Alkaliphilic Bacillus species show potential application in concrete crack repair by virtue of rapid spore production and germination then extracellular calcite formation

Characterization of alkaliphilic Bacillus species for spore production and germination and calcite formation as a prelude to investigate their potential in microcrack remediation in concrete.


Introduction
Metabolic activities in microbes yield insoluble organic and inorganic materials, intra-or extracellularly (Lappin-Scott et al. 1988). The processes that lead to production by living organisms of inorganic material such as phosphorites, carbonates, silicates and iron and manganese oxides in the form of shells, skeletons and teeth are termed as bio-mineralization (Beveridge et al. 1983;Ghiorse 1984). Bio-calcification involves precipitation of calcium carbonate polymorphs and occurs commonly in soil, fresh water and marine environments (Shirakawa et al. 2011).
Research into and potential application of bio-calcification has included restoration of limestone on historic monuments and ornamental stone (Rodriguez-Navarro et al. 2003), soil stabilization and microbiologically enhanced crack repair (De Muynk et al. 2010). More generally, CaCO 3 precipitation induced by bacterial activity can increase stability of structures in civil engineering (Sarmast et al. 2014). The methods mimic what has been occurring naturally, since many carbonate rocks have been cemented over eons by microbe-induced calcium carbonate precipitation (Rodriguez-Navarro et al. 2003).
The most widely accepted hypotheses related to CaCO 3 precipitation are based on accumulation of unused calcium ions in the extracellular medium (Rodriguez-Navarro et al. 2003;Shirakawa et al. 2011). There are at least two pathways for microbial precipitation of CaCO 3 : passive and active. Ureolytic bacteria degrade urea to yield CO 2 and NH 3 , then accumulation of NH 3 in the vicinity increases pH resulting in passive deposition of CaCO 3  . Active CaCO 3 precipitation is a result of ion exchange through the cell membrane and activation of Ca 2+ and/or Mg 2+ ion pumps, possibly combined with carbonate ion production (Le M etayer-Levrel et al. 1999). In some micro-organisms a series of bioprocesses like photosynthesis, ammoxification, denitrification, sulphate reduction and anaerobic sulphide oxidation induce extracellular precipitation of carbonate (Baskar et al. 2009). Calcite, vaterite and aragonite are three polymorphs of calcium carbonate precipitated by bacteria (Wei et al. 2015).
The strength and relatively low cost of concrete explain it as the most widely used construction component worldwide (Jonkers et al. 2010). However, the low tensile strength of concrete and its susceptibility to cracking often compromise the structural integrity. Although crack widths smaller than 0Á2 mm do not impose a structural threat, ingress of chlorides, sulphates and acids result in corrosion of steel reinforcement, or expansion of the hardened cement paste, and this can lead to catastrophic structural damage (Edvardsen 1999;Reinhardt and Jooss 2003). Enormous expenditure is required for maintenance and repair of cracks. It is estimated that the United States spends 4 billion dollars annually on concrete highway bridges as an outcome of reinforcement corrosion (Koch et al. 2001).
In order to create a sustainable and cost effective alternative, microbiologically induced calcite precipitation is being studied for application in healing of cracks in concrete (van Tittelboom et al. 2010). Bacteria in the form of Bacillus species endospores can be added to the concrete mix (naked or encapsulated) while casting but they must withstand extreme alkaline pH values, dehydrating condition within concrete and survive high mechanical compressive forces during concrete curing (Jonkers et al. 2010). Previous investigations on sealing of cracks with bacterially precipitated CaCO 3 have involved Bacillus cohnii, B. pasteurii (syn, Sporosarcina), B. sphaericus and B. alkalinitrilicus (Bang et al. 2010;Jonkers et al. 2010;van Tittelboom et al. 2010;Wiktor and Jonkers 2011). Calcium carbonate precipitation on application of immobilized Sporosarcina pasteurii cells to the cracks enhanced concrete performance (Bang et al. 2010) and Shewanella species have also been shown to achieve crack remediation (Ghosh et al. 2005), but these species do not produce endospores. Application of Myxococcus xanthus to restore surfaces of ornamental stones induces new carbonate crystal grains that are more durable than the original limestone (Rodriguez-Navarro et al. 2003). These various studies have typically taken an engineering perspective and often neglected optimization and understanding of the basic microbiology and calcite formation required for the potentially important application of these refractile, crystal-inducing, bacterial strains.
In this work we have avoided use of ureolytic bacteria, because oxidation of urea to nitrous acid and then nitric acid can lead to severe corrosion of building materials (Allan 1999). We describe, often for the first time in the context of bacteria used in concrete remediation, selection from three alkaline-tolerant Bacillus sp., evaluation of speed and extent of endospore formation, triggers for spore germination, ability of spores to survive in concrete, and the kinetics, amounts and structural confirmation of calcite production. These features are prerequisites to function, in combination with suitable nutrients and precursors, to create a product effective in crack healing. The findings are highly encouraging because spore production could be readily commercialized, spores survive the formation of mature concrete and bacterial cells are efficient producers of calcite not only in vitro but also in situ.

