Engineered Ureolytic Microorganisms Can Tailor the Morphology and Nanomechanical Properties of Microbial-Precipitated Calcium Carbonate

We demonstrate for the first time that the morphology and nanomechanical properties of calcium carbonate (CaCO3) can be tailored by modulating the precipitation kinetics of ureolytic microorganisms through genetic engineering. Many engineering applications employ microorganisms to produce CaCO3. However, control over bacterial calcite morphology and material properties has not been demonstrated. We hypothesized that microorganisms genetically engineered for low urease activity would achieve larger calcite crystals with higher moduli. We compared precipitation kinetics, morphology, and nanomechanical properties for biogenic CaCO3 produced by two Escherichia coli (E. coli) strains that were engineered to display either high or low urease activity and the native producer Sporosarcina pasteurii. While all three microorganisms produced calcite, lower urease activity was associated with both slower initial calcium depletion rate and increased average calcite crystal size. Both calcite crystal size and nanoindentation moduli were also significantly higher for the low-urease activity E. coli compared with the high-urease activity E. coli. The relative resistance to inelastic deformation, measured via the ratio of nanoindentation hardness to modulus, was similar across microorganisms. These findings may enable design of novel advanced engineering materials where modulus is tailored to the application while resistance to irreversible deformation is not compromised.


Microorganism viability, urease activity, and solution chemistry during the 7-day experiment.
We monitored the viability and urease activity for two engineered ureolytic E. coli strains, E. coli HB101/pBU11 and E. coli HB101/ure-integration, and the native producer S. pasteurii. Over 7 days, colony forming units (CFUs) (Fig. 1A) and urease activity (Fig. 1B) decreased for all three ureolytic microorganisms. However, urease activity normalized to CFU (Fig. 1C) decreased with time for E. coli HB101/pBU11 but reached a maximum at 2d for the other two ureolytic microorganisms. Only S. pasteurii had measurable urease activity at 7d.
We then quantified the precipitation kinetics of each microorganism through measuring pH, calcium (Ca), and phosphorus (P) at each timepoint. Each of the three ureolytic microorganisms raised the solution pH compared to Media (no cell) and Wild-Type (E. coli HB101 with no urease genes) controls over the course of the 7d experiment (Fig. 1D). Soluble Ca measured from inductively coupled optical emission spectroscopy (ICP-OES) initially decreased in all cultures, including the two controls. However, as expected, Ca was most reduced for solutions containing ureolytic microorganisms (Fig. 1E), with most of the Ca in solution depleted by day 2. The rate of Ca depletion (dCa/dt) varied by microorganism (Fig. 1F). Through 1d, dCa/dt was not different between E. coli HB101/ure-integration and controls. At 2d, dCa/dt was at a maximum for the low-urease activity engineered microorganism. For the higher-urease activity engineered microorganism, E. coli HB101/pBU11, as well as the native producer S. pasteurii, dCa/dt was greatest at 1d and then approached zero at 2-7d. At the timepoints most closely representing calcite nucleation as verified by XRD and SEM (12h for S. pasteurii and E. coli HB101/pBU11; 1d for E. coli HB101/ure-integration), dCa/dt increased in the order of measured urease activity. Specifically, dCa/ dt was lowest for E. coli HB101/ure-integration, followed by S. pasteurii, and then E. coli HB101/pBU11. Soluble P concentration varied considerably throughout the week-long experiment (Fig. 1G). The rate of P depletion (dP/dt) was initially large for all three ureolytic microorganisms and later reversed as P returned to solution, indicating dissolution of P-containing precipitates (Fig. 1H). As with dCa/dt, the timing of this transition depended on microorganism. For the low-urease activity microorganism, P returned to solution at days 2-7, whereas P dissolution occurred a day earlier for the two more ureolytic microorganisms. Finally, consistent with SEM and XRD observation of calcium phosphate crystals, the wild-type and media controls demonstrated high (C) Urease activity normalized to CFUs was at a maximum at 1d for E. coli HB101/pBU11 and 2d for the other two ureolytic microorganisms (D) Native and engineered ureolytic microorganisms raised the solution pH, although at different rates. (E) Soluble Ca was depleted at different times for ureolytic strains; the slowest total Ca depletion was observed for the engineered strain with the lowest urease activity, E. coli HB101/ure-integration. Ca was slightly lowered in controls with the precipitation of CaP. (F) The rate of calcium depletion, dCa/dt, differed by microorganism. The dCa/dt at beginning of calcite nucleation (12h for S. pasteurii and E. coli HB101/pBU11, 1d for E. coli HB101/ure-integration) was related to the urease activity of the microorganism and inversely related to calcite crystal size at the study endpoint. (G) Soluble P initially decreased for ureolytic strains and control conditions with the spontaneous precipitation of CaP. (H) The rate of P depletion in solution, dP/dt, varied with microorganism such that phosphorus was initially depleted and then returned to solution with calcite nucleation.
