Research article | Open | Published:
The cyanobacterial cell division factor Ftn6 contains an N-terminal DnaD-like domain
BMC Structural Biologyvolume 9, Article number: 54 (2009)
DNA replication and cell cycle as well as their relationship have been extensively studied in the two model organisms E. coli and B. subtilis. By contrast, little is known about these processes in cyanobacteria, even though they are crucial to the biosphere, in utilizing solar energy to renew the oxygenic atmosphere and in producing the biomass for the food chain. Recent studies have allowed the identification of several cell division factors that are specifics to cyanobacteria. Among them, Ftn6 has been proposed to function in the recruitment of the crucial FtsZ proteins to the septum or the subsequent Z-ring assembly and possibly in chromosome segregation.
In this study, we identified an as yet undescribed domain located in the conserved N-terminal region of Ftn6. This 77 amino-acids-long domain, designated here as FND (Ftn6 N-T erminal D omain), exhibits striking sequence and structural similarities with the DNA-interacting module, listed in the PFAM database as the DnaD-like domain (pfam04271). We took advantage of the sequence similarities between FND and the DnaD-like domains to construct a homology 3D-model of the Ftn6 FND domain from the model cyanobacterium Synechocystis PCC6803. Mapping of the conserved residues exposed onto the FND surface allowed us to identify a highly conserved area that could be engaged in Ftn6-specific interactions.
Overall, similarities between FND and DnaD-like domains as well as previously reported observations on Ftn6 suggest that FND may function as a DNA-interacting module thereby providing an as yet missing link between DNA replication and cell division in cyanobacteria. Consistently, we also showed that Ftn6 is involved in tolerance to DNA damages generated by UV rays.
DNA replication and cell division are probably the most fundamental processes in the cell life cycle. Both proceed through a remarkably conserved general mechanism and are inextricably intertwined to each others and to the cell metabolism .
The DNA replication cycle can be divided into three distinct stages; initiation, elongation, and termination. Replication is initiated by the highly conserved AAA+ superfamily ATPase-member DnaA that binds the oriC, inducing DNA strand melting [2–4]. In E. coli, DnaA also interacts with the ring helicase DnaB and directs the loading of DnaB/DnaC onto the single stranded DNA (ssDNA) region. After binding of DnaB on the ssDNA region of the oriC, DnaC is released in an ATPase dependent manner. Then, DnaB recruits the DnaG primase and DNA polymerase III to form the replication fork . In B. subtilis, two additional essential proteins, called DnaB (be aware of the confusing nomenclature between the E. coli and B. subtilis) and DnaD, are engaged in entry of the ring helicase at oriC. DnaB could function as a membrane anchoring factor for the replication initiation machinery  or, together with DnaI, the functional homolog of E. coli DnaC , in the recruitment of the ring helicase . DnaD interacts with both DnaA and DnaB [8, 9]. It exhibits DNA remodelling activity, enhancing partial melting of the DNA strands, and could, therefore, function in early steps of replication such as initiating the recruitment of the ring helicase [9–15]. During elongation, the replication forks constituted at the oriC travel in opposite directions to achieve replication of the entire chromosome. When the replication forks reach the terminus, terC, the replication complexes are dismantled in a process involving specific termination factors .
The earliest event in bacterial cytokinesis is the definition of the future cell division site. This occurs through the dynamic assembly/disassembly of the Z-ring structure resulting from the self-polymerization of the ubiquitous tubulin-like protein FtsZ [17, 18]. Placement of the Z-ring is mainly dependent on the Min system both in E. coli and in B. subtilis [17, 18]. Once assembled, the Z-ring is believed to serve as a scaffold for recruitment of the cell division machinery to activate septation and physical separation of the daughter cells. In contrast to the model organisms E. coli and in B. subtilis, the molecular basis of cell division has not been as well studied in cyanobacteria. Nevertheless, studies with the two unicellular cyanobacteria Synechocystis PCC6803 and Synechococcus PCC7942 and the filamentous Anabaena PCC7120, have allowed the identification and the characterisation of clear Fts and Min orthologs as well as ZipN/Ftn2 and Ftn6, two cell division factors restricted to cyanobacteria [19–21]. Although ftn6 deletion leads to cell division defects, resulting in cells dramatically elongated in Synechococcus PCC7942 or enlarged in Anabaena PCC7120 [19, 21], the molecular function of Ftn6 remains unclear. Nevertheless, recent data suggest an involvement of Ftn6 in recruitment of FtsZ proteins to the septum or subsequent Z-ring assembly, as cells deleted for ftn6 do not exhibit condensed Z-rings, but rather diffuse localization of FtsZ .