Bacterial isolates and growth
Three alkalinophilic species Bacillus pseudofirmus DSM 8715, B. cohnii DSM 6307 and Bacillus halodurans 497 were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ), Braunschweig, Germany and stored in 50% (v/v) glycerol at À80°C. They were routinely cultured in buffered Luria-Bertani (LB) broth which contained 100 ml l À1 Na-sesquicarbonate (Na-sesquicarbonate composition per litre, NaHCO 3 42 g and Na 2 CO 3 anhydrous 53 g) to achieve pH 9Á5.

Spore preparation, sporulation and germination
Spores were prepared in a sporulation medium and resulting spores were harvested after 48 h growth by centrifugation (Jonkers et al. 2010). The spore pellet was rinsed thrice with chilled 10 mmol l À1 Tris-HCl buffer pH 9 followed by chlorohexidine digluconate treatment (0Á3 mg ml À1 , 30 min) to kill vegetative cells and further washed thrice with chilled 10 mmol l À1 Tris-HCl buffer pH 9. For studying spore formation, spore samples were stained (Schaeffer and Fulton 1993).
Spore germination was determined by coupling: (i) a qualitative procedure involving the percentage decrease in OD 600 nm of spore suspensions during germination (Kudo and Horikoshi 1983;Hornstra et al. 2006) and (ii) direct counting of refractile, nongerminated spores by phase-contrast microscopy (91000). Spores were suspended in buffer (Tris-HCl, 0Á03 mol l À1 , pH 9Á5) containing germination triggers 10 lmol l À1 inosine and L-alanine, and 0Á1 mol l À1 NaCl. They were incubated at 30°C in a final volume of 3 ml and adjusted to OD 0Á25 before assessments at 15, 30, 45, 60, 90 and 120 min, then 16 h. Phase-contrast microscopy of freshly harvested spores revealed c. 10% dark and nonrefractile vegetative cells, which could be mistaken for germinated spores. Therefore, before testing potential stimulants of spore germination, viable vegetative cells were selectively killed with chlorhexidine digluconate. Post treatment, bacterial spores appeared bright and refractile.

Preparation of cement paste specimens and viability of bacterial spores
The survivability of spores in cement paste was studied by adding spores to cement at the same time as the addition of water. Freeze drying involved washing, as described earlier, spore pellet thrice with 10 mmol l À1 Tris-HCl pH 9, then chlorohexidine digluconate, then a further triple wash with Tris-HCl before lyophilization at À40°C in an Edwards Modulyo Freeze Dryer. Dried spores were stored desiccated at 4°C. Fifty milligrams of freeze-dried spore powder was added to cement (sterilized at 100°C overnight) made to a paste with sterile distilled water, ratio 1 : 2. This paste was placed in moulds (45 9 45 9 5 mm) and demoulded after 24 h and cured at room temperature (c. 20°C) in sterile conditions. The moulds had been first surface sterilized, once with chlorohexidine digluconate then with 70% ethanol. The specimens contained either 1Á21 9 10 6 spores per gram of B. pseudofirmus, or 1Á44 9 10 6 spores per gram of B. cohnii or 4Á4 9 10 6 spores per gram of B. halodurans. Three cement pastes were made for each of the bacterial species. The cement used was a Portland fly ash cement, CEM II/ B-V 32.5R, conforming to BS EN 197-1. All specimens were stored in a sterile container, which was opened under sterile conditions. Spore viability was estimated by total viable count from cement specimens aged between 1 and 93 days. One gram of each specimen was comminuted using sterile mortars and pestles, thoroughly vortexed to enhance spore extraction, then 1-ml aliquots were serially diluted (in triplicate) in sterile Tris-HCl buffer pH 9Á5. Aliquots of suitable dilution were plated on LB alkaline agar plates at 30°C overnight and total viable count recorded after 18 h.