www.nature.com/scientificreports www.nature.com/scientificreports/ initial dP/dt that then tapered towards zero. Unlike with ureolytic microorganisms, the control conditions did not return P to solution during the 7d experiment.

Mineral phases assessed by XRD.
We employed x-ray diffraction (XRD) to characterize which phases of CaCO 3 and calcium phosphate (CaP) were precipitated by each microorganism across the 7d experiment. XRD revealed that Media and Wild-Type controls produced CaP, namely brushite and apatite (Fig. 2D,E). The urea-CaCl 2 growth media contains P in the ingredient Difco ('nutrient broth'). No CaCO 3 was identified for the controls at any time point. For S. pasteurii, only CaCO 3 was detected across the course of the experiment, although CaCO 3 polymorphs varied with time. For the E. coli HB101/pBU11 strain, only calcite was detected at each time point ( Fig. 2A). E. coli HB101/ure-integration produced calcite and CaP at 12h and 1d, and only calcite for 2d-7d ( Fig. 2B). At 12h, S. pasteurii produced vaterite and calcite, while only calcite was detected for 1d-7d (Fig. 2C).
The XRD results demonstrated good correspondence with the changes in solution chemistry noted over the 7d experiment. Control conditions did not raise pH and only produced CaP. Changes to pH and soluble Ca were evident as early as 12h for S. pasteurii and E. coli HB101/pBU11, which aligned with the detection of prominent CaCO 3 peaks at 12h onwards. By contrast, for the lower urease-activity E. coli HB101/ure-integration, neither pH nor soluble Ca were appreciably changed until 2d. Calcite and CaP were detected during this first 2d. From 2-7d, when pH was elevated by E. coli HB101/ure-integration to similar levels as achieved the other two microorganisms, XRD only detected calcite. crystal morphology. To evaluate the hypothesis that precipitation kinetics influence crystal morphology, we assessed crystal size and shape using scanning electron microscopy (SEM) at all time points. S. pasteurii produced large (~100 µm) structures at 12h with regions of well crystallized material with bacterial casts alongside neighboring regions of disorganized and uncasted material (Fig. 3A-D). By 1d, and continuing through 7d, crystals precipitated by S. pasteurii recrystallized to smaller and more uniform crystalline structures with bacterial casts. The crystals precipitated by E. coli HB101/pBU11 demonstrated a similar transition from large, mixed structures at 12h to smaller calcites with clear bacterial casts (Fig. 3E-H). E. coli HB101/ure-integration produced nanocrystalline sheets at 12h and 1d ( Fig. 3I-L). By 2d, and continuing through 7d, precipitates from E. coli HB101/ure-integration were well-crystallized, and casted calcites and were often very large. From measurement of crystal sizes at 7d, the average size of calcite crystals produced by E. coli HB101/ure-integration was larger than for crystals precipitated by S. pasteurii (+371%, p < 0.001) and much larger than for crystals precipitated by E. coli HB101/pBU11 (+1924%, p < 0.001) ( Table 1). SEM images revealed that Media and Wild-Type controls www.nature.com/scientificreports www.nature.com/scientificreports/ produced nanocrystalline sheets that did not vary in morphology from 1-7d ( Fig. 3M-R). The morphologies observed in the control conditions were similar to those from the earliest timepoints (12h-1d) for E. coli HB101/ ure-integration. chemical composition of minerals via SeM-eDS and WDS. Because XRD revealed the presence of CaP phases for controls as well as for ureolytic microorganisms at early time points, we employed SEM-EDS to identify CaP in study precipitates. SEM-EDS confirmed that nanocrystalline sheets produced by Media and Wild-Type controls contained both Ca and P. For biogenic precipitates from S. pasteurii and E. coli HB101/pBU11 collected at 12h, the faceted and casted regions of these precipitates were identified as CaCO 3 . Meanwhile, disorganized regions of the same structures were identified as CaP (Fig. 