Anal. Chem. 2009, 81, 473–478
Screening of Biomineralization Using Microfluidics Huabing Yin,*,† Bozhi Ji,‡ Phillip S. Dobson,† Khedidja Mosbahi,§ Andrew Glidle,† Nikolaj Gadegaard,† Andy Freer,§ Jonathan M. Cooper,† and Maggie Cusack*,‡ Department of Electronics and Electrical Engineering, University of Glasgow, Glasgow, G12 8QQ, U.K., Department of Geographical and Earth Sciences, University of Glasgow, Glasgow G12 8QQ, U.K., and Glasgow Biomedical Research Centre, University Place, Glasgow G12 8TA, U.K. Biomineralization is the process where biological systems produce well-defined composite structures such as shell, teeth, and bones. Currently, there is substantial momentum to investigate the processes implicated in biomineralization and to unravel the complex roles of proteins in the control of polymorph switching. An understanding of these processes may have wide-ranging significance in health care applications and in the development of advanced materials. We have demonstrated a microfluidic approach toward these challenges. A reversibly sealed T-junction microfluidic device was fabricated to investigate the influence of extrapallial (EP) fluid proteins in polymorph control of crystal formation in mollusk shells. A range of conditions were investigated on chip, allowing fast screening of various combinations of ion, pH, and protein concentrations. The dynamic formation of crystals was monitored on chip and combined with in situ Raman to reveal the polymorph in real time. To this end, we have demonstrated the unique advantages of this integrated approach in understanding the processes involved in biomineralization and revealing information that is impossible to obtain using traditional methods. Biomineralization is the process whereby control is exerted by biological systems on the formation of minerals, defining both the polymorph that is formed and the orientation and crystal habit.1,2 Polymorph control is seen clearly in the shell of the mollusk Mytilus edulis (Figure 1A), where calcium carbonate, CaCO3, is deposited as calcite and aragonite, providing an extremely robust environment for the animal. Under laboratory conditions, this switch from low energy calcite to the metastable aragonite would require a temperature and pressure well above ambient. With the recruitment of certain proteins and polysaccharides, the mollusk can accomplish polymorph control at ambient conditions.3 This model has been proposed as a catalyst to stimulate a search for definitive evidence of a mechanism for biomineralization.3 It is already known that the extrapallial (EP) fluid, found in * To whom correspondence should be addressed. E-mail:
[email protected] (H.Y.) and
[email protected] (M.C). † Department of Electronics and Electrical Engineering, University of Glasgow. ‡ Department of Geographical and Earth Sciences, University of Glasgow. § Glasgow Biomedical Research Centre. (1) Weiner, S. J. Struct. Biol. 2008, 163 (3), 229–234. (2) Cusack, M.; Freer, A. Chem. Rev. 2008, DOI 10.1021/cr078270o. (3) Addadi, L.; Joester, D.; Nudelman, F.; Weiner, S. Chem.sEur. J. 2006, 12, 981–987. 10.1021/ac801980b CCC: $40.75 2009 American Chemical Society Published on Web 11/20/2008
Figure 1. (A) Scanning electron micrograph of a fracture section of the biomineralic shell of Mytilus edulis with calcite prisms to the exterior (top) and aragonite nacre to the interior (bottom). (B) Microfluidic chip used for the study.
the compartment between the shell and mantle tissue, contains many proteins that are involved in shell formation and are key protagonists in this process of biomineralization.4 In order to identify the influence of specific proteins involved in biomineralization, using microfluidics, we developed a rapid, low volume technique of screening protein activity in a highly controlled and reproducible manner, within a large parameter space. Traditionally, the Kitano process5 offers a robust way to grow calcite crystals and thereby assess protein function. However, this is a lengthy procedure that can take between 24-48 h to attain carbonate equilibrium and stable calcite crystals. This is not conducive to protein experiments since there is a possibility of denaturing sensitive biomolecules over this time span. In this later situation, however, the reduced diffusion times, afforded by microfluidic systems, offers distinct advantages. Typical microfluidic channels have dimensions of tens to hundreds of micrometers and have already found diverse applications in the life sciences,6,7 including drug discovery,8,9 diagnos(4) Cusack, M.; Parkinson, D.; Freer, A.; Pe´rez-Huerta, A.; Fallick, A. E.; Curry, G. B. Mineral. Mag. 2008, 72, 567–575. (5) Kitano, Y. Bull. Chem. Soc. Jpn. 1962, 35, 1973–1980. (6) Sims, C. E.; Allbritton, N. L. Lab Chip 2007, 7, 423–440.
Analytical Chemistry, Vol. 81, No. 1, January 1, 2009
473
Figure 2. On chip crystal formations with Na2CO3 and CaCl2 solutions in deionized water: (A) initial confluence of two streams; (B) single line crystal formation at the initial interface; (C) crystal formation on the CO32- side in the upper channel; and (D) uniform distribution of crystals across the channel near the channel outlet. Solutions of 10 mM Na2CO3 and 10 mM CaCl2 were applied for crystal formation from parts A-D.
tics,10 combinatory chemistry,11 and protein crystallization.12,13 Such devices provide a large surface to volume ratio with a low thermal mass.6 Under the correct flow conditions, mass transfer can also be controlled such that there are well-defined (diffusion limited) reagent and temperature gradients. We have developed here a novel platform for understanding the subtle reactions and processes of biomineralization, with the possibility of better understanding the biological process and, in the future, of creating new biomimetic materials. By combining a T-junction microfluidic device (to create a well defined reagent mixing profile) with in situ Raman microspectroscopy, we show that we can characterize the polymorph of the resultant crystals and gain real time information about the structure of individual crystals formed across a wide range of experimental conditions. EXPERIMENTAL SECTION The EP fluid of fresh blue mussels M. edulis (Linnaeus, 1758) was extracted from the space located between the outer mantle epithelium and shell by syringe. After extraction, heterogeneous EP fluid was centrifuged at 28000g for 30 min at 4 °C, and the supernatant was then dialysed against 50 mM Tris pH 7.5 at 4 °C, overnight. T-junction microfluidic chips (Figure 1B) were made by casting polydimethylsiloxane (PDMS) elastomer against a silicon master as described previously.14 The total length of the serpentine microchannel was 5 cm, with all microchannels having a uniform cross section of 130 µm × 100 µm. The PDMS was clamped against a cleaned glass substrate, forming a reversible seal, and microfluidic interconnects were created, using established methods. Reagents were delivered through the two inlets at a range of flow rates using syringe pumps (KD Scientific). Microfluidic Experimental Settings and in Vitro Mineral Crystallization. CaCO3 crystals were formed from aqueous solutions of calcium chloride (CaCl2) and sodium carbonate (Na2CO3). Small volumes of CaCl2 and Na2CO3 solutions (between 2 and 10 µL) were delivered using carrier liquids at the same flow rate through the two inlets (1 and 2, respectively). Initially (7) Teh, S. Y.; Lin, R.; Hung, L. H.; Lee, A. P. Lab Chip 2008, 8, 198–220. (8) Kang, L. F.; Chung, B. G.; Langer, R.; Khademhosseini, A. Drug Discovery Today 2008, 13, 1–13. (9) Weigl, B. H.; Bardell, R. L.; Cabrera, C. R. Adv. Drug Delivery Rev. 2003, 55, 349–377. (10) Price, C. P.; Kricka, L. J. Clin. Chem. 2007, 53, 1665–1675. (11) Khandurina, J.; Guttman, A. Curr. Opin. Chem. Biol. 2002, 6, 359–366. (12) Lau, B. T. C.; Baitz, C. A.; Dong, X. P.; Hansen, C. L. J. Am. Chem. Soc. 2007, 129, 454–455. (13) Sauter, C.; Dhouib, K.; Lorber, B. Cryst. Growth Des. 2007, 7, 2247–2250. (14) Yin, H. B.; Zhang, X. L.; Pattrick, N.; Klauke, N.; Cordingley, H. C.; Haswell, S. J.; Cooper, J. M. Anal. Chem. 2007, 79, 7139–7144.
474
Analytical Chemistry, Vol. 81, No. 1, January 1, 2009
experiments were performed with solutions of different concentrations prepared in deionized water to monitor crystal formation without contributions from protein or buffer: e.g., 10 mM CaCl2 (inlet 1) with 10 mM Na2CO3 (inlet 2). In a second set of experiments CaCl2 and Na2CO3 were maintained at constant concentrations of 20 mM in 50 mM Tris buffer pH 7.5 to provide a buffered solution, and a crude extract of EP proteins was added to determine the influence that buffer may have on crystallization. Bovine serum albumin (BSA, 1 mg/mL) buffered in 50 mM Tris pH 7.5 was used as a nonfunctional protein control. Local pH was measured under flow conditions through the addition of the indicator dye, phenolphthalein (0.1 mM), to both solutions. Optical microscopy was used to record in situ crystal formation and the resultant crystal morphogenesis, while Raman spectra were obtained on a LabRam INV Raman spectrometer (Jobin Yvon Ltd.) using a 632.81 nm laser beam as the excitation light source. An integration time of 10 s was used for each spectrum, and an average of two spectra was used in the analysis. Once the two fluid streams formed an interface, in situ Raman spectra were recorded continuously as crystal formation progressed. The crystals that formed were subsequently fixed in pure methanol, and the PDMS channels were removed and imaged using a Quanta 200F SEM (FEI) using low vacuum mode at an accelerating voltage of 10 kV for uncoated samples. RESULTS AND DISCUSSION Device Design and on Chip Crystallization. In nature, latent diffusion-controlled processes are behind the formation of many of the intricate structures found in inorganic matter. These processes are undoubtedly important in the formation and arrangement of calcite and/or aragonite crystals in the shells of mollusks, such as Mytilus edulis. Nucleation and growth of calcite or aragonite crystals is a thermodynamic and kinetic process15 that is dependent on the concentration of components present and the surrounding microenvironment. Traditional laboratory methods used to understand and control polymorph formation have been performed iteratively utilizing a range of experimental conditions. Given appropriate flow conditions, it is well-known that complex and stable concentration gradients can be readily achieved in microfluidic devices.16 By designing a T-junction fluidic device with a Reynolds number, Re, of 9.5 using 548 nm illumination).
(17) Verdoes, D.; Kashchiev, D.; van Rosmalen, G. M. J. Cryst. Growth 1992, 118, 401–413. (18) Kelland, J. J. Am. Chem. Soc. 1937, 59, 1675–1678.
CaCl2 solution (pH 4.9) inhibits the formation of stable crystals, whereas the alkalinity of the Na2CO3 side (pH 9.5) enables crystal Analytical Chemistry, Vol. 81, No. 1, January 1, 2009
475
Figure 4. Off-chip characterization: (A) SEM images of crystals formed on chip; (B) representative Raman spectrum of crystals formed on chip using Na2CO3 and CaCl2 solutions in deionized water. All the crystals were calcite, which has characteristic peaks at 281, 712, and 1086 cm-1.
Figure 5. Crystal formation in the presence of EP proteins: (A) initially, unstable branched crystals were formed and (B) subsequently, they transform to stable ovoid-shaped crystals.
growth. To illustrate the variation in local pH, phenolphthalein (0.1 mM) was added to both solutions. Figure 3C clearly showed that the boundary between the initial low pH of CaCl2 solution and high pH of the Na2CO3 solution is maintained for a significant distance along the channel. It was observed that this pH gradient disappeared with time (i.e., distance along the channel), and it is correlated with uniformly distributed crystals, as shown in Figure 2D. This phenomenon reveals that the whole system is driven both by pH and by the diffusion of ions and serves to demonstrate that microfluidics, enhanced by fluidic computations, is a powerful method for the interrogation of a large set of complex conditions within a single experiment. Ex Situ Micro-Raman. The reversible seal of the microfluidic chip enabled the retrieval of crystals formed for further off-chip characterization. Polymorph and structural identification using micro-Raman are shown in parts A and B of Figure 4. Well separated crystals, following the interface between the two streams, are shown in Figure 4A. Raman measurements from about 30 randomly selected crystals (5 in each experiment) show the same calcite spectra (Figure 4B), which features the strongest peak at 1086 cm-1, the vibration band of calcite 710 cm-1, and lattice mode vibration peak at 281 cm-1.19 No characteristic Raman peaks for aragonite (a single narrow peak at 205 cm-1) and vaterite (a narrow double-peak at ∼1088 cm-1 and broad peaks at ∼260 and 300 cm-1)19 were observed. This demonstrates that calcite is the dominant polymorph formed on chip for a wide range of concentrations and ratios of Na2CO3 and CaCl2. Screening of Biomineralization on Chip. In reality, the formation of biominerals in nature is not as simple as the model (19) Dandeu, A.; Humbert, B.; Carteret, C.; Muhr, H.; Plasari, E.; Bossoutrot, J. M. Chem. Eng. Technol. 2006, 29, 221–225.
476
Analytical Chemistry, Vