derivative that will be detailed elsewhere.
For pFB241 (Plac::rlpA-rfp), rlpA was amplified with primers 5′-GCTCTAGAGAAAATGTTGTCGAAAAGCGTGTAAGAGGTGCGC-3′ and 5′-CCGCTCGAGCTGCGCGGTAGTAATAAATGACTGTAATTGGGC-3′. The product was treated with XbaI and XhoI, and the resulting 1,127-bp fragment was used to replace the 584-bp XbaI-XhoI fragment of pMG36 (Plac::pal-rfp).
For pFB269 (Plac::yfp-damX), damX was amplified with primers 5′-GCGGATCCGAATGGATGAATTCAAACCAGAAGACGAGC-3′ and 5′-CCCAAGCTTTTACTTCAGATCGGCCTGTACC-3′. The product was treated with BamHI and HindIII, and the resulting 1,294-bp fragment was ligated to the 8,406-bp BamHI-HindIII fragment of pCH305 (Plac::yfp-mreD-le).
For pMG4 (Plac::TTgfp-ftsN241-319), primers 5′-AGAGGGATCCGAGAAAAAAGACGAACGCCGCTGG-3′ and 5′-TGAGAAGCTTAACCCCCGGCGGCGAG-3′ were used to amplify a 258-bp fragment of pCH201. The product was digested with BamHI and HindIII, and the 244-bp fragment was used to replace the 802-bp BamHI-HindIII fragment of pCH282.
For pMG5 (Plac::TTgfp-ftsN281-319), primers 5′-GTAAGGATCCAATGGCTGGAATCGTGTGGTC-3′ and 5′-TGAGAAGCTTAACCCCCGGCGGCGAG-3′ were used to amplify a 138-bp fragment of pCH201. The product was digested with BamHI and HindIII, and the 124-bp fragment was used to replace the 802-bp BamHI-HindIII fragment of pCH282.
Plasmids pMG12 (Plac::gfp-ftsN1-105), pMG13 (Plac::gfp-ftsN1-81), and pMG47 (Plac::gfp-ftsN1-90) were constructed in a similar fashion as pCH276 (see above), using primers 5′-CGCAGCTAGCAAAGGAGAAGAACTTTTCACTGG-3′ (in each case) and 5′-GCCACTCGAGACCGGCAGAAGGTTCTGTGGGCGC-3′ (pMG12), 5′-AGCGCTCGAGTTCTTCTGGTTTTGGTGGTAGTCCGTTTCC-3′ (pMG13), or 5′-ATTTCTCGAGCTCTTTAATGTAGCGCCAGCGTTCTTCTGG-3′ (pMG47).
For pMG14 (Plac::TTgfp-ftsN71-105), primers 5′-GCCAGGATCCACCGGAAACGGACTACCACCAAAACC-3′ and 5′-GACGTTGTAAAACGACGGCCAGTG-3′ were used to amplify a 170-bp fragment of pMG12. The product was digested with BamHI and HindIII (internal to fragment), and the 140-bp fragment was used to replace the 802-bp BamHI-HindIII fragment of pCH282.
Plasmid pMG15 (Plac::TTbfp-ftsN71-105) was obtained by replacing the 549-bp NcoI-NheI fragment of pMG14 with that of pDB391 (Plac::bfp-minD).
For pMG20 (PBAD::TTbfp-ftsN71-105), primers 5′-ATTTTCTAGATTAAGAAGGAGATATACCATGAACAATAACGATCTCT-3′ and 5′-AAAGGGGGATGTGCTGCAAG-3′ were used to amplify a 1,140-bp fragment from pMG15. The product was digested with XbaI and SalI (internal to fragment), and the 1,025-bp fragment was used to replace the 2,216-bp XbaI-SalI fragment of pCH264 (PBAD::gfp-ftsA), a derivative of pBAD33 (28) that will be detailed elsewhere.
For pMG44 (Plac::TTgfp-dedD140-220), pFB239 was used as a template in a PCR with primers 5′-CGGGATCCACGGGTAAAGCCTATGTTGTGC-3′ and 5′-CCCAAGCTTTTAATTCGGCGTATAGCCCATTACCACGCC-3′. The product was treated with BamHI and HindIII, and the resulting 252-bp fragment was used to replace the 802-bp BamHI-HindIII fragment of pCH282.
For pMG50 (Plac::TTgfp-ftsN71-90), primers 5′-CGCAGCTAGCAAAGGAGAAGAACTTTTCACTGG-3′ and 5′-ATTTCTCGAGCTCTTTAATGTAGCGCCAGCGTTCTTCTGG-3′ were used to amplify an 821-bp fragment of pMG14. The product was digested with NcoI (internal to fragment) and XhoI, and the 642-bp fragment was used to replace the 789-bp NcoI-XhoI fragment of pCH288.
For pMG59 (attHK022 bla lacIq Plac::gfp-ftsN1-319), the 1,745-bp XbaI-HindIII fragment of pCH201 was used to replace the 1,778-bp XbaI-HindIII fragment of pTB222 (attHK022 bla lacIq Plac::zipA-gfp).
For pMG60 (attHK022 bla lacIq Plac::gfp-ftsN1-319), the 1,539-bp XbaI-HindIII fragment of pCH354 was used to replace the 1,778-bp XbaI-HindIII fragment of pTB222.
λFB236 (Plac::gfp-dedD) and λFB241 (Plac::rlpA-rfp) were obtained after crossing pFB236 and pFB241, respectively, with λNT5 (imm21), as described previously (14).
Relevant E. coli strains are listed in Table Table22.
Amidase knockout strains BL7 to BL12 and BL25 were created by λRed-mediated recombineering (13, 61). A 1,093-bp amiA<>cat fragment was generated by amplifying the cat cassette of pKD3 using primers 5′-CCGTCGTAACGGATTACGCGATACGATATAACATCTGGACCATATGAATATCCTCCTTAG-3′ and 5′-TAACAGCCGTTCTTCTTCCGGGTTGGTGATAAACGAGGTGTGTAGGCTGGAGCTGCTTCG-3′, while the aph cassette of pKD13 was amplified using 5′-CTGGCGCGTTTAGCCGGTTAACCTTTGAAAGGTGGCGGGAATTCCGGGGATCCGTCGACC-3′ and 5′-CCATACGGAAGCGCCAGTCGCACTGCGGTTCGGTGCGGCGTGTAGGCTGGAGCTGCTTCG-3′ to yield a 1,383-bp amiB<>aph fragment, or by using 5′-GTTTCCGGTCAAATTAGTCGTTTACTTGTTACACAGCTTAATTCCGGGGATCCGTCGACC-3′ and 5′-CTGGTCTCCCATAAAAAAGCGCCATTCAGCGCCTTTTTAGTGTAGGCTGGAGCTGCTTCG-3′ to generate a 1,383-bp amiC<>aph fragment (chromosomal sequences underlined). Each fragment was recombined with the chromosome of TB10, resulting in three strains that served as donors for transduction of each knockout allele into TB28, to generate BL7 (amiB<>aph), BL9 (amiC<>aph), and BL25 (amiA<>cat). In BL25, cat replaces 751 bp of the amiA gene (from bp −25 to +726), while aph replaces 811 bp of the amiB gene (from bp +2 to +813) in BL7 and the entire amiC gene (from bp −22 to +1343) in BL9. FLP-mediated eviction of aph (13) from BL7 yielded BL8 (amiB<>frt), and transduction of amiC<>aph into BL8 resulted in BL10. Transduction of amiA<>cat into BL9 and BL10 then gave rise to BL11 and BL12, respectively.