Based on sequence and structure analyses, we here propose that the cyanobacterial-specific cell division factor Ftn6 contains a not hitherto described N-terminal domain related to the DnaD-like domain found in the DnaD chromosomal replication protein family. Identification of the Ftn6 N-terminal Domain, we termed FND (Ftn6 N-terminal Domain), opens up very interesting perspectives about Ftn6 function in cell division and possibly in chromosome segregation as well as on their necessary interplay. Consistently, we also showed that Ftn6-depleted cells are sensitive to DNA damages generated by UV rays.
Results and discussion
Ftn6 orthologs contain a conserved N-terminus domain (FND)
To identify putative Ftn6 motifs amenable for functional analysis, we performed database searches using the Synechocystis PCC6803 Ftn6 protein sequence (Syn6803) as query. BLAST search of the NCBI-database allowed the identification of 27 Ftn6 orthologs, all belonging to the cyanobacteria phylum. No ortholog was found in plastids or other prokaryotes. Interestingly, Ftn6 orthologs were found in all Nostocales (5 out of 5), Oscillatoriales (4/4) and Gloeobacterales (1/1), in some Chroococcales (17/29), but not in Prochlorales (0/13) even in those whose genome is fully sequenced. This finding, along with the viability of ftn6-depleted mutants [19, 21], suggests that other cell division factors functionally overlap with Ftn6 in cytokinesis. Alignment of all Ftn6 amino-acids sequences identified by BLAST (Additional file 1) revealed a single conserved region encompassed within the first 90 first amino-acids of Syn6803 Ftn6 (Figure 1). This 77 amino-acids-long domain (L18 to L94 in Syn6803 Ftn6), termed here FND for F tn6 N-terminal D omain, is bipartite with the first 28 amino-acids (L18 to L45) poorly conserved and the 49 remaining ones (W46 to L94) characterized by the W-X3-A-X2-E-X4-G-R-Y-X3-S-X4-L-X2-W consensus (Figure 1). The high degree of conservation of FND in Ftn6 orthologs suggests that this domain plays an important part of the function(s) of Ftn6 in cell division.
FND is related to the DnaD-like domain
Interrogation of protein domain databases did not allow identification of any FND-related domain (data not shown). Conversely, PSI-BLAST searches of the NCBI non-redundant database using the Syn6803 FND sequence as query identified significant hits for the alignment of this sequence with its orthologs and, interestingly, with members of the DnaD protein family. DnaD sequences that were reported from the 3th PSI-BLAST iteration share low, typically less than 20% identity, but significant similarities with Ftn6 orthologs (Figure 2). For instance, the Psi-BLAST returned alignments with highly significant E-values between Syn6803 FND and several members of the DnaD-like family (Geobacillus sp., GenBank:ZP_03559496, E-value 8e-15; Lactobacillus hilgardii, GenBank:ZP_03954546, E-value 1e-13; Clostridium beijerinckii, GenBank:YP_001308753, E-value 1e-10; Streptococcus mutans, GenBank: 2ZC2_A, E-value 4e-08 or Bacillus subtilis, GenBank:ABN10247, E-value 3e-08;...). DnaD consists of two domains with distinct biochemical properties . The N-terminal Domain is involved in the oligomerization of the protein and interactions with DnaA, while the C-terminus, listed in the PFAM database as the DnaD-like domain (pfam04271), binds DNA [8, 14]. Very interestingly, the sequence similarity we observed between the Ftn6 orthologs and the DnaD family is restricted to the DnaD-like domain (Figure 2), suggesting a DNA-binding activity for FND.