Selection of an appropriate medium for quantification of CaCO 3 formation
One millilitre of a culture of B. pseudofirmus DSM 8715 grown overnight in LB broth at 30°C, 150 rev min À1 was seeded to 100 ml of the following two media: 1/10th strength LB acetate medium (LBA, containing per litre tryptone 1 g, yeast extract 0Á5 g, NaCl 1 g, calcium acetate 2Á5 g); B4 medium containing per litre (yeast extract 4 g, glucose 5 g, calcium acetate 2Á5 g). All media were adjusted to pH 8 with NaOH.

Quantification of calcite precipitation in vitro
Collection and processing of crystals from liquid cultures was the same for subsequent analyses by atomic absorption spectroscopy (AAS), Fourier transform-infrared spectroscopy (FTIR), Raman spectroscopy, scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDX). Crystals were extracted daily until 8 days for AAS and after 8 days for other analytical methods. Cultures were inoculated and incubated as above, then fluids from triplicate cultures were centrifuged (3800 g, 10 min) and pellets washed once with 1/10th LB without Ca acetate and then twice with SDW; the washed pellet was then filtered on nitrocellulose membranes (3 lm, 45 mm diam.). Crystals from the membrane were collected, washed thrice with SDW then air-dried. The dried crystals were dissolved in 0Á6 mol l À1 HCl (Sch€ afer et al. 2011) added slowly to avoid effervescence, then analysed for elemental Ca 2+ by AAS (Perkin Elmer AAnalyst 100, Beaconsfield, Bucks, UK) with an acetylene nitrous oxide flame at wavelength 422Á7 nm; slit width 0Á7 nm. All samples standards and blank contained 2000 ppm potassium as an ionization suppressant.
Chemical and physical characterization of bacterial crystalline precipitates

FTIR
The crystals from 8-day cultures were subjected to FTIR and all spectra were recorded in the range 4000-600 cm À1 , using a PerkinElmer Frontier NIR/IR FTIR spectrometer with a Pike Technology Miracle ATR sampling accessory. Each spectrum was the average of five scans using a spectral resolution of 2 cm À1 .The IR spectra were recorded and stored with software Perki-nElmer Spectrum (ver. 10.03.09). Spectra were matched with a reference spectrum of calcite from the RRUFF database (http://rruff.info/Calcite/R050128) (Lafuente et al. 2015).

Raman spectroscopy
Raman spectra were recorded using a Renishaw inVia confocal Raman Microscope with a 509 objective lens. Measurements were taken using a 785 nm laser in high confocality mode (20 lm slit width) with a 1200 l mm À1 grating. Laser power was 100% using a 10-s exposure with 100 accumulations.

SEM and EDX
Morphology and characteristics of the crystals was determined by SEM (JSM 6480LV (JEOL, Welwyn, Herts, UK)): resolution, 3 nm, magnification, 8-300 000, accelerating voltage, 0Á3-30 kV. Dried precipitates from 8-day liquid cultures were uniformly spread on aluminium stubs with two-way adherent tabs, air-dried then gold-coated by sputtering for 3 min at RT. For EDX, specimens were uniformly spread directly on adherent tabs and dried before analysis.