Strain BL18 (ftsAE124A) was obtained by exchance of the chromosomal wild-type ftsA allele of strain TB28 with the ftsAE124A allele on plasmid pBL12 [cat repA(Ts) ftsAE124A] (Table (Table1),1), using the method of Hamilton et al. (33). BL18 was transformed with pBL12, and transduction of the ftsN<>aph allele from CH34/pCH201 yielded BL20/pBL12. Strains BL20(iBL5)/pBL12 and BL20(iBL6)/pBL12 were obtained by transduction of the integrated forms of pBL5 and pBL6 from TB28(iBL5) and TB28(iBL6), respectively, into BL18/pBL12 prior to the introduction of ftsN<>aph.
For BL38 (damX<>cat), the cat cassette of pKD3 was amplified using primers 5′-TGCAACGTGGTAAGCATTAACCTTTTAGTGGGGTGTTAACCATATGAATATCCTCCTTAG-3′ and 5′-CGTAGAAGTCGCAGCCGGTGCCTTCGTGCTCGCCACAGGGTGTAGGCTGGAGCTGCTTCG-3′. Recombination of the 1,093-bp fragment with the chromosome of TB10/pFB269 yielded BL36/pFB269, in which bp 1 to 869 of damX were replaced with a cat gene that was transcribed in the same direction, leaving promoters P4 and P5 of the downstream dam gene intact (42). Transduction of damX<>cat from the latter to TB28 then yielded BL38.
BL40 was obtained by transduction of damX<>cat from BL36/pFB269 to MG14, BL41 was obtained by transduction of dedD<>cat from FB71/pFB239 to MG13, BL42 (rlpA<>frt damX<>cat) was obtained by transduction of damX<>cat from BL36/pFB269 to MG13, BL43 was obtained by transduction of ftsN::EZTnKan-2 (ftsNslm117) from TB77 to BL38, BL44 was obtained by transduction of ftsNslm117 from TB77 to MG13, BL45 was obtained by transduction of ftsNslm117 from TB77 to BL42, BL47 (rlpA<>frt damX<>frt) was obtained by eviction of cat from BL42, and BL48 was obtained by transduction of dedD<>cat from FB71/pFB239 to BL47.
Strain CH31(aph araC PBAD::ftsN) was obtained by transduction of PftsN<>(aph araC PBAD) from TB54 to TB28 and required arabinose for survival.
To create CH34 (ΔftsN), the aph cassette of pKD13 was amplified with primers 5′-TCCGGGGATCAACACGCGCGTTACCTTAAAGTAAACCTTTGTGTAGGCTGGAGCTGCTTC3′ and 5′-TCCAGCCATTGTTGGTGGTGATTTTCGAGTCAAAGCCTTCATTCCGGGGATCCGTCGACC-3′, resulting in a 1,384-bp fragment with end sequences homologous to the chromosomal ftsN locus (underlined). Recombination of this fragment with the chromosome of TB10/pCH201 yielded CH33/pCH201, in which 890 bp of the ftsN locus (from bp −80 to bp +810) was replaced with aph, whose transcription was directed away from the downstream hslV gene. Transduction of ftsN<>aph from CH33/pCH201 to TB28/pCH201 yielded CH34/pCH201.
Strain FB55 (dedD::EZTnKan-2) resulted from a new screen we recently described for cell shape/division mutants in which an increased level of FtsZ provides a selective advantage (4). Strain Rod2361/pFB184 (ΔlacIZYA dedD::EZTnKan-2/bla lacIq Plac::sdiA::lacZ) was recovered as a solid blue colony after plating an EZTnKan-2 transposon library of host strain TB28/pFB184 (ΔlacIZYA dedD::EZTnKan-2/bla lacIq Plac::sdiA::lacZ) at room temperature on LB agar supplemented with isopropyl-β-d-thiogalactopyranoside (IPTG) and 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) (LB-IX), as described previously (4). Mapping the EZTnKan-2 insertion site as before (6) revealed that the transposon had inserted just after bp 48 of dedD. Subsequent tests showed that extra FtsZ is not essential for viability of the mutant but provides a modest advantage by reducing the average length of cell chains formed by dedD mutants (data not shown). FB55 was obtained by transduction of dedD::EZTnKan-2 to TB28.
For FB70 (rlpA<>cat), the cat cassette of pKD3 was amplified with primers 5′-ATGCGTAAGCAGTGGCTCGGGATCTGCATCGCGGCAGGAATGCTCGCGGCTGTGTAGGCTGGAGCTGCTTCG-3′ and 5′- CTGCGCGGTAGTAATAAATGACTGTAATTGGGCTTCGGTTTGCAAACGTTTTAGTTCCTATTCCGAAGTTCC-3′, yielding a 1,099-bp fragment with end sequences homologous to the chromosomal rlpA locus (underlined). Recombination with the chromosome of TB10/pCX16 yielded FB69/pCX16, in which 986 bp of the rlpA gene (from bp +51 to bp +1036) was replaced with cat and transcription of the latter was directed away from the downstream dacA gene. Transduction of rlpA<>cat from FB69/pCX16 to TB28 resulted in FB70.
For FB71 (dedD<>cat), the cat cassette of pKD3 was amplified with primers 5′-CGCACATGTCATGGAAGTGATTGACGCGAGGAGAAGCGGTGGCAAGTAAGTAATGTGTAGGCTGGAGCTGCTTCG-3′ and 5′-TTAATTCGGCGTATAGCCCATTACCACGCCACTTAAGCCAGAAAGTTGCTTTAGTTCCTATTCCGAAGTTCC-3′, resulting in a 1,102-bp fragment with end sequences homologous to the chromosomal dedD locus (underlined). Recombination of this fragment with the chromosome of TB10/pCX16 yielded FB68/pCX16, in which 598 bp of the dedD gene (from bp +14 to bp +612) was replaced with cat and transcription of the latter was directed away from the downstream cvpA gene. Transduction of dedD<>cat from FB68/pCX16/pFB239 to TB28 resulted in FB71.
Strains MG13 and MG14 were obtained by eviction of cat from FB70 and FB71, respectively.
Strain MG19(λFB236) was obtained by transduction of ftsN::EZTnKan-2 (ftsNslm117) from TB77 into MG14(λFB236) with selection on LB plates containing 20 μg/ml kanamycin and 100 μM IPTG.
Strain Slm117/pTB8 (ΔlacIZYA ΔminCDE ftsN::EZTnKan-2/bla lacIq Plac::minCDE lacZ) was recovered as a solid blue colony after plating an EZTnKan-2 transposon library of host strain TB43/pTB8 (ΔlacIZYA ΔminCDE/bla lacIq Plac::minCDE lacZ) at 30°C on LB-IX, as described previously (7). Restreaking of the colony on LB-IX gave rise to both solid blue (Min+) as well as solid white (Min−) colonies, and the latter were still about half the size of the former. This indicated that, relative to previously described slm alleles (6, 7), the ftsNslm117 lesion conferred a more subtle growth defect on Min− cells. The exact site of the EZTnKan-2 insertion was determined as described before (6).