The belonging of FND to the DnaD-like domain family was further supported by fold recognition techniques. Indeed, submission of each FND domains to the PHYRE server constantly returned the two DnaD structures currently deposited in PDB as best hits, i.e. the DnaD-like domain of the replication proteins from Streptococcus mutans (PDB: 2ZC2) and from Enterococcus faecalis (PDB: 2I5U). DnaD-like domains reported from the PHYRE search share low but noticeable similarities with all FND tested (data not shown), suggesting structural similarities between FND and the DnaD-like domain. Then, Streptococcus mutans DnaD-like domain has been included in the alignment shown in Figure 2. As expected for proteins sharing a low level of sequence identity, we noticed that the nature of the hydrophobic residues conserved in DnaD-like domains (noted by red dots below the WebLogo profile  shown in the additional file 2 and shaded in pink background in Figure 2) was not preserved in FND (shaded by light green background in Figure 2). By contrast, their positions were highly conserved. The high degree of conservation (80%; 16 aminoacids residues out of 20; compare positions shaded in pink and green in the bottom of Figure 2) of the hydrophobic pattern between FND and DnaD-like domain strongly argues in favour of a similar fold. This is particularly evident for the helices 3 and 4, in which the hydrophobic pattern is not only conserved in position (83%; 10 out of 12), but is also highly similar, particularly the two alanines of the third helix (A50 and A54 in Syn6803 FND) and the leucine and the tyrosine (L69 and W72 respectively) at the extreme C-terminus of the fourth helix (Figure 2).
Based on these results and the alignment shown in Figure 2, we constructed a homology 3D-model of Syn6803 FND with MODELLER  Normalized DOPE z-score: -0.533). Overall model quality assessed by ProSA-Web returned a Z-Score of -4.12, which is in agreement with the Z-Scores of all experimentally determined chains currently deposited in the PDB database. Most of the hydrophobic positions of FND conserved in the DnaD-like domain (Figure 2 and additional file 2) are buried within the structure (Figure 3A), emphasizing their importance for structure stability and/or folding. Note that two hydrophobic positions conserved in DnaD-like domains (noted by a star in the Figure 2), but missing in FNDs, are not buried in the DnaD-like domains structure (data not shown) and, hence, are probably not required for the folding of this domain. The highly conserved G58R59Y60 [K/R]61 motif that is specific for FNDs (Figures 1 and 2), localizes in a loop between the helices H3 and H4 (Figures 1 and 3B). The tyrosine is buried within the structure, whereas the glycine and the two basic amino-acids are exposed on the surface of FND. Consurf-based mapping of the evolutionarily conserved residues exposed on the surface of the Syn6803 FND domain  (Additional file 3) show that G58, R59 and K61 residues cluster with F24 and D25 from the first helix, E53, L55 and Y56 from the third helix and S64 from the fourth helix. All these residues are either strictly or functionally conserved (Figure 1) and hence, could be engaged in Ftn6-specific interactions.
Functional prediction for Ftn6
So far, DnaD-like proteins have only been found in some low G+C content gram positive bacteria and their associated phages , where they exhibit pleiotropic functions all related with DNA metabolism. For instance, DnaD was shown to be involved in initiation of chromosome and plasmid replication [25, 26], sporulation , DNA repair  and recombination . Furthermore, the DnaD-related protein from the thermophilic bacteriophage GBSV1 exhibits an unspecific nuclease activity . The exact function of the DnaD-like domain in these processes remains unclear, but the DnaD-like domain from B. subtilis was found to exhibit DNA-binding and DNA-remodelling activities [11–15]. Altogether, these data strongly suggest that the DnaD-like domain does not define a common structural fold occurring in functionally unrelated proteins, but rather that the DnaD-like domain-containing proteins, including Ftn6, share common functions involving DNA.
What is the function of Ftn6? FND could suggest a function in DNA replication for Ftn6. However, this hypothesis is unlikely as Ftn6-depleted mutants do not appear to affect chromosome replication and do not produce anucleate cells . Furthermore, the N-terminal extension in DnaD proteins, which interacts with DnaA , is missing in Ftn6 orthologs (data not shown). Alternatively, Ftn6 could function in the cross talk between chromosome replication and cell division, a fundamental biological process not yet investigated in cyanobacteria. In most bacteria, both processes are intimately co-ordinated, as formation and placement of the future division septum is regulated by nucleoid occlusion and only occurs after replication of a significant portion of the chromosome . By contrast, Z-ring can assemble at nucleoid-occupied sites and nucleoid separation occurs during Z-ring constriction in cyanobacteria . This lack of nucleoid occlusion supposes an efficient mechanism to segregate chromosome trapped at the midcell during Z-ring constriction. It has recently been proposed that Ftn6 could be involved in chromosome segregation in Synechococcus PCC7942 . How Ftn6 is functionally connected to chromosome segregation remains unknown. Nevertheless, identification of the putative DNA-binding domain, FND, strongly supports the involvement of Ftn6 in this pathway and its interplay with cell division.
To test this hypothesis, we reasoned that defective chromosomal segregation should generates DNA damages thereby increasing the sensitivity of the cells to DNA damaging agents. As expected, we found that the Ftn6-depleted mutant we recently constructed  is significantly more sensitive to UV rays than the wild-type strain (Figure 4). Together, our findings strengthen both the functional relationship between DnaD-like domain-containing proteins and DNA metabolism, and the potential function of Ftn6 at the interface between DNA replication and cell division.
Although depletion of Ftn6 leads to cell division defects, the molecular function of this cynobacterial-specific divisome component remained unclear. Sequence alignment of Ftn6 orthologs beforehand identified by BLAST allowed us to uncover a new conserved domain localized within the N-terminus of the proteins. Combining several approaches, we then shown that this domain, designated here as FND, exhibits sequence and structure similarities with the DnaD-like domains found in several factors involved in DNA metabolism. The structure similarities between FND and DnaD-like domains together with the sensitivity of the Ftn6-depleted mutant to UV rays, led us to propose that Ftn6 is functionally linked to DNA metabolism, possibly playing a role at the interface between DNA replication and cell division. Whether this function involves or not other cell division factors and what is (are) the DNA target(s) of Ftn6 remain to be determined.
In silico methods
Databases search of Ftn6 and DnaD domain-containing proteins were performed using BLAST (e < 10-4) [33, 34] and PsiBLAST [34, 35] algorithms. Multiple sequence alignment of the DnaD-like-containing proteins or/and Ftn6 orthologs were generated using ClustalW2 [36, 37] (Matrix: BLOSUM, Gap penality: 10 and penality for Gap extension: 0,1), and visualized with Boxshade . Further details are given in the relevant figure legends and in the additional Files 1 and 4. Fold recognition was performed with PHYRE [39, 40]. 3D-structure of Syn6803 FND was modelled using the MODELLER 9v6 program  and visualized with Pymol . Briefly, 10 models of Syn6803 FND were first built based on the alignment shown in Figure 2. All 10 models were then evaluated with DOPE from the MODELLER package and the best chosen as final model. The overall model quality was additionally validated with ProSA-Web [42, 43].
WT Synechocystis PCC6803 and its derivative ftn6Δ::Kmr/FTN6+  were grown as described . Cells were then 4-fold serially diluted in MM medium and then spotted onto MM plates. Finally, the plates were or not exposed to either 250 or 500 J.m-2 UV radiation and incubated 7 days at 30°C under the above described light conditions.
Haeusser DP, Levin PA: The great divide: coordinating cell cycle events during bacterial growth and division. Curr Opin Microbiol 2008, 11: 94–9. 10.1016/j.mib.2008.02.008
Messer W: The bacterial replication initiator DnaA. DnaA and oriC, the bacterial mode to initiate DNA replication. FEMS Microbiol Rev 2002, 26: 355–74.
Kaguni JM: DnaA: controlling the initiation of bacterial DNA replication and more. Annu Rev Microbiol 2006, 60: 351–75. 10.1146/annurev.micro.60.080805.142111
Schaeffer PM, Headlam MJ, Dixon NE: Protein – protein interactions in the eubacterial replisome. IUBMB Life 2005, 57: 5–12. 10.1080/15216540500058956
Rokop ME, Auchtung JM, Grossman AD: Control of DNA replication initiation by recruitment of an essential initiation protein to the membrane of Bacillus subtilis. Mol Microbiol 2004, 52: 1757–67. 10.1111/j.1365-2958.2004.04091.x
Soultanas P: A functional interaction between the putative primosomal protein DnaI and the main replicative DNA helicase DnaB in Bacillus. Nucleic Acids Res 2002, 30: 966–74. 10.1093/nar/30.4.966
Velten M, McGovern S, Marsin S, Ehrlich SD, Noirot P, Polard P: A two-protein strategy for the functional loading of a cellular replicative DNA helicase. Mol Cell 2003, 11: 1009–20. 10.1016/S1097-2765(03)00130-8
Ishigo-Oka D, Ogasawara N, Moriya S: DnaD protein of Bacillus subtilis interacts with DnaA, the initiator protein of replication. J Bacteriol 2001, 183: 2148–50. 10.1128/JB.183.6.2148-2150.2001
Bruand C, Velten M, McGovern S, Marsin S, Sérèna C, Ehrlich SD, Polard P: Functional interplay between the Bacillus subtilis DnaD and DnaB proteins essential for initiation and re-initiation of DNA replication. Mol Microbiol 2005, 55: 1138–50. 10.1111/j.1365-2958.2004.04451.x
Marsin S, McGovern S, Ehrlich SD, Bruand C, Polard P: Early steps of Bacillus subtilis primosome assembly. J Biol Chem 2001, 276: 45818–25. 10.1074/jbc.M101996200
Turner IJ, Scott DJ, Allen S, Roberts CJ, Soultanas P: The Bacillus subtilis DnaD protein: a putative link between DNA remodeling and initiation of DNA replication. FEBS Lett 2004, 577: 460–4. 10.1016/j.febslet.2004.10.051
Zhang W, Carneiro MJ, Turner IJ, Allen S, Roberts CJ, Soultanas P: The Bacillus subtilis DnaD and DnaB proteins exhibit different DNA remodelling activities. J Mol Biol 2005, 351: 66–75. 10.1016/j.jmb.2005.05.065
Zhang W, Allen S, Roberts CJ, Soultanas P: The Bacillus subtilis primosomal protein DnaD untwists supercoiled DNA. J Bacteriol 2006, 188: 5487–93. 10.1128/JB.00339-06
Carneiro MJ, Zhang W, Ioannou C, Scott DJ, Allen S, Roberts CJ, Soultanas P: The DNA-remodelling activity of DnaD is the sum of oligomerization and DNA-binding activities on separate domains. Mol Microbiol 2006, 60: 917–24. 10.1111/j.1365-2958.2006.05152.x
Zhang W, Machón C, Orta A, Phillips N, Roberts CJ, Allen S, Soultanas P: Single-molecule atomic force spectroscopy reveals that DnaD forms scaffolds and enhances duplex melting. J Mol Biol 2008, 377: 706–14. 10.1016/j.jmb.2008.01.067
Neylon C, Kralicek AV, Hill TM, Dixon NE: Replication termination in Escherichia coli: structure and antihelicase activity of the Tus-Ter complex. Microbiol Mol Biol Rev 2005, 69: 501–26. 10.1128/MMBR.69.3.501-526.2005
Harry E, Monahan L, Thompson L: Bacterial cell division: the mechanism and its precison. Int Rev Cytol 2006, 253: 27–94. 10.1016/S0074-7696(06)53002-5
Lutkenhaus J: Assembly dynamics of the bacterial MinCDE system and spatial regulation of the Z ring. Annu Rev Biochem 2007, 76: 539–62. 10.1146/annurev.biochem.75.103004.142652
Koksharova OA, Wolk CP: A novel gene that bears a DnaJ motif influences cyanobacterial cell division. J Bacteriol 2002, 184: 5524–8. 10.1128/JB.184.19.5524-5528.2002
Mazouni K, Domain F, Cassier-Chauvat C, Chauvat F: Molecular analysis of the key cytokinetic components of cyanobacteria: FtsZ, ZipN and MinCDE. Mol Microbiol 2004, 52: 1145–58. 10.1111/j.1365-2958.2004.04042.x
Miyagishima SY, Wolk CP, Osteryoung KW: Identification of cyanobacterial cell division genes by comparative and mutational analyses. Mol Microbiol 2005, 56: 126–43. 10.1111/j.1365-2958.2005.04548.x
Crooks GE, Hon G, Chandonia JM, Brenner SE: WebLogo: a sequence logo generator. Genome Res 2004, 14: 1188–90. 10.1101/gr.849004
Eswar N, Eramian D, Webb B, Shen MY, Sali A: Protein structure modeling with MODELLER. Methods Mol Biol 2008, 426: 145–59. full_text
Landau M, Mayrose I, Rosenberg Y, Glaser F, Martz E, Pupko T, Ben-Tal N: ConSurf 2005: the projection of evolutionary conservation scores of residues on protein structures. Nucleic Acids Res 2005, 33: W299–302. 10.1093/nar/gki370
Bruand C, Sorokin A, Serror P, Ehrlich SD: Nucleotide sequence of the Bacillus subtilis dnaD gene. Microbiology 1995, 141: 321–322. 10.1099/13500872-141-2-321
Bruand C, Ehrlich SD, Jannière L: Primosome assembly site in Bacillus subtilis. EMBO J 1995, 14: 2642–2650.
Lemon KP, Kurtser I, Wu J, Grossman AD: Control of initiation of sporulation by replication initiation genes in Bacillus subtilis. J Bacteriol 2000, 182: 2989–2991. 10.1128/JB.182.10.2989-2991.2000
Li Y, Kurokawa K, Matsuo M, Fukuhara N, Murakami K, Sekimizu K: Identification of temperature-sensitive dnaD mutants of Staphylococcus aureus that are defective in chromosomal DNA replication. Mol Genet Genomics 2004, 271: 447–457. 10.1007/s00438-004-0996-6
Bruand C, Farache M, McGovern S, Ehrlich SD, Polard P: DnaB, DnaD and DnaI proteins are components of the Bacillus subtilis replication restart primosome. Mol Microbiol 2001, 42: 245–255. 10.1046/j.1365-2958.2001.02631.x
Song Q, Zhang X: Characterization of a novel non-specific nuclease from thermophilic bacteriophage GBSV1. BMC Biotechnol 2008, 8: 43. 10.1186/1472-6750-8-43
Koksharova OA, Klint J, Rasmussen U: Comparative proteomics of cell division mutants and wild-type of Synechococcus sp. strain PCC 7942. Microbiology 2007, 153: 2505–17. 10.1099/mic.0.2007/007039-0
Marbouty M, Saguez C, Cassier-Chauvat C, Chauvat F: Characterization of the FtsZ-interacting septal proteins SepF and Ftn6 in the spherical-celled cyanobacterium Synechocystis PCC6803. J Bacteriol 2009.
Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ: Basic local alignment search tool. J Mol Biol 1990, 215: 403–10.
BLAST and PsiBLAST at NCBI[http://blast.ncbi.nlm.nih.gov/Blast.cgi]
Altschul SF, Madden TL, Schäffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ: Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res 1997, 25: 3389–402. 10.1093/nar/25.17.3389
Thompson JD, Higgins DG, Gibson TJ: CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position specific gap penalties and weight matrix choice. Nucleic Acids Res 1994, 22: 4673–4680. 10.1093/nar/22.22.4673
EBI: ClustalW2 server.[http://www.ebi.ac.uk/Tools/clustalw2/index.html]
Bennett-Lovsey RM, Herbert AD, Sternberg MJ, Kelley LA: Exploring the extremes of sequence/structure space with ensemble fold recognition in the program Phyre. Proteins 2008, 70: 611–25. 10.1002/prot.21688
Wiederstein M, Sippl MJ: ProSA-web: interactive web service for the recognition of errors in three-dimensional structures of proteins. Nucleic Acids Res 2007, 35: W407–10. 10.1093/nar/gkm290
Marbouty M, Mazouni K, Saguez C, Cassier-Chauvat C, Chauvat F: Characterization of the Synechocystis PCC6803 penicillin-binding proteins and cytokinetic proteins FtsQ and FtsW, and their network of interactions with ZipN. J Bacteriol 2009, 191: 5123–33. 10.1128/JB.00620-09
We are particularly indebted to Isabelle Callebaut, Fransisco Malagon and Silvia Jimeno-Gonzalez for their critical reading of the manuscript. This work was supported by grants from the Commissariat à l'Energie Atomique (CEA). CS and MM were supported by CEA post-doctoral and doctoral fellowships respectively.
MM and CS collected the data, participated in the sequences alignments and the modelling of the 3D-structures. FC and CS conceived the study, carried out the analysis of the data and drafted the manuscript. FC coordinated the study. All authors read and approved the final manuscript.
Martial Marbouty, Cyril Saguez contributed equally to this work.