4A,B). At later timepoints, only CaCO 3 was identified in crystals precipitated by S. pasteurii and E. coli HB101/pBU11 (Fig. 4C,D). By contrast, E. coli HB101/ure-integration produced mostly CaP until day 2, at which point all crystals studied with SEM-EDS were composed of only CaCO 3 . We also sought to identify where P was present in the interior of biogenic CaCO 3 at the study endpoint. Cross-sections of crystals precipitated by each of the three ureolytic microorganisms were mapped using wavelength dispersive spectroscopy (WDS). Crystals from all three ureolytic microorganisms contained substantial P content (P:Ca ~ 5-10%) in the core of the crystal, as well as within a ring at the crystal surface (Fig. 4E). and E. coli HB101/pBU11 (E-H) both had an evolution over time from large, poorly faceted structures at 12h to smaller, faceted polycrystalline structures with abundant bacterial casting at 1-7d. E. coli HB101/ure-integration (I-L) initially formed calcium phosphate nanocrystals until 2d, when large, round, highly casted CaCO 3 crystal structures appeared. At 7d, the average calcite crystal size was smallest for E. coli HB101/pBU11, followed by S. pasteurii and then E. coli HB101/ure-integration. Wild-Type (M-O) and Media (P-R) controls produced only calcium phosphate nanocrystalline sheets which did not vary in morphology from 1-7d. All SEM images shown at 500×. www.nature.com/scientificreports www.nature.com/scientificreports/ Mechanical properties via nanoindentation. We utilized nanoindentation to evaluate whether microscale mechanical properties of biogenic CaCO 3 depend on the precipitating microorganism. From ANOVA, there was a main effect of microorganism (p = 0.031) on mean nanoindentation modulus, E i . Tukey post-hoc testing revealed that the E. coli HB101/ure-integration strain produced calcite crystals with significantly higher moduli (E i , +14.8%, p = 0.025) than E. coli HB101/pBU11 (Fig. 5A). Calcites from S. pasteurii exhibited moduli that were not statistically different from either engineered E. coli strain.
Histograms of raw (i.e., not averaged) nanoindentation moduli for each microorganism showed that all ureolytic microorganisms produced calcites with a range of E i , yet E. coli HB101/ure-integration produced stiffer crystals with greater frequency (Fig. 5B). While 34.8% of indents from crystals from E. coli HB101/ure-integration had E i of at least 40 GPa, this percentage declined to 22.1% and 14.7% for crystals precipitated by S. pasteurii and for E. coli HB101/pBU11, respectively. Average crystal size and average nanoindentation modulus, measured for crystals collected from the same cultures, were found to significantly correlate (p = + 0.830, p < 0.05).
Neither indentation hardness, H i , nor hardness, H (as determined from Eq. 5), significantly differed with ureolytic microorganism ( Table 1). The ratio of H i /E i , interpreted as the relative material resistance to inelastic deformation during nanoindentation, were similar for all three ureolytic microorganisms (Fig. 5A inset). Thus, although E i depended on which microorganism precipitated the crystal, the relative resistance to inelastic deformation was similar for all crystals across all ureolytic microorganisms.  www.nature.com/scientificreports www.nature.com/scientificreports/ Assessment of crystal phase and composition at indentation sites. We aimed to understand why the low-urease activity engineered strain (E. coli HB101/ure-integration) produced significantly higher modulus CaCO 3 crystals than the low-urease activity engineered strain (E. coli HB101/pBU11). A Raman spectrum collected at the same location as each indent was used (1) to identify the phase(s) of CaCO 3 present, (2) to measure the relative size and perfection of the crystal, and (3) to identify the presence of one or more CaP phases. Of note, though CaP was not identified in bulk 7d samples using XRD, Raman spectroscopy can detect trace phases below the XRD detection limit at microscale observation sites.
All Raman spectra contained calcite peaks, while no additional polymorphs of CaCO 3 (i.e., vaterite or aragonite) were identified, indicating that differences in nanoindentation properties were not due to CaCO 3 phase. Weak CaP peaks were detected for some indent locations for each of the three ureolytic microorganisms, revealing the presence of minor quantities of CaP. No differences in mean crystal size (A280/A712) or perfection (FWHM of v 1 carbonate) or CaP phases (brushite:v 1 carbonate, apatite:v 1 carbonate, carbonated apatite:v 1 carbonate) of precipitates were found between microorganisms (Table 2). However, relationships between CaP content and E i differed by microorganism. For crystals precipitated by E. coli HB101/pBU11, for example, reduced E i was significantly correlated with greater apatite:v 1 carbonate (r 2 = 0.25, p = 0.001) and brushite:v 1 carbonate (r 2 = 10.2%, p = 0.042). These measures demonstrated a high degree of intercorrelation (r 2 = 0.514) and thus cannot be interpreted separately. Instead, these findings show that E i is reduced for higher amounts of CaP phases  www.nature.com/scientificreports www.nature.com/scientificreports/ at indentation sites on precipitates produced by E. coli HB101/pBU11. By contrast, for crystals precipitated by E. coli HB101/ure-integration and S. pasteurii, the presence of CaP was not significantly correlated with reduced E i .

Discussion
The purpose of this study was to determine whether morphology and material properties of biogenic CaCO 3 could be controlled by engineering ureolytic microorganisms with different urease activities and, consequently, precipitation kinetics. We studied biogenic crystal precipitates from three ureolytic microorganisms: the native ureolytic S. pasteurii, the engineered E. coli HB101/pBU11 with similar urease activity as S. pasteurii, and the engineered E. coli HB101/ure-integration strain with lower urease activity than either of the other two microorganisms. To understand how nucleation, growth, and stability of crystal precipitates differed for these three microorganisms, the solution chemistry, crystal phase, and crystal morphology were compared at multiple time points from 12h through 7d and compared with control conditions. At 7d, microscale chemistry and elemental analysis of biogenic CaCO 3 from the three microorganisms were performed with Raman spectroscopy and WDS, respectively, while microscale mechanical properties were assessed using nanoindentation.
All three microorganisms produced only calcite and no other polymorphs of CaCO 3 by the end of the 7d experiment. Vaterite was observed as a transient phase at 12h for crystals precipitated by S. pasteurii, while no metastable phases of CaCO 3 were observed at any timepoint for either engineered E. coli strain. Importantly, metastable phases may have been present for engineered E. coli strains before 12h, or in sub-detectable quantities. Alternatively, vaterite precipitated by S. pasteurii may have been stabilized by extracellular polymeric substances (EPS), as has been witnessed in prior work 1,14 . Unlike S. pasteurii, E. coli HB101 does not produce biofilm 26,27 . Ultimately, the transient vaterite may not have great consequence to the mature calcite crystals, as all detectable vaterite transformed into calcite during the experimental timeframe.
The kinetics of mineralization and resulting crystal morphology depended on microorganism ureolytic activity. The initial rate of calcium depletion from solution (i.e., the rate of calcium depletion most closely associated with calcite nucleation for each microorganism) increased with the urease activity of the precipitating microorganism. The low-urease activity microorganism (E. coli HB101/ure-integration) had the lowest initial calcium depletion rate and the high-urease activity microorganism (E. coli HB101/pBU11) had the highest initial calcium depletion rate. The average size of crystals measured at 7d was also inversely related to both the urease activity and the initial calcium depletion rate. At the study endpoint, the average crystal size for the low-urease engineered microorganism was almost 20 times larger than for the high-urease engineered microorganism. The growth of cells is another important consideration for crystal nucleation and growth, as cells produce urease and can also serve as nucleation centers. S. pasteurii maintained higher viability than the two engineered E. coli strains throughout the study. This may have influenced why S. pasteurii most quickly depleted total Ca in solution. However, in our study, neither initial calcium depletion rate nor crystal size were clearly related to differences in colony formation units at any point of the experiment.
Our findings that lower urease activity microorganisms produced slower initial nucleation kinetics and larger crystals is consistent with work by Cuthbert and coworkers, where S. pasteurii biofilms were cultured with either high (13 g/L) or low (0.13 g/L) nutrient broth concentration 28 . The biofilms produced in low nutrient conditions were sparser and consequently had slower initial ureolysis and calcite mineralization. Calcite crystals were larger when produced in the low-nutrient conditions, which was explained by lower initial saturation state. It is expected from classical nucleation theory that the driving force for nucleation is increased with saturation state, and therefore that the stable crystal nucleus size decreases with increased saturation state 29 . While we did not measure saturation state, our measurements of urease activity and solution calcium are consistent with the interpretation that differences in ureolytic activity, modulated in our study via genetic engineering, alter the solution saturation state and therefore mineralization kinetics and crystal size. The size difference of biogenic calcite precipitates may be of specific importance for some engineering applications. Mitchell and coworkers, for example, found that larger calcite crystals exhibit decreased solubility and may therefore be more useful for co-precipitating metals or radionuclides 13 .
We were able to observe that calcium phosphate phases have a role in both nucleation and growth of CaCO 3 precipitated by S. pasteurii and both engineered E. coli microorganisms. Brushite and apatite nanocrystalline sheets were observed by XRD and SEM throughout the 7d experiment in control conditions, as well as for the low-urease engineered E. coli strain before the onset of substantial calcite precipitation. Brushite and apatite were also observed concomitant with calcite nucleation in S. pasteurii and the high-urease activity engineered E. coli strain. Our results from XRD and SEM are in agreement with changes to solution P throughout the 7d experiment. After an immediate drop in P, some portion of P returned to solution with the onset of calcite nucleation and was available to participate in additional precipitation. The saturation state with respect to calcium phosphate would be expected to change continuously during calcite precipitation. In addition, some CaP phases, such as brushite, have decreased solubility in alkaline conditions 30 . We are not the first to observe that CaP plays a role in the nucleation of biogenic CaCO 3 crystals. Dupraz et al. visualized nanoscale CaP nucleated on S. pasteurii before CaCO 3 deposition 31 . In addition, ureolytic microorganisms harvested from cave environments precipitated CaCO 3 and CaP in lab cultures 1 . However, novel to this study, we directly observed nucleation of calcite on calcium phosphate, and also demonstrate that P is evident in a ring near the polycrystal structure exterior, as well as within the nucleation center (Fig. 4).
The nanoindentation modulus (E i ) of biogenic calcite was higher for microorganisms with lower urease activity and slower precipitation kinetics. Specifically, E i was approximately 15% higher for the engineered microorganism with lower urease activity, E. coli HB101/ure-integration, compared with E. coli HB101/pBU11 with higher urease activity. While both microorganisms produced biogenic CaCO 3 with a range of moduli, E. coli HB101/ure-integration produced 63% more crystals with moduli of at least 40 GPa than the engineered E. coli strain with higher urease activity. S. pasteurii, which has urease activity between that of the two engineered www.nature.com/scientificreports www.nature.com/scientificreports/ strains, produced calcite with moduli that were between and not significantly different from either engineered E. coli strain. The nanoindentation moduli observed in this study were lower than for inorganic CaCO 3 (~60-80 GPa) 20 . However, nanoindentation moduli of biogenic CaCO 3 demonstrate a wide range of values; crystals precipitated in lab cultures using non-engineered microorganisms had nanoindentation moduli from 36-65 GPa 21 , while crystals precipitated in the presence of carbonic anhydrase isolated from the sponge Sycon raphanus had nanoindentation moduli of 43-47 GPa 32 . On the other hand, cave speleotherms produced via bacterial precipitation have much stiffer crystals, with nanoindentation moduli from 64-126 GPa 33 .
We sought to understand if differences in E i for biogenic CaCO 3 produced by engineered E. coli with low-high urease activity could be attributed to variation in CaCO 3 polymorphs, crystal porosity, crystal perfection, or inclusion of CaP phases. Differences in nanoindentation modulus and hardness for biogenic CaCO 3 have been previously attributed to the presence of metastable vaterite or aragonite, in both lab-grown cultures as well as cave speleotherms 19,21,33 . While variation in CaCO 3 nanomechanical properties may be desirable for engineered applications such as functionally-graded composites and films, achieving this variation via the inclusion of metastable CaCO 3 polymorphs such as vaterite is not ideal. In our study, only calcite (i.e., no vaterite or aragonite) was detected using XRD for any of the 7d samples tested by nanoindentation. Site-matched Raman spectroscopy revealed that every crystal measured with nanoindentation was only calcite and did not include another CaCO 3 polymorph. Porosity could be altered through differences in crystal nucleation and growth modulated by changing microorganism urease activity, but all indents with significant compaction events (indicating pore collapse) were excluded from analysis. Differences in porosity at the nanoscale cannot be excluded but were not specifically interrogated by this study. Crystal perfection, as measured through crystallinity and also the A280/A712 ratio, also did not significantly correlate to E i for any microorganism. However, from Raman spectra obtained at the same sites as indentation, we detected a significant, negative linear relationship between the presence of CaP phases (brushite and apatite) and E i . Both E. coli HB101/ure-integration and S. pasteurii had similar mean CaP, yet quantity of CaP did not predict E i . The specific mechanism by which calcium phosphate influences the nanomechanics of calcite crystals precipitated by the high-urease microorganism is not clear. Calcium phosphates measured by nanoindentation have a wide range of mechanical properties, depending on phase, crystallization, and organic content. Amorphous hydroxyapatite printed on a titanium substrate had a nanoindentation modulus of 65 GPa, which increased to 127 GPa after annealing 34 . Yet lab-grown hydroxyapatite-brushite coatings can have moduli as low as 3-15 GPa 35 . Brachiopods shells containing calcium phosphate demonstrate nanoindentation moduli from 5-50 GPa, with lower values associated with more organic content 20 . In addition, calcium phosphate could, in theory, alter calcite crystal mechanics through assembling at grain boundaries, increasing porosity of the polycrystalline assembly, altering calcite orientation, or potentially myriad other mechanisms. Future investigations varying the phosphate concentration in solution would facilitate investigation of the likely complex role of phosphate on cell growth as well as crystal nucleation, growth, and resulting mechanical properties.
While nanoindentation modulus varied with the kinetics of the precipitating microorganism, hardness (H) did not. Neither H, which was determined from E i and nanoindentation hardness (H i ) using assumptions of elastic-plastic deformation, nor H i /E i significantly differed with microorganism. The ratio of H i /E i , sometimes referred to as "resilience", is related to the elastic strain to yield 36 and is generally higher for biogenic materials when toughness is advantageous for the organism 20,36 . For all biogenic calcites in this study, H i /E i values were generally quite high (~0.07), and were equal or greater to those collected from other biogenic materials (e.g., brachiopods and otoliths) 19,20 and much greater than for inorganic CaCO 3 (~0.03-0.04), which has relatively lower hardness and higher modulus 20 . Of key importance, we observed that the modulus of biogenic calcite can be selected for higher or lower values through selection of microorganism with appropriate kinetics, while the material hardness is not compromised.
There were several important limitations to this study. First, while we did note that the presence of CaP reduces nanoindentation modulus for crystals precipitated by E. coli HB101/pBU11, we did not determine the specific mechanism by which it does so. Second, while engineered E. coli improves control of the microorganism and its crystal precipitates compared with a native ureolytic microorganism, these engineered strains are not readily field-deployable, while S. pasteurii has already been used successfully in field-scale applications 7 . It is possible that increases to calcite moduli may also be accomplished with S. pasteurii via environmental conditions that slow reaction kinetics (e.g., lower temperature or nutrient concentration). However, for applications where a high level of control over the microorganism is required (e.g., hybrid living building materials, functionally-graded composites and films), engineered E. coli may facilitate the desired level of tailorability over crystal morphology and material properties.

conclusions
In this study, we performed physical, chemical, and microscale mechanical assessments of biogenic CaCO 3 crystals deposited by native ureolytic (S. pasteurii) and engineered (E. coli) ureolytic microorganisms. Our key findings were that calcite crystal size and nanoindentation moduli could be modulated through metabolically engineering E. coli strains to have different urease activities and precipitation kinetics. Crucially, hardness was not changed amongst microorganisms, demonstrating that a choice of higher or lower modulus for biogenic calcite can be made without compromising the resistance of the crystal to irreversible deformation. We also identified that CaP phases have a role in the nucleation and growth of biogenic calcite precipitated by engineered and native ureolytic species. Furthermore, increased CaP was negatively correlated with nanoindentation modulus for the high-urease activity engineered E. coli strain (E. coli HB101/pBU11). Taken together, the results from this study demonstrate, for the first time, that the morphology and material properties of biogenic CaCO 3 can be tailored with engineered ureolytic microorganisms-a finding that enables new possibilities for bacteria-mediated design of advanced functional materials. www.nature.com/scientificreports www.nature.com/scientificreports/

Materials and Methods
Microorganisms and culture conditions. Three microorganisms were cultured for MICP experiments, including S. pasteurii ATCC 11859 and two engineered E. coli strains: E. coli HB101/ure-integration and E. coli HB101/pBU11. Pre-culture growth conditions for S. pasteurii, at 30 °C, were consistent with previous description 37 . Details of the construction of the engineered E. coli strains and pre-culture growth conditions are described in Liang et al. 24 . In summary, for the E. coli HB101/ure-integration strain, a single copy of the urease gene cluster from S. pasteurii was inserted into the genome using a CRISPR-based technique. E. coli HB101/pBU11 is a plasmid strain constructed with medium copy number (15-20 copies) of the urease gene cluster. Both E. coli strains were pre-cultured at 37 °C in lysogeny broth (LB) containing 50 µM NiCl 2 , which is necessary for ureolysis by E. coli 12 . Media for E. coli HB101/pBU11 also included ampicillin (100 µg/mL) for plasmid maintenance.
For the experiments, pre-cultured microorganisms were added (at OD 600 = 0.2) to sterile filtered urea-CaCl 2 media (20 g/L urea, 10 g/L NH 4 Cl, 3 g/L Difco nutrient broth, 25 mmol NaHCO 3 , 25 mmol CaCl 2 ). This media was consistent with media utilized in previous S. pasteurii biomineralization experiments 11 except NiCl 2 (5 µM) was also included in E. coli cultures. Cultures were prepared in 50 mL sterile plastic conical vials and incubated in a shaker at 20 °C, which is a temperature that represents expected conditions for engineering applications. Cultures were sacrificially sampled at four timepoints: 12h, 1d, 2d, and 7d. At the first three timepoints, three replicate cultures per microorganism were sampled. At the 7d timepoint, four replicates per microorganism were sampled to improve statistical power for nanoindentation measurements. Controls included (1) wild-type E. coli HB101 with no genetic modifications (Wild-Type control) in urea-CaCl 2 media and (2) urea-CaCl 2 media with no microorganisms (Media control). Controls were sampled in triplicate at 1d, 2d, and 7d.

Solution chemistry and reaction kinetics.
At each time point, 10 µL of culture was taken to measure colony forming units (CFUs). This assessment was performed in triplicate. The remaining volume of the 50 mL solution was then vacuum filtered, using 0.22 µm nylon filter paper (Sigma-Aldrich). Precipitates collected from this filtration were used to assess crystal morphology and phase (Section 2.3). From the filtrate, pH was measured on a calibrated Beckman-Coulter electrode. The filtrate was then adjusted with HNO 3 to pH 1.5 in order to stabilize solution chemistry for inductively coupled optical emission spectroscopy (ICP-OES) measurement of soluble Ca and P. These samples were further prepared using a modified version of the technique developed by Farrell, Matthes, and Mackie 38 . Briefly, 5 mL of a 7:3 mixture of hydrochloric acid and hydrofluoric acid and then 2 mL of nitric acid were added to digestion tubes and heated to 95 °C in a digestion block (HotBlock by Environmental Express) for approximately 2 hours. Samples were then cooled via water bath and brought up to 50 mL with a 1.5% by weight boric acid solution. The samples were then reheated to 95 °C for about 15 minutes and cooled before analysis. Samples were analyzed with a Perkin Elmer SCIEX ICP-OES, model # Elan DRC-e.
The solution depletion rates of Ca (dCa/dt) and P (dP/dt) were evaluated by dividing the difference in elemental concentration between two timepoints by the elapsed time. Because cultures were sacrificially sampled at each timepoint (i.e., different cultures at 12h, 1d, 2d, 7d), the average Ca and P value amongst replicates was used for depletion rate assessment. Depletion rates were calculated at all timepoints for ureolytic microorganisms and controls.
Urease activity was assessed for additional cultures prepared identically to those described in Section 2.1. Briefly, 5 mL of unfiltered solution for each ureolytic microorganism was collected at 1d, 2d, and 7d. These samples were centrifuged and sonicated, and the extract collected for urease activity analysis (Urease Activity Assay Kit, Sigma-Aldrich 357 MAK120).
Mineralogical assessment of crystal morphology and phase. Precipitates collected from filtration were dried overnight at room temperature in a hood before analysis. These precipitates were divided into samples for scanning electron microscopy (SEM), SEM-energy dispersive spectroscopy (SEM-EDS), and XRD analyses. A portion of the 7d precipitates was also allocated for nanoindentation and site-matched Raman spectroscopy analyses. For SEM characterization of crystal morphology, precipitates were mounted on carbon tape and sputter-coated with a 10 nm coating of platinum. Crystal morphology was evaluated with a Hitachi SU3500 set at an accelerating voltage of 5 kV, working distance of 8 mm, and spot size of 30. Crystals were visualized at 500x and 1500x for each sample. Crystal area was evaluated at 7d for each replicate using 500x images and ImageJ. All in-focus and well-defined crystals were measured for each image. For SEM-EDS, samples were sputter-coated with gold. Composition of crystals was assessed with a JEOL 6480 at an accelerating voltage of 15 kV, working distance of 10 mm, and spot size of 50. For XRD, dried precipitates were rinsed in DI water and dried on no-background silicon disks for fingerprint XRD. A Siemens D500 X-ray diffractometer analyzed samples from 5 to 65° 2θ using CuKαX-ray radiation with a stepsize of 0.02° and a dwell time of 2 s per step. Crystal phases were identified using Jade software (MDI, version 9) and the International Centre for Diffraction Data (ICDD) 2003 database.
Samples for nanoindentation and Raman spectroscopy were prepared by embedding precipitates in Epoxy (EpoxiCure 2, Buehler). Embedded disks were then ground with 600 and 1200 grit papers and polished with progressively fine diamond-oil suspension (9, 6, 3, 1, 0.025 µm grit) (Buehler). A polished disk was prepared for each of the four replicates per microorganism. After nanoindentation and Raman spectroscopy (described below), embedded crystals were assessed for elemental composition using wavelength-dispersive spectroscopy (WDS). The polished disks were first sputter-coated with 20 nm of carbon. Elemental maps for Ca and P were obtained for several crystals for each microorganism using a JEOL-8230, using 30 nA, 50 ms dwell, 2 µm spot size and collecting over 512 × 512 pixels. The elemental ratio of P:Ca was calculated for all WDS maps.
Following Raman spectroscopy, nanoindentation was performed with a Berkovich tip in displacement-control mode to a maximum displacement of 500 nm (Hysitron TI950). The load function was 10 s ramp, 10 s hold, 10 s unload. Indents were placed a minimum of 10 µm away from any sample edges and previous indents. Indents were excluded from analysis if load-displacement curves showed evidence of a compaction event or poor surface contact ( Supplementary Information Fig. 1). A total of 15-20 indents with corresponding Raman spectra were analyzed for each of the four disks per microorganism, with a total of 68-69 locations assessed for each microorganism.
Load-displacement curves were analyzed using the Oliver-Pharr approach 42 and custom MATLAB script. Briefly, stiffness (S) was determined from the derivative of the initial portion of the unloading curve with the assumption of elastic behavior. The tip contact area (A c ) is calibrated as a function of contact depth (h c ). The reduced modulus (E r ) was calculated according to: The reduced modulus is a function of sample (s) and tip (t) modulus (E s , E t ) as well as Poisson's ratios of the sample (v s ) and tip (v t ): (2) If tip properties are known (here, E t = 1140 GPa, v t = 0.07) and if the sample Poisson's ratio is known, the sample modulus can be determined. Here, the Poisson's ratio of biogenic CaCO 3 was not determined. Therefore, we present the indentation modulus (E i ), in which no assumption of Poisson's ratio is necessary: The value of E i quantifies the resistance to elastic (reversible) deformation during indentation. Indentation hardness (H i ) was also calculated for every indent by measuring the maximum load and the corresponding contact area created by the maximum load: i c H i necessarily incorporates contributions from both elastic and inelastic deformations. For materials that are not perfectly plastic, such as CaCO 3 , H i is different than hardness (H) assessed by bulk-scale techniques (e.g., Rockwell testing) 36 . However, assuming elastic-plastic material deformation-an assumption that is not valid for substantially viscoelastic materials-a constitutive relationship exists between hardness, H, and H i and E i 36 : The resulting value of H is independent of E i . The value β is the equivalent cone angle of the indenter and equals 70.32° for a Berkovich tip. For each nanoindentation curve, E i , H i , and H were calculated. The ratio of H i / E i was also calculated, which is interpreted as the relative resistance to inelastic deformation in the contact zone during nanoindentation 20,36 . Statistical methods. Measurements from nanoindentation (E i , H i , H, H i /E i ) as well as morphology (crystal area) were averaged for each replicate for each microorganism. Nanoindentation measures were compared between microorganisms (S. pasteurii, E. coli HB101/ure-integration, E. coli HB101/pBU11) using one-factor ANOVA. Post-hoc testing was performed using a Tukey correction to adjust critical alpha for family-wise error. The definition of significance was set a priori to p < 0.05. For all models, residuals were checked for normality and homoscedasticity. The Pearson product-moment correlation was performed to assess the strength of the relationship between Raman spectroscopy measures of crystal size and composition, as well as crystal nanoindentation modulus. All analyses were performed using Minitab (v18).