Strain TB77 (ftsNslm117) was obtained by transduction of ftsN::EZTnKan-2 from Slm117 to TB28.
Unless stated otherwise, cells were grown at 30°C in LB (0.5% NaCl) medium or in M9 minimal medium supplemented with 0.2% Casamino Acids, 50 μg/ml l-tryptophan, 50 μM thiamine, and 0.2% of either glucose (M9-glucose) or maltose (M9-maltose). When appropriate, medium was supplemented with 50 μg/ml ampicillin (Amp), 50 μg/ml spectinomycin (Spec), 20 μg/ml kanamycin (Kan), 25 μg/ml chloramphenicol (Cam), or 12.5 μg/ml tetracycline (Tet). Amp and Cam concentrations were reduced to 15 and 10 μg/ml, respectively, when cells carried bla or cat integrated in the chromosome. Other details are specified below in the text.
MG13(λFB241) [ΔrlpA(Plac::rlpA-rfp)] cells (see Fig. 10B and C, below) were imaged on a Nikon 90i microscope equipped with a 100× (1.45 numerical aperture) oil objective, an X-Cite 120 fluorescence illuminator (EXFO), a red fluorescent protein (RFP)-specific filter set (532- to 587-nm excitation, 595-nm dichroic mirror, 608- to 683-nm barrier), and a Photometrics QuantEM:512SC back-illuminated electron-multiplying charge-coupled-device camera. All other imaging was performed on a Zeiss Axioplan-2 microscope system as previously described (35). Live cells were imaged on pads of 1.0 or 1.2% agarose with 0.5% NaCl.
In synthetic lethality screens for transposon insertions that are particularly detrimental in the absence of a functional Min system (6, 7), we recovered a mutant strain (Slm117) (Table (Table2)2) in which EZTnKan-2 disrupts ftsN at codon 119. Consequently, the FtsNslm117 peptide still contains the cytoplasmic (FtsN1-30), transmembrane (FtsN31-54), and periplasmic H1 (~FtsN62-67) and H2 (~FtsN80-93) portions of native FtsN, but it lacks most of H3 (~FtsN117-123) and all of the linker (FtsN124-242) and C-terminal SPOR (FtsN243-319) domains (Fig. (Fig.1).1). Since ftsNslm117 cells are viable, this further narrowed the minimal essential domain of FtsN (EFtsN) as previously inferred from complementation experiments (55). Although capable of growing at normal rates, cells carrying the ftsNslm117 allele did show a distinct division phenotype. When TB77 (ftsNslm117) cells were grown in LB medium, many formed chains of two or more cell units (Fig. (Fig.2A2A and Table Table3),3), and this division defect became more pronounced when cells were grown in M9 minimal medium (Fig. (Fig.2B).2B). The absence of a functional Min system was not lethal to ftsNslm117 cells but, consistent with recovery of the insertion in the slm screen, further aggravated their chaining/filamentation phenotype (data not shown).
Two possible causes of cell chaining are inefficient splitting of septal murein (e.g., in ami or envC mutants) or inefficient tethering of the outer membrane (OM) to the freshly separated septal murein layers (e.g., in tol-pal mutants) (24, 34, 49). Both types of defect result in a delayed invagination of the OM and, therefore, in a localized increase in the intermembrane distance at sites of constriction. At least in the case of envC and tol-pal chains, this can be readily detected by ring-like accumulations of periplasmic Tat-targeted GFP (TTGFP) at these sites (6, 24). In contrast, we detected no obvious TTGFP “rings” in chains of TB77/pTB6 cells (Fig. (Fig.2E),2E), indicating that ftsNslm117 cells do not suffer a gross delay in OM invagination. In turn, this suggests that the three envelope layers still invaginate coordinately in ftsNslm117 cells but that the entire process takes significantly longer to complete than normal.
Previous work showed that FtsN variants that either lacked the first 55 (12) or the last 181 (55) residues were still capable of carrying out its essential function in cell division. Together with the viability of ftsNslm117 cells, this indicated that the essential part of FtsN (EFtsN) might be confined to a small periplasmic juxtamembrane peptide of less than 65 residues.
To test this prediction, we performed complementation analyses with plasmids producing fusions of GFP to various portions of FtsN. The plasmids were derivatives of pMLB1113, and the synthesis of fusions was placed under the control of the lac regulatory region of the vector (Fig. (Fig.1;1; Table Table1).1). In one assay for FtsN function, plasmids were tested for their ability to correct the division defect of the FtsN depletion strain CH31 (araC Para::ftsN). The native ftsN promoter region in this strain is replaced with that of araBAD, such that ftsN transcription from the chromosome depends on the presence of arabinose in the medium (8). In a more stringent assay, plasmids were introduced in the wild-type strain TB28, and transformants were tested for their ability to survive P1-mediated introduction of a chromosomal ftsN<>aph (ΔftsN) knockout allele in the presence of IPTG. For each fusion capable of rescuing ΔftsN cells, the minimal concentration of IPTG yielding cells of close to normal size was then determined.
As summarized in Fig. Fig.1,1, production of GFP-FtsN1-319 (full length), GFP-FtsN1-243, GFP-FtsN1-105, or GFP-FtsN1-90 was sufficient to support division in either FtsN-depleted or ΔftsN cells (Fig. (Fig.3),3), but production of GFP-FtsN1-81 was not. Western analyses indicated that the latter fusion was not any less stable than the former (see Fig. S1A in the supplemental material). These results confirmed that most of the periplasmic portion of FtsN is dispensable and indicated that residues in the residue 81 to 90 interval are required for its essential function(s).
To assess if small periplasmic peptides covering this region are also sufficient, we next tested fusions to TTGFP, which were directed to the periplasm via the twin-arginine transporter (Tat) by use of the TorA signal peptide (Fig. (Fig.1).1). This showed that production of TTGFP-FtsN71-105 from pMG14 was indeed sufficient to support division of both FtsN-depleted (CH31) and ΔftsN (CH34) cells (Fig. (Fig.11 and and3D).3D). As might be expected, this periplasmic fusion supported cell division less efficiently than the functional transmembrane GFP-FtsN variants we tested. Thus, whereas ΔftsN cells carrying pCH201 (Plac::gfp-ftsN), pCH354 (Plac::gfp-ftsN1-243), pMG12 (Plac::gfp-ftsN1-105), or pMG47 (Plac::gfp-ftsN1-90) already attained a WT division phenotype in the presence of 5 μM IPTG, those carrying pMG14 (Plac::TTgfp-ftsN71-105) required at least 50 μM to do so. Western analyses indicated that the periplasmic (processed) form of TTGFP-FtsN71-105 reached a level that was about 25-fold higher than that of the transmembrane fusions under these conditions (see Fig. S1 in the supplemental material).
Expression of TTGFP-FtsN71-90 from pMG50 failed to correct division at any concentration of inducer. This was unexpected, as the transmembrane GFP-FtsN1-90 fusion encoded by pMG47 was fully capable of compensating for a complete lack of native FtsN (Fig. (Fig.11 and and3C).3C). Western analyses suggested that the TTGFP-FtsN71-90 fusion was not any less stable than TTGFP-FtsN71-105 (see Fig. S1B, lanes 3 and 4, in the supplemental material), so it is possible that FtsN function becomes dependent on residues in the 1 to 70 interval when residues in the 91 to 105 interval are lacking.
In any event, these results showed that the essential function of FtsN can be carried out by a surprisingly small periplasmic portion of the protein. This essential domain (EFtsN) comprises 35 residues (FtsN71-105) at most and likely includes most of the residues thought to form helix H2 (~FtsN80-93) (60). Those corresponding to H1 (~FtsN62-67) and H3 (~FtsN117-123), however, as well as the rest of the protein are dispensable for FtsN′s essential activity in cell division.
The cytoplasmic portion and most of the transmembrane helix of FtsN (FtsN1-45) can be replaced without abrogating its ability to localize to constriction sites, suggesting that a septal localization determinant resides in its periplasmic portion (1, 12).
As the small periplasmic EFtsN domain is sufficient to support division, it is also a good candidate for directing FtsN to the SR. However, we failed to detect any specific accumulation of the functional TTGFP-EFtsN fusion at the division sites of CH34/pMG14 (ΔftsN/Plac::TTgfp-ftsN71-105) cells. This was so at the minimal concentration of IPTG needed to stimulate a close-to-normal division frequency (50 μM), as well as at lower concentrations of inducer, at which cells became progressively more filamentous (Fig. (Fig.3D3D and data not shown). Rather, the fusion appeared dispersed throughout the periplasm, and similar fluorescent halos were seen when TTGFP-EFtsN was expressed in WT or ftsNslm117 cells, even at very low levels (Fig. (Fig.2D2D and data not shown). Western analyses indicated that the TTGFP-EFtsN fusion was not completely stable (see Fig. S1B, lane 3, in the supplemental material), raising the possibility that the dispersed periplasmic fluorescence was a consequence of proteolytic release of an intact GFP moiety. Although this caveat is difficult to rule out, the TTGFP-EFtsN fusion appeared no less stable than another TTGFP-FtsN fusion that did localize to division sites (TTGFP-SFtsN [see below]), rendering this possibility unlikely. Taken together, these results suggest that EFtsN by itself has little, if any, specific affinity for the division apparatus.
To more precisely identify the domain(s) of FtsN responsible for targeting to the septal ring, we inspected live TB28 (WT) cells and, where possible, CH34 (ΔftsN) cells producing various other GFP fusions by fluorescence microscopy. Fusions were qualitatively judged to have accumulated at (prospective) constriction sites only if fluorescence intensity at the site appeared markedly higher than at the remainder of the cell's periphery. For comparison, we also scored the localization patterns of ZipA1-183-GFP, a transmembrane ZipA derivative that lacks an FtsZ-binding domain and no longer accumulates at SRs. As seen before (36), this fusion appeared evenly distributed along the membrane in the large majority (96%; n = 381) of TB28/pCH178 (WT/Plac::zipA1-183-gfp) cells (Fig. (Fig.4B;4B; Table Table4).4). The remaining cells (n = 14) were scored as showing a notably higher signal at a cell constriction site than elsewhere. Although this could reflect some remaining affinity of the ZipA1-183-GFP fusion for the division apparatus in some cells, it more likely reflects inaccuracies in image aquisition and/or interpretation.
When produced as the sole FtsN derivative in CH34 (ΔftsN) cells, the fusion of GFP to full-length FtsN (GFP-FtsN) showed a clear ring-like accumulation at midcell in 42% of cells in an exponentially growing culture (Fig. (Fig.3A;3A; Table Table4).4). In a small subpopulation of these (5%), a ring was present in cells that did not yet show a visible sign of constriction when observed with differential interference contrast (DIC) optics. In all other cases, the ring coincided with readily detectible constrictions. Not all constricting cells were scored as showing a clear ring, and this was particularly so with deeply constricted cells. Although the fusion was always present at constriction sites, many of these cells showed an additional fluorescent signal along the entire membrane that was too high relative to the “ring” at the constiction site to confidently classify the latter as an obvious accumulation of the fusion.
The GFP-FtsN1-243 fusion lacks the SPOR domain, and GFP-FtsN1-90 additionally lacks the linker region and H3 (Fig. (Fig.1).1). Even though these fusions were fully capable of supporting division when produced in CH34 (ΔftsN) cells (Fig. (Fig.11 and and3)3) and did not appear any less stable than GFP-FtsN1-319 (see Fig. S1A in the supplemental material), they both showed an obvious and similar localization defect (Fig. 3B and C and Table Table4).4). Either fusion appeared to accumulate in rings in only a small subpopulation of cells (~15%), and none of these was at the early stage of constriction, when cell envelope invagination is still difficult to detect by DIC. The fusions appeared homogenously distributed along the membrane in the remaining cells, which, in addition to all unconstricted cells, included the majority (~75%) of constricted individuals. These results indicated that the SPOR domain is required for efficient accumulation of FtsN at constriction sites.
To assess if this domain (SFtsN) is also sufficient for septal targeting, we examined WT cells producing TTGFP-SFtsN, a periplasmic fusion containing just this domain and encoded on pMG4 (Plac::TTgfp-ftsN241-319) (Fig. (Fig.1).1). Indeed, this fusion localized sharply at sites of constriction (Fig. (Fig.4C).4C). A clear medial ring was observed in 60% of the total population, including 5% of unconstricted and in almost all (95%) constricted cells (Table (Table4).4). In contrast, cells expressing a fusion to an incomplete SPOR domain (TTGFP-FtsN281-319) (Fig. (Fig.1)1) only showed fluorescent halos (Fig. (Fig.4D).4D). Little, if any, of the intact periplasmic form of this fusion could be detected by immunoblotting, suggesting it was rapidly degraded (see Fig. S2, lane 5, in the supplemental material).
We conclude that the SPOR domain of FtsN (SFtsN) is an important septal localization determinant. Even though the SPOR-less GFP-FtsN1-243 and GFP-FtsN1-90 fusions localized poorly relative to GFP-FtsN and GFP-SFtsN, they did appear to accumulate to some degree at the constriction site in about one-fifth of constricting CH34 cells (Table (Table4).4). This suggests that the first 90 residues of FtsN include an additional septal localization determinant(s) but that any such determinant is significantly weaker than SFtsN.
As in ΔftsN cells (see above), production of TTGFP-EFtsN was sufficient to restore a normal division frequency in TB77 (ftsNslm117) cells, even though the fusion showed no sign of specific accumulation at constriction sites (Fig. 2C and D). Conversely, the TTGFP-SFtsN fusion failed to suppress the chaining phenotype of TB77, even though it accumulated sharply at constriction sites in these chains (Fig. (Fig.2F2F).
These results imply that the chaining phenotype of ftsNslm117 cells is caused by a partial deficiency in EFtsN activity, which could be due to instability of the FtsNslm117 peptide, a failure to concentrate this SPOR-less peptide at the presumed site of EFtsN action (the SR), or a combination of both. The inference that a reduction of EFtsN activity at the SR leads to a prolonged constriction period and cell chaining is consistent with the phenotype of cells with suboptimal FtsN levels (Fig. (Fig.5B)5B) (see below).