Microbiologically enhanced crack remediation
Bacillus pseudofirmus DSM 8715 was grown in 500 ml of alkaline LB broth (30°C, 150 rev min À1 ) for 18-20 h, cells (OD 0Á8-0Á95) were collected by centrifugation (3800 g, 4°C) then washed thrice with 0Á03 mol l À1 Tris-HCl pH 9Á5 buffer. A suspension containing 3 9 10 6 cells was made up in 10 ml of a designed crack repair medium (CRM) of pH 9Á0 containing per litre (calcium acetate 100 g, yeast extract 4 g, glucose 2 g) and was injected once daily for 8 days into microcracks (≤0Á25 mm generated by using a flexural strength machine) of prewetted concrete blocks (40 9 40 9 160 mm) and incubated in a humid chamber (100% RH) at 30°C.

Initial surface absorption test
The initial surface absorption test described in BS 1881-208 (British Standards Institution, 1996 was used as measure of the efficiency of the repair material(s) to exclude water. Although the test does not measure the bulk permeability of the concrete, it provides a measure of the quality of near surface properties by the rate at which water absorbs into the surface of concrete. This zone is where cracks will manifest and heal, thus most often determines durability of concrete and protection of the steel reinforcement therein. Mortar prisms (40 9 40 9 160 mm) were prepared in accordance with BS EN 196-1. The cement was a Portland fly ash, CEM II/B-V 32.5 R conforming to BS EN 197-1, and the fine aggregate was standard CEN sand conforming to BS EN 196-1. The water : cement ratio was 1 : 2 and the aggregate : cement ratio was 3 : 1. In order to avoid cracks penetrating throughout, mortars were reinforced with a fibre polymer mesh placed 15 mm from the base. All mortars were cast in stainless steel moulds, then cured at 95% RH and 20°C for 24 h, followed by immersion in water at 20°C. At 28 days, cracks 0Á12-0Á14 mm maximum width were generated in the centre of the mortar prisms under three-point bending. Controls comprised noncracked mortars.
Following cracking, three mortars were injected with: (i) water only, (ii) CRM only (iii) B. pseudofirmus in CRM with three control prisms per treatment.
A scaled-down and slightly modified version of the method described in BS 1881-5 was used. The specimens were dried for 72 h at 50°C before testing. A cap of 1600 mm 2 was sealed to the surface and filled with water. The rate at which the water was absorbed into the concrete under a pressure head of 200 mm was measured by movement along a capillary tube attached to the cap. The rate of surface absorption was measured at intervals of 10 min, 30 min, 1 h and 2 h from the start of the test.

Sporulation and sporulation kinetics
Application of endospores would be required in order to survive the hostile conditions of a fresh concrete mix of pH c. 13. Therefore, it was important to study timing, speed and the degree of sporulation in the selected Bacillus isolates. Eventually this process could be required on a commercial scale for use in a self-healing agent. Initially three alkalinophilic Bacillus species were chosen and surveyed for spore production in vitro.
Sporulation in B. pseudofirmus (following growth overnight then inoculation into a sporulation medium) was detectable at 20 h and apparently 100% of cells viewed by phase-contrast microscopy contained endospores by 26 h, at which stage divisional growth had ceased (Fig. 1a,b). Bacillus cohnii spores were produced rapidly, beginning by 3 h and 98% sporulation was achieved after 6 h (data not shown). Bacillus halodurans formed long chains that intertwined causing practical difficulty in microscopic quantification and hence percentage sporulation data could not be recorded. Later this isolate was removed from the survey.

Germination inducers and germination kinetics
For application in self-healing concrete, spores need to germinate rapidly to generate cells that precipitate calcite. Germination inducers have been widely studied in Bacillus sp. For example, B. cereus T most effectively germinates with L-alanine and purine ribosides (Shibata et al. 1976). Kudo and Horikoshi (1983) reported that germination of alkaliphilic B. pseudofirmus 2b-2 required the presence of Na + ions along with alanine and inosine.
Germination of B. pseudofirmus spores increased with combined inducer concentrations ranging from 52% at 2 lmol l À1 to >99% at 10 lmol l À1 (Table 1). No germination occurred with NaCl alone in the absence of alanine and inosine.
Kinetics of germination was studied using the optimal conditions above (0Á1 mol l À1 NaCl, 10 lmol l À1 L-alanine and inosine), when germination of B. pseudofirmus spores exceeded 99% by 120 min with a corresponding OD reduction of 43Á8% (Fig. 2). Evaluation of germination by B. cohnii was very difficult because of its small spore size. In any event, germination under these conditions was low and the isolate was not used further.

Survival of spores in cement stone
It was essential to test the survivability of endospores in cement paste during mixing and subsequent hardening, where they would initially experience extreme pH, then compression. The outcome would assist in designing an appropriate way of delivery and predicted longevity in concrete. Estimation of viable spores (CFU per gram) in the ageing cement paste specimens was conducted by serial dilution of finely crushed specimens. The noninoculated cement control samples did not show any viable growth. In specimens containing B. pseudofirmus spores, after 1 and 3 days of curing, c. 4Á4 and 1Á4%, (i.e. 5Á3 and 1Á7 9 10 4 spores per gram of the originally incorporated 1Á2 9 10 6 spores per gram) were recovered as viable colonies (Fig. 3). This number declined to 0Á58 then 1% between 7 and 28 days when spore survival resulted in 7Á1 then 1Á4 9 10 3 CFU per gram. Beyond 42 days recovery was relatively consistent at c. 1-4% recovery.
In specimens containing B. cohnii spores, after 1 and 3 days of curing, c. 2Á1 and 0Á25% of the originally incorporated 1Á4 9 10 6 spores per gram were recovered as   ) in the presence of alanine (10 lmol l À1 ), inosine (10 lmol l À1 ) and NaCl (0Á1 mol l À1 ) in 30 mmol l À1 Tris-HCL (pH 9Á5). Error bars show SD. viable colonies; this number declined to 0Á15-0Á09% between 7 and 28 days. Beyond 42 days and up to 93 days recovery again was relatively consistent at c. 0Á52-2Á4% (data not shown).
Bacillus halodurans, after 1 and 3 days of curing, was recovered as c. 0Á24% viable colonies of the originally incorporated 4Á4 9 10 6 spores per gram; this number declined below 0Á1% from 7 days onwards (data not shown). Based on the earlier difficulty of assessing spore production microscopically and limited survival in concrete, this isolate was not used further.

Optimization of medium for calcite precipitation for calcite quantification
Bacterial isolates for potential use in self-healing concrete must precipitate calcite rapidly and in sufficient quantity in response to crack formation. In order to quantify calcite formation in vitro it was essential to design a suitable medium. B4 medium has been used for screening bacteria for calcium carbonate precipitation (Shirakawa et al. 2011). Although B4 medium supported the greatest growth (Fig. 4a), precipitation in controls precluded its use, questioning its value in previous studies. However, 1/10th LBA was selected as it supported good bacterial growth with no precipitation in uninoculated controls.
In vitro precipitation of calcite in trace amounts was detectable within 48 h of inoculation. The levels of CaCO 3 precipitated (Fig. 4b) increased up to 6-8 days when yield was mean 2145-2330 mg l À1 . Although every effort was made to remove bacteria from crystals, inevitably some remained. One hundred and sixty-five, 171 and 181 cells were detected by CFU in crystals (weights as above) from 6, 7 and 8 days respectively. However, no elemental Ca 2+ was detected from equivalent numbers of bacteria, grown in the same medium.