The sharp localization of TTGFP-SFtsN in TB77 (ftsNslm117) cells, furthermore, ruled out the possibility that the fusion accumulated at constriction sites via SPOR-SPOR interactions with native FtsN.
The results above suggested that the SPOR domain of FtsN aids cell division by helping to concentrate EFtsN at its site of action. If so, a poorly localizing SPOR-less FtsN derivative, such as GFP-FtsN1-243, would be expected to be less effective in restoring cell division to ΔftsN cells than full-length GFP-FtsN. In order to assess this with some accuracy, we used ΔftsN strains in which the GFP-FtsN and GFP-FtsN1-243 fusions were encoded by single-copy CRIM constructs that had been integrated at the chromosomal attHK022 site. These strains afforded better control over the production levels of the fusions than the corresponding strains encoding the fusions on multicopy plasmids described above. In Fig. Fig.5A,5A, cells of strains CH34(iMG59) [ΔftsN(Plac::gfp-ftsN)] and CH34(iMG60) [ΔftsN(Plac::gfp-ftsN1-243)] were subjected to spot titer analyses on LB agar supplemented with various concentrations of IPTG. This showed that ΔftsN cells encoding the SPOR-less fusion indeed required a significantly higher concentration of inducer for good growth (100 μM) than ones rescued by the fusion to native FtsN (25 μM). The division phenotypes of cells growing in liquid M9-glucose medium were consistent with this result, although the division phenotypes of both strains were less severe than on solid medium. In the absence of inducer, most CH34(iMG59) cells already grew as short chains and they resembled WT cells at 25 μM IPTG and above, suggesting that little gfp-ftsN expression is needed to support cell division under these conditions (Fig. (Fig.5B,5B, upper row). In contrast, most CH34(iMG60) cells formed long filaments with few constrictions in the absence of IPTG. At a 25 μM concentration of inducer, they roughly resembled the chains of CH34(iMG59) seen with no inducer, and they required 100 μM or more to resemble WT cells (Fig. (Fig.5B,5B, lower row). Western analyses indicated that these differences were not due to significant differences in the stability of the fusions and that cells required about fourfold more of the SPOR-less GFP-FtsN1-243 fusion than GFP-FtsN in order to divide normally (see Fig. S3 in the supplemental material).
While cells that lack EFtsN fail to constrict (11), SFtsN (or full-length FtsN) appeared to accumulate at the SR only at or after the onset of constriction (Table (Table4).4). Hence, septal localization of SFtsN appeared to depend on the essential activity of EFtsN. As in the context of the native protein septal localization of SFtsN will cause EFtsN to become concentrated at constriction sites as well, this predicts that FtsN accumulates at the SR by an unusual self-enhancing mechanism that requires both domains. To assess this scenario more directly, we studied the localization of SFtsN in the EFtsN depletion strain CH34/pMG20/pMG4 (ΔftsN/PBAD::TTbfp-ftsN71-105/Plac::TTgfp-ftsN241-319) in which production of a fusion of the essential domain to Tat-targeted blue fluorescent protein (TTBFP-EFtsN) can be modulated with arabinose and that of TTGFP-SFtsN with IPTG. In the presence of both inducers, cells appeared to divide normally, and TTGFP-SFtsN accumulated at each constriction site (Fig. (Fig.6B).6B). In the absence of arabinose (EFtsN−), however, TTGFP-SFtsN did not form regularly spaced rings, as might be expected if SFtsN could target SRs independently of EFtsN, but accumulated in rings only at rare constrictions still remaining in some of the filaments (Fig. (Fig.6C6C and data not shown). Western analyses indicated that the failure of TTGFP-SFtsN to form rings in the EFtsN-depleted filaments was not due to excessive degradation of the fusion (see Fig. S2, lanes 3 and 4, in the supplemental material). We conclude that SFtsN targeting to constriction sites normally indeed depends on the presence of EFtsN.
Two scenarios might explain the EFtsN-dependent accumulation of SFtsN at SRs. The first is that SFtsN can interact directly with EFtsN and that the latter acts as a pilot for the former. We considered this unlikely, because all GFP fusions that lacked SFtsN, especially TTGFP-EFtsN, localized poorly, if at all (Fig. (Fig.1;1; Table Table4).4). The second is that, even though EFtsN by itself fails to localize, it can still stimulate cell constriction when its concentration in the periplasm is sufficiently high, and SFtsN is attracted to some other feature of the constricting SR. The recent discovery that a mutant FtsA protein (FtsAE124A) allows cells to grow and divide in the complete absence of FtsN (5) provided us with a means to conclusively rule out the first senario.
To this end, we first replaced ftsA on the chromosome of strain TB28 with ftsAE124A and then introduced the ΔftsN (ftsN<>aph) knockout lesion by P1-mediated transduction. While cells of the BL18 (ftsAE124A) recipient strain divided at a normal frequency as expected (5), the resultant ftsAE124A ΔftsN transductant cells were very filamentous and barely viable (data not shown). The viability of transductants improved significantly, however, when recipients carried additional copies of the ftsAE124A allele, such as on plasmid pBL12. Even then, division was not completely normal, as cells of the resulting strain, BL20/pBL12 (ΔftsN ftsAE124A/ftsAE124A) displayed a chaining phenotype not unlike that of TB77 (ftsNslm117) (Fig. 6D and E).
We next compared the localization patterns of periplasmic TTGFP and TTGFP-SFtsN in BL20/pBL12 derivatives that carried a single copy of pBL5 (Plac::TTgfp) or pBL6 (Plac::TTgfp-ftsN241-319) integrated at attHK022. Expression of unfused TTGFP in strain BL20(iBL5)/pBL12 resulted in an homogenous fluorescent halo around each cell without obvious ring-like accumulations at constriction sites (Fig. (Fig.6E),6E), suggesting that chaining was due to slow constriction rather than to any defect that would cause a grossly expanded intermembrane distance at sites of constriction. In contrast, TTGFP-SFtsN accumulated at each constriction site of BL20(iBL6)/pBL12 cells (Fig. (Fig.6D).6D). We conclude that SFtsN is readily able to recognize these sites in the complete absence of any other domain of FtsN, including EFtsN.
Given that recruitment of SFtsN to division sites appeared to depend on the onset of the constriction phase of the fission process (Fig. (Fig.6C;6C; Table Table4)4) and that the SPOR domains of both E. coli FtsN (55) and B. subtilis CwlC (44) have an affinity for murein, it is reasonable to suppose that the feature of constriction sites that is recognized by SFtsN includes some form of septal murein.
To test this idea further, we monitored localization of TTGFP-SFtsN after treatment of cells with cephalexin, a β-lactam that specifically interferes with the murein transpeptidase activity of PBP3 (FtsI), which is specifically required for the synthesis of septal murein during cell constriction (56). As illustrated in Fig. Fig.6F,6F, TTGFP-SFtsN localized throughout the periplasm in cephalexin-treated filaments of TB28(iBL50) [WT(Plac::TTgfp-ftsN241-319)] cells, save for some very weak accumulations at some of the remaining constrictions. (Note that the integrative plasmid pBL50 is identical to pBL6, except that bla has been replaced with cat.) The failure of TTGFP-SFtsN to accumulate in rings under these conditions was not due to excessive degradation of the fusion (see Fig. S4, lanes 2 and 3, in the supplemental material). In contrast, and as expected (30, 31), a ZipA-GFP fusion accumulated in regularly spaced rings in cephalexin-treated TB28(iBL48) [WT(Plac::zipA-gfp)] filaments (Fig. (Fig.6G6G).
The purified periplasmic domain of FtsN (FtsN58-318) was previously found to preferentially bind the longest (>25 disaccharide units) glycan strands that remained upon amidase treatment of purified sacculi and to bind poorly to sacculi from cells that were deficient in murein amidase activities (55). This suggested that the septal localization of SFtsN might be impaired in amidase-deficient cells as well, and we tested this by using a set of isogenic strains lacking one, two, or all three of the amiA, -B, and -C genes.
TTGFP-SFtsN still accumulated sharply at constriction sites in cells of strains BL25 (ΔamiA), BL7 (ΔamiB), and BL9 (ΔamiC), as well as in the (relatively short) cell chains formed by strains BL10 (ΔamiB ΔamiC) and BL11 (ΔamiA ΔamiC) (Fig. (Fig.7A7A and data not shown), indicating that none of the amidases is specifically required for accumulation of SFtsN at septa. In the long cell chains formed by the triple knockout strain BL12 (ΔamiA ΔamiB ΔamiC), however, the localization pattern of TTGFP-SFtsN (Fig. (Fig.7C)7C) closely resembled that of unfused TTGFP (Fig. (Fig.7D).7D). Thus, fluorescence appeared throughout the periplasm, although some weak septal accumulations along the cell chains were apparent as well. As the latter were similarly apparent with the TTGFP control, we assume these reflected an increased intermembrane distance at such sites. Western analyses indicated that the lack of specific accumulation at septa was not due to excessive processing of TTGFP-SFtsN in strain BL12 (see Fig. S2, lane 2, in the supplemental material).
We conclude that efficient targeting of SFtsN to sites of constriction indeed requires the activity of at least one of the three periplasmic murein amidases in E. coli.
In addition to FtsN, three E. coli proteins with unknown functions (DamX, DedD, and RlpA) are predicted to also contain a C-terminal SPOR domain that resides in the periplasm.
Among these, DedD (named for downstream E. coli DNA D ) drew our particular attention because we also recovered a dedD::EzTn transposon insertion allele in a genetic screen for new cell shape/division mutants in a parallel project (4). This allele caused a distinct division phenotype that was recapitulated in strains in which dedD had been removed by targeted deletion (Fig. (Fig.8A).8A). Thus, both dedD::EzTn and ΔdedD cells showed a chaining phenotype that was reminiscent of that of ftsNSlm117 cells, but milder (Fig. 8B and D, Table Table3,3, and data not shown). Cell chaining was fully suppressible by expression of dedD or gfp-dedD from plasmid or lysogenic phage constructs (Fig. (Fig.8C8C and data not shown), implying that the phenotype was not due to polar effects of the dedD lesions and that the GFP-DedD fusion retained function (Fig. (Fig.8C8C and data not shown). Moreover, GFP-DedD accumulated in septal rings. Localization was especially sharp in constricting cells, although weaker accumulations were apparent in some preconstricting cells as well (Fig. (Fig.8C8C).
To assess if the SPOR domain of DedD (DedD140-220, SDedD) by itself was sufficient for septal localization, we monitored a TTGFP-SDedD fusion containing just this domain. Indeed, the fusion localized sharply to constriction sites, although it failed to correct the chaining phenotype of ΔdedD cells (Fig. (Fig.8D8D).
Notably, while the absence of DedD in otherwise WT cells caused only mild cell chaining, it proved lethal to ftsNslm117 cells. Initial attempts to introduce ΔdedD in ftsNslm117 cells, or vice versa, were fruitless, indicating that the combination was synthetically lethal. To elucidate the cause of lethality, we used depletion strains in which GFP-DedD is the sole source of DedD and its production can be modulated with IPTG. Figure Figure8E8E illustrates that strain MG14(λFB236) [ΔdedD(Plac::gfp-dedD)] grew equally well in the presence (row 1) or absence (row 3) of inducer, but that strain MG19(λFB236) [ftsNslm117ΔdedD(Plac::gfp-dedD)] indeed failed to propagate in its absence (row 4). Cells of the latter strain showed the typical chaining phenotype associated with ftsNslm117 in the presence of inducer (DedD+) (data not shown). However, they formed very long nonseptate filaments in its absence (DedD−) (Fig. (Fig.8F),8F), implying that DedD becomes essential for cell constriction in cells that produce only the truncated form of FtsN.
We conclude that DedD is a new nonessential division protein in E. coli that is required for efficient cell constriction and that becomes essential in ftsNslm117 cells.
The damX gene (37, 43) lies immediately upstream of that for DNA adenine methylase (dam) (Fig. (Fig.9A).9A). Like FtsN and DedD, DamX is a bitopic type II membrane protein with a large periplasmic domain. Earlier insertional mutagenesis of damX resulted in no obvious phenotype, although overexpression of the gene led to inhibition of division, suggesting some involvement of DamX in the division process (43). To test the prediction that DamX also associates with the septal ring, we monitored cells producing fluorescent fusions to the full-length protein and to the isolated SPOR domain (DamX338-428, SDamX) (Fig. (Fig.9A).9A). Both YFP-DamX (Fig. (Fig.9B)9B) and TTGFP-SDamX (Fig. (Fig.9C)9C) indeed localized sharply to division sites in WT cells, further implicating the protein in some division-associated function and showing that its SPOR also targets constriction sites autonomously.
Inspection of BL38 (ΔdamX) cells confirmed the original observation (43) that the absence of the protein has no obvious phenotype in otherwise wt cells (Table (Table33 and data not shown). However, introduction of ΔdamX into either ftsNslm117 or ΔdedD cells clearly aggravated their chaining phenotypes (Table (Table3).3). This aggravated phenotype of ΔdamX ftsNslm117 or ΔdamX ΔdedD double mutant cells could be suppressed to roughly that seen with ftsNslm117 or ΔdedD single mutants by expression of damX from plasmid pBL47 (Plac::damX), confirming that aggravation of the division defects observed in the double mutants was indeed caused by the absence of DamX (data not shown). In contrast, production of YFP-DamX failed to alleviate the chaining phenotypes of the double mutants, indicating that appending YFP to its N terminus interfered with DamX function (data not shown).
We conclude that DamX is also a new division protein in E. coli and that the protein contributes in a nonessential manner to the constriction process when FtsN or DedD function is compromised.
RlpA is an outer membrane lipoprotein (53), setting it apart from the other three E. coli SPOR domain proteins, which are all type II (N-in) bitopic inner membrane species. To monitor localization of the full-length protein, we used an RlpA-RFP (RlpA-mCherry) fusion encoded on lysogenic phage λFB241 (Plac::rlpA-rfp). To monitor that of its SPOR domain (RlpA281-362, SRlpA), we used a Tat-targeted fusion of GFP to this domain (Fig. 10A). As shown in Fig. 10B, the RlpA-RFP fusion accumulated at constriction sites as seen above for FtsN, DamX, and DedD. Unlike the latter proteins, however, RlpA-RFP additionally accumulated in a pattern of foci, arcs, or perhaps more extended stuctures along the rest of the cell envelope in both dividing and nondividing cells (Fig. 10B and C and data not shown). This suggests that a portion of RlpA molecules is recruited to the site of division while another portion remains associated with some other structures/complexes along the cell cylinder. In contrast, and much like the other four SPOR fusions studied above, the TTGFP-SRlpA fusion accumulated sharply at division sites and not in any other obvious nonhomogenous pattern (Fig. 10D). This indicates that, while SRlpA is sufficient to recognize division sites, the incorporation of full-length RlpA in the nonseptal accumulations involves other parts of the protein.
RlpA is not essential for cell viability (2, 23) and deletion of rlpA, by itself or when combined with ΔdamX, ΔdedD, and/or ftsNslm117, did not result in any obvious additional growth or division defects (Table (Table3).3). Thus, even though RlpA accumulates at the SR, whether it fullfills any role during the constriction process remains unclear.
Bacillus subtilis CwlC is a murein amidase that is produced at a late stage of endospore formation, where it contributes to lysis of the mother cell compartment and release of the mature spore (39, 41, 52). In addition to an N-terminal catalytic domain, the protein contains a C-terminal SPOR domain (Fig. 11A) that has affinity for amidase-digested B. subtilis murein, as inferred from nuclear magnetic resonance chemical shift analyses (44, 51).
To assess if a SPOR domain from a heterologous gram-positive source has similar properties in vivo as that of the four E. coli proteins, we examined the localization pattern of TTGFP-SCwlC, a Tat-targeted fusion containing just this domain (CwlC181-255), in various E. coli strains. Interestingly, this fusion localized very sharply to constriction sites (Fig. 11B to E). In addition, expression of the fusion induced chaining of WT cells, suggesting it may compete with DamX, DedD, and/or FtsN for a septal localization target (Fig. 11B).
These results suggest that all five SPOR domains examined here recognize the same or closely related target molecules that are highly enriched at constriction sites in E. coli. The fact that this includes a SPOR domain from a distantly related organism suggests this may be true for such domains from many other bacterial species as well.
Finally, as for SFtsN above (Fig. (Fig.55 and and6),6), we examined the localization patterns of the SPOR domains of CwlC, DamX, DedD, and RlpA in amidase-deficient chains and EFtsN-depleted filaments. As summarized in Table Table55 and illustrated for TTGFP-SCwlC in Fig. Fig.11,11, all four SPOR domains behaved similarly to that of FtsN in amidase mutants. Each still localized sharply to constriction sites in chains of single or double ami mutants but failed to do so in chains of the triple amiABC mutant (Fig. 11E and F, Table Table5,5, and data not shown).
However, we did observe distinct localization patterns in EFtsN-depleted filaments of strain CH34/pMG20. SRlpA behaved like SFtsN and appeared distributed throughout the periplasm (Table (Table5).5). Western analyses indicated that this was not caused by excessive degradation of the TTGFP-SRlpA fusion (see Fig. S5, lanes 7 and 8, in the supplemental material). In contrast, the other three SPOR domains still accumulated in readily discernible ring-like structures (Table (Table5),5), which are illustrated for TTGFP-SCwlC in Fig. 11C. In addition to relatively strong accumulations at remaining sites of constriction, the TTGFP fusions to these three SPOR domains also formed multiple weaker rings along the length of these filaments, and the spacing between these was often quite irregular (Fig. 11C and data not shown).
The finding that septal localization of these three SPOR domains appeared significantly less sensitive to depletion of EFtsN than that of FtsN itself suggested that their localization might also be less sensitive to interference with the activity of PBP3. We tested this for TTGFP-SCwlC by treatment of strain TB28(iBL57) [WT(Plac::TTgfp-ScwlC)] with cephalexin. In contrast to TTGFP-SFtsN (Fig. (Fig.6F),6F), TTGFP-SCwlC indeed still accumulated in one or more distinct rings in cephalexin-treated filaments (Fig. 11D).
We conclude that septal targeting of SCwlC, SDamX, and SDedD is significantly less sensitive to a lack of the essential division activity of FtsN than SFtsN and SRlpA. This raises the possibility that the former three recognize a septal target that is distinct from that of the latter two. On the other hand, we note that blocking cell constriction by EFtsN depletion or cephalexin treatment was not fully effective as judged by the presence of one or more mostly shallow constrictions in the resulting filaments. Therefore, it is also possble that all SPOR domains engage the same septal target but that SCwlC, SDamX, and SDedD have a significantly higher affinity for the target than SFtsN and SRlpA. In this case, one has to assume that sufficient EFtsN and PBP3 activities remained in the filaments to create enough of the target to efficiently attract the high-affinity but not the low-affinity SPORs. Either way, the results illustrate that different SPORs can have distinct properties that may be of physiological relevance.
We have shown that the essential function of E. coli FtsN in cell division can be performed by a small domain (EFtsN) of at most 35 residues centered about the presumed H2 helix in the periplasmic juxtamembrane region (60), and we have identified the C-terminal SPOR domain (SFtsN) as a strong septal localization determinant. In addition, we showed that three additional SPOR-containing proteins in E. coli, DamX, DedD, and RlpA, are also septal ring components and that DamX, and especially DedD, play a demonstrable role in the cell constriction process. We found that the isolated SPOR domains of all four E. coli proteins, as well as that of the heterologous B. subtilis murein amidase CwlC, accumulate sharply at sites of constriction, indicating this is a common property of such domains. In turn, this predicts that many of the SPOR domain proteins that can be identified in a wide range of bacterial species (20) are targeted to constriction sites as well. As the manuscript was being prepared for publication, this was indeed demonstrated for three additional SPOR-containing proteins from three additional species (45). One of these (CC_2007) is essential for division in Caulobacter crescentus and may be functionally equivalent to E. coli FtsN, as it shares many of its properties (45).
Our results have interesting implications for the role of FtsN in cell division and the manner by which it joins the division apparatus. The abilities of some of the SPOR-less GFP fusions to accumulate at constriction sites to some degree suggests that, besides SFtsN, an additional localization determinant(s) resides within the first 90 residues of FtsN (Table (Table4).4). However, SPOR-less fusions localized poorly at best, suggesting that any specific interactions between SPOR-less FtsN and SR-associated binding sites are weak or that the number of available sites is limiting.
We further found that the accumulation of SFtsN in a ring occurs at about the time cells initiate constriction and normally depends on EFtsN activity, although this dependency could be bypassed in cells producing the FtsAE124A protein. An interesting implication of these results is that FtsN joins the division apparatus in a self-enhancing manner, which is a property that is consistent with a role in triggering the constriction phase of the division process.
Taken together with existing biochemical evidence (44, 46, 55), our results suggest the following order of events at the onset of constriction. (i) Once the SR matures to the point where the murein transpeptidase PBP3 (FtsI) has joined the assembly, the SR is poised to initiate the constriction process. (ii) However, cell constriction requires a sufficient level of EFtsN activity at the SR, and the majority of FtsN is spread along the membrane prior to synthesis of any septal murein. (iii) Even so, a fraction of the FtsN molecules in the membrane interact with the poised SR structure, likely aided by weak specific interactions between the N-terminal portion of FtsN with other SR components (26), and their associated EFtsN domains stimulate a small amount of constriction and septal murein synthesis. (iv) This murein is processed by multiple activities, including murein amidases, leading to the generation of a sharply localized and transient form of septal murein that also constitutes a high-affinity binding substrate for SFtsN. (v) Accumulation of this substrate at the SR recruits additional FtsN molecules from elsewhere in the membrane by binding their SFtsN moieties. (vi) This, in turn, leads to an increased local concentration of EFtsN at the SR and an enhanced progression of steps iii to vi, and so on.
To understand why FtsN is normally essential for cell division, it is now pertinent to understand what EFtsN does. Although a self-enhanced accumulation of FtsN at division sites is consistent with a function in triggering constriction, the cell chaining phenotypes of cells with a partial EFtsN deficiency imply that EFtsN function is required to sustain the process as well. It seems unlikely that EFtsN acts as a structural stabilizer of the SR, as both its small size and its inability to noticeably accumulate at SRs by itself argue against it specifically interacting with multiple SR partners. Thus, even though full-length FtsN may be able to interact with multiple SR partners and this may help to stabilize the SR structure (10, 16, 17, 26, 38), this is probably not the reason why the protein is required for division. It is rather more likely that FtsN is essential because EFtsN is specifically required to stimulate the constriction process, as suggested in the scheme above. The small size of this domain also argues that its activity is more likely to be allosteric than catalytic, and it is attractive to speculate that an interaction between EFtsN and another essential SR component generates some signal or state that is required for constriction to start and proceed. One straightforward possibility is that EFtsN stimulates septal murein synthesis directly, for example, by activating the murein transpeptidase PBP3 (46). However, more circuitous routes are possible. For example, binding of EFtsN to a periplasmic domain of any of several essential transmembrane SR proteins may generate a stimulatory signal that is first transmitted to the Z ring assembly in the cytoplasm, the constriction of which then secondarily induces septal murein synthesis in the periplasm.
Whatever the mode of action of EFtsN, it is clear that FtsAE124A can, at least partially, compensate for its absence (5) (Fig. (Fig.6).6). FtsA is an attractive candidate for helping to coordinate septal ring activities in the cytoplasmic and periplasmic compartments during cell envelope invagination, and the mutant FtsAE124A protein in the cytoplasm may inadvertently induce the septal murein synthesis machinery in the periplasm to go ahead, even when EFtsN is lacking in the latter compartment (5).
What does SFtsN recognize, and how important is this to the division process? Although alternative mechanisms cannot be excluded, it seems likely that SFtsN and other SPOR domains are attracted to constriction sites by the local accumulation of some form of murein. As neither GFP-SFtsN (Fig. (Fig.4C)4C) nor any of the other four SPOR domains we studied (Fig. (Fig.99 to to11)11) lingered at newly created poles, this material must only be transiently available for binding at the constricting SR. The failure of all five SPOR domains to accumulate at constriction sites in cells lacking the three murein amidases AmiA, -B, and -C (Table (Table5)5) correlates well with the biochemical evidence that the periplasmic domain of FtsN (including SFtsN) bound poorly to sacculi from a similar mutant and that both this domain as well as the isolated SPOR domain of CwlC (SCwlC) still interacted with material that remained upon digestion of murein sacculi with purified amidases (44, 55). The simplest interpretation of these results is that SFtsN specifically recognizes glycan strands that have been, at least partially, denuded of peptide side chains (presumably both free and cross-linked) by amidase action (55). Such denuded strands are likely subject to (partial) detachment from the sacculus and/or to subsequent processing by lytic transglycosylases, which could be responsible for the transient nature of the SPOR-binding substrate. Another possibility is that SPOR domains recognize a murein geometry that is uniquely present at the SR and inherently transient in normally dividing cells, such as the junction between newly split and as-yet-unsplit septal murein.
In any event, our results suggest that the ability of SFtsN to recognize its binding substrate contributes to FtsN function by helping to concentrate EFtsN at the SR as soon as constriction begins and that suboptimal levels of EFtsN activity at the SR lead to increased constriction periods and cell chaining. Helping to increase EFtsN activity at the SR may be the only function of the SPOR domain, but it is attractive to speculate on additional roles. For example, FtsN seems well placed to also serve a quality control function during the constriction process. Thus, any problem in septal murein assembly or processing that reduces the availability or accessibility of the SFtsN substrate might well lead to a reduction in the concentration of EFtsN at the SR and a commensurate reduction in the constriction rate. The existence and importance of such a function may only become apparent under special circumstances, and further work will be needed to explore this possibility further.
Lastly, the identification of DamX, DedD, and RlpA as new SR components raises the issues of how and when they are recruited to the assembly and, especially, what their roles are. As with FtsN, their SPOR domains are likely dominant septal localization determinants, but additional domains may help recruit each of these proteins to the SR as well. The localization patterns of isolated SPOR domains in EFtsN-depleted filaments raise the possibility that DamX and DedD localize to division sites prior to RlpA and FtsN itself, but an accurate determination of their time and order of recruitment to the SR will require careful analyses.
Next to Pal, RlpA is now the second outer membrane lipoprotein found to accumulate at constriction sites. As part of the Tol-Pal system, Pal is needed for proper invagination of the OM during cell constriction (24). By connecting the OM with the underlying murein layer via its SPOR domain, RlpA could fulfill a similar function. However, we failed to detect any division defect in ΔrlpA mutants (Table (Table33 and data not shown). Whether RlpA has some redundant role in cell constriction itself or whether its accumulation at division sites is related to some other function remains to be seen. In addition to its septal localization, RlpA-RFP also accumulated into foci (or, possibly, more extended structures) along the cell periphery in both dividing and nondividing cells, suggesting involvement in a nondivision process as well. In this regard it may be relevant that rlpA lies in an operon with the genes for the cell shape proteins PBP2 (mrdA) and RodA (mrdB) (21, 53), although ΔrlpA cells showed no obvious shape defect (data not shown).
Whereas the function of RlpA remains unclear, our mutant analyses indicated that DamX and DedD are genuine cell division proteins. Otherwise-wild-type cells lacking DamX showed no obvious phenotype, but absence of the protein aggravated the division defects of ftsNslm117 or ΔdedD cells, suggesting that DamX plays a relatively minor accessory role in the division process that is redundant to that of FtsN and/or DedD.
The role of DedD was more obvious, as cells lacking the protein displayed a distinct chaining phenotype. Moreover, depletion of DedD caused lethal filamentation in ftsNslm117 cells, indicating that cells with limited EFtsN activty come to rely on DedD to accomplish fission. This could mean that DedD is required for full EFtsN function, or that it can partially subtitute for it, and it will be interesting to elucidate the role of DedD in more detail.
We thank Steven Sandler and Peter Setlow for materials and Andrea Möll, Martin Thanbichler, and David Weiss for communicating results prior to publication.
This work was supported by NIH GM57059 (to P.D.) and NIH NRSA Institutional Training Grant T32GM08056 (to F.B.). Thomas G. Bernhardt holds a Career Award in the Biomedical Sciences from the Burroughs Wellcome Fund.
Published ahead of print on 14 August 2009.
†Supplemental material for this article may be found at http://jb.asm.org/.