Characterization of precipitate as CaCO 3 by FTIR, Raman spectroscopy, SEM and EDX
FTIR spectra have been extensively used to distinguish crystalline calcium carbonate polymorphs (calcite, aragonite and vaterite) (Ghosh 2001). FTIR spectra (Fig. 5a) of crystals from B. pseudofirmus, B. cohnii and B. halodurans when analysed in the range of 400-4000 cm À1 , showed three major absorption peaks at 1530, 1426, 875 and 712 cm À1 peculiar to calcite. In the Raman spectra of Fig. 5b, the most intense Raman bands are observed at low frequencies (50-400 cm À1 ), which corresponds to the lattice mode vibrations. Calcite has minor bands at $ 282 and 713 cm À1 ; aragonite has minor bands at $ 207 and 704 cm À1 (Bischoff et al. 1985). In the inset figure of Fig. 5b, a peak at 281 cm À1 is observed at frequencies (50-400 cm À1 ) and peaks at 702 and 1085 cm À1 occur. The FTIR and Raman spectrum reveals the precipitated calcium carbonate as a mixture of calcite and aragonite. The spectra of crystals precipitated by B. cohnii and B. halodurans showed similar peaks (data not shown).
SEM analysis (Fig. 5c) revealed the crystals of B. pseudofirmus from 1/10th LBA medium were uniformly sized, and appeared as multifaceted structures. Energydispersive X-ray spectroscopy (Fig. 5d) showed that different regions in contained Ca, C, O and P. (Fig. 5d).

Bacterially induced microcrack sealing
For proof of concept, a medium (CRM), which provided calcium carbonate precipitation in situ (T. K. Sharma, unpublished data), was used. The in vitro yields of calcium carbonate by B. pseudofirmus cells in this medium were 1403, 3283 and 3870 mg l À1 after 3, 6 and 8 days respectively. Levels increased slowly to 4023 mg l À1 by 15 days.
After about 8 days, a white precipitate resulted in occlusion of microcracks. The precipitated bio-mineral was hard and adhered strongly to the concrete (resisted physical extraction using a scalpel blade). Precipitates were most prominent at the crack surface. SEM images (Fig. 6a-c) clearly shows the healing of cracks by B. pseudofirmus. The EDX spectrum (Fig. 6d) of the precipitated material reveals Ca, C and O. Portland cement contains traces of Si, Al and S hence these respective peaks are present.

Initial surface absorption test
As is typical of mortars the rate of surface absorption reduced with time as the moisture content of the mortar progressively increases during the test (Fig. 7a). The mean initial surface absorption of the noncracked mortar at 10 min was 0Á10 ml m À2 s À1 and at 2 h was 0Á02 ml m À2 s À1 . Typically mortars are considered to have excellent resistance to ingress of water if they have values ≤0Á25 and 0Á07 ml m À2 s À1 at 10 min and 2 h (Concrete Society 1985). These properties were evident in the noncracked mortars.
The two cracked, control mortars had very high initial water surface absorption values. The water-treated control had initial surface absorptions of 4Á1 and 2Á4 ml m À2 s À1 at 10 min and 2 h; while the medium control gave even higher values of 8Á2 and 6Á3 ml m À2 s À1 respectively. These values reflect minimal resistance to water ingress. Absorption values ≥0Á5 and 0Á15 ml m À2 s À1 are normally regarded as high. In both cases it is likely that water was flowing through the crack without obstruction, rather than ingressing via absorption.
Cracks injected with B. pseudofirmus gave, in contrast, initial surface absorption values of a similar magnitude to that of the noncracked samples: 0Á10 ml m À2 s À1 at 10 min and 0Á03 ml m À2 s À1 at 2 h. These results clearly show that bacterial activity within the created cracks has resulted in a compound(s) that prevents flow of water through the specimen (Fig. 7b). The conferred resistance to water absorption is similar to that of the surrounding intact mortar.

Discussion
It was crucial to determine the timing and quantity of sporulation for experimental purposes and eventually for efficient commercial production. Sporulation in various Bacillus species is normally induced by reduced levels of nutrients (Errington 2003), but additionally elemental Mn 2+ is required during the active growth phase. Apparently 100% sporulation occurred in B. pseudofirmus and B. cohnii within 24 h under optimum sporulation conditions. Therefore, commercial production of spores could be rapid, reducing production times and cost.
Various amino acids, ribosides, sugars and related compounds can act as germinants, with generally at least one inducer or a combination of inducers necessary for germination. Bacillus pseudofirmus germinated rapidly in the presence of L-alanine and inosine with 0Á1 mol l À1 NaCl, but there was no germination in the absence of NaCl. Germination of B. cereus was reported at 10 mmol l À1 inosine and 10 or 100 mmol l À1 alanine (Shibata et al. 1976;Gounina-Allouane et al. 2008), whereas the optimal concentration of inducers L-alanine and inosine was 0Á4 mmol l À1 for B. pseudofirmus No. 2b-2 (Kudo and Horikoshi 1983). However, for application in self-healing concrete, cost effectiveness is crucial and hence inosine and L-alanine were tested here at lower concentrations than reported elsewhere. Germinants at 10 lm were found to successfully trigger germination.
In the current study, a correlation coefficient of 0Á88 between percentage germination and reduction in optical density confirmed OD as a more rapid and facile technique than microscopy to determine germination, albeit at lower accuracy. For example, c. 50% reduction in OD corresponded to 100% germination in B. pseudofirmus DSM 8715, which was similar to that found for Bacillus firmus No.2b-2 (Kudo and Horikoshi 1983).
Spore survival in cement appears to be low, when compared to the theoretical potential CFU. Detection of viable B. pseudofirmus cells long term (42-93 days) was only between c. 1-4%. Similarly, viability after 9 days of B. cohnii spores was about 0Á5-2Á5%. The well documented resilience of Bacillus endospores to adverse conditions and the relatively consistent survival over a long time period (93 days) in cement might suggest recovery of spores from cement matrix is the limiting factor, rather than viability loss due to factors such as initial alkaline pH and consequent compression on curing. In practical terms, provided that the initial inoculum is high enough, then sufficient numbers of spores could potentially survive and take part in the healing of concrete. In contrast, Jonkers et al. (2010) found that viability of spores in cement specimens decreased to negligible levels within 4 months and commented on the likely role of compression when pore sizes in concrete decreased to c. 1 lm.
This is the first critical analysis of quantification and kinetics of bacterial calcite formation. Based on AAS analysis, B. pseudofirmus precipitated >2Á3 g l À1 of CaCO 3 after 8 days of incubation but production was detectable by 2 days. Amounts produced on a different medium as used in crack healing exceeded 4 g l À1 . Clearly extrapolation of timing and amounts from in vitro to in situ are currently a matter for speculation. The speed of germination, extent of growth and the duration of conditions suitable for bacterial growth in microcracks within concrete are unknown, but will be addressed in future studies. According to weight gain, Rodriguez-Navarro et al. (2003) found most carbonate precipitation occurred in the first 5-10 days and Jonkers reported maximal crystal formation after c. 7 days. Analogously, here we chose 8 days based on our observations of crack occlusion following bacterial treatments.
We obtained a correlation value of 0Á99 between dry weight of precipitated, microfiltered, extracellular material from B. pseudofirmus cultures and calculated values of CaCO 3 precipitation based on AAS, suggesting that for this isolate at least, AAS could be by-passed for quantification. However, for some species other bacterial products might contribute to dry weight analyses, with extracellular polymeric substances (EPS) an obvious candidate (Aslam et al. 2008). Using dry weight determination, B. pseudofirmus produced in vitro significantly more precipitated carbonate than either of the other two species (data not shown).
Combined with data for sporulation, germination extent and spore survival in concrete, calcite precipitation levels led to the selection of B. pseudofirmus as the isolate with the greatest potential for use in self-healing concrete.
Crystals were inevitably contaminated with bacteria, so these were analysed for Ca 2+ content. However, there was no elemental Ca 2+ detected when bacterial cells solubilized in HCl were subjected to AAS. Notably, related B. subtilis does not accumulate calcium (Silver and Kralovic 1969). Living cells were found within carbonate crystals by Krumbein (1974), and Morita (1980) -Navarro et al. 2003). All three polymorphs are formed in calcareous deposits of living organisms, with the organic matrix affecting the form in molluscs, but ionic composition greatly influencing the polymorph in solutions (Simkiss (1964). The SEM images of calcium carbonate crystals by B. pseudofirmus revealed layers of bioliths (Ferrer et al. 1988;Baskar et al. 2009); the layered SEM structure of the bio-mineral could result from the mixture of calcite and aragonite. Different culture media and bacterial isolates can yield variable crystal morphologies (De Muynk et al. 2008a).
Application of B. pseudofirmus to microcracks resulted in occlusion, by virtue of deposition of strong, adherent material suggested by EDX to be CaCO 3 . Cracks were effectively sealed based on water absorption tests. The calcium carbonate precipitated remains intact, due to its low solubility at high pH, therefore should provide a permanent as well as effective seal. Wang et al. (2014) found that water permeability was reduced about 10-fold by addition of encapsulated B. sphaericus. A decrease of 65-90% in water absorption by mortar and greater resistance to freeze-thaw resulted from deposition of calcium carbonate crystals by surface applied bacteria (De Muynk et al. 2008b).
For initial proof of concept, we conducted experiments with unprotected, vegetative cells. Clearly, for eventual practical use, bacteria will have to be encapsulated. Also, separate encapsulation of nutrients will be necessary, to include germination inducers and organic source of calcium, as detailed above. Some reports have emphasized the need for protection of bacteria before incorporation into cement mixtures (Bang et al. 2010;Jonkers et al. 2010;Wang et al. 2014). Encapsulation will not only prevent premature spore germination but also reduce the bacterial inoculum volumes required and avoid potential damage by nutrients to the properties of concrete (unpublished data).
The germination of endospores in cement stone a realistic situation, which would be dependent on water ingress following microcrack formation, rupture of nutrient-containing capsules and suitable growth temperatures remains to be assessed. Incorporated oxygen probes might reveal the initial occurrence and even extent of growth. Suitable imaging of crack surfaces should show resulting vegetative cells and associated calcite crystals. Release of nutrients from incorporated capsules could be monitored by basic chemical analysis of one major component (e.g. glucose) from washings made from cracked surfaces. Onset of adverse conditions, such as lowered hydration level, could then trigger a round of sporulation providing potential cells for the next event.
The size of cracks must be a significant factor in terms of the ability of bacterial precipitates to result in effective occlusion. Crystals produced in vitro from shaken, liquid cultures, after processing (centrifugation and filtration) were c. 2Á5 lm (light microscopic analysis), yet crack widths 100 times greater were effectively sealed by applying bacterial cells. Crystal size and morphology can be influenced by many factors including nucleation density and localized supersaturation. Presumably, various factors contribute to the presentation of crystals, to include incorporation within a bacterial biofilm (Aslam et al. 2008). Negatively charged bacterial EPSs are effective at binding cations and act as nucleation sites for mineral deposition. Also, bacterial cell wall functional groups such as hydroxyl and phosphate become increasingly negative at high pH with affinity for cations and isolated bacterial membranes will precipitate CaCO 3 (see Rodriguez-Navarro et al. 2003). These authors added that in situ, bacterial attachment to limestone then formation of carbonate around bacteria ensures strong adhesion of the newly formed cement. Nucleation points within the static environment could create larger crystal accumulations (Rivadeneyra et al. 1996). Bio-mineralized calcite is likely to be more resistant to dissolution than calcite formed by inorganic precipitation as the former is less soluble (Morse 1983). Crack healing of up to 0Á46 mm wide cracks was recorded by Wiktor and Jonkers (2011), but the specimens were submerged in water for 100 days, in contrast to an 8-day period used here.
Based on our studies of sporulation, germination, in vitro calcium carbonate precipitation in alkaline conditions and ability of spores to survive within concrete, B. pseudofirmus appears to be an appropriate strain to use as a self-healing agent for bio-mineralization in cementbased materials. Ongoing work includes testing Bacillus sp. with a wider range of optimal growth temperatures, for applications from temperate to tropical regions. Also, we are investigating spore-forming Bacillus sp. with neutral pH growth optima to function within aged concrete as the pH reduces towards neutrality as an outcome of carbonation (Pade and Guimaraes 2007). In order to provide convincing evidence as to applicability and durability, future studies need to be in depth and applied to treated cement stone in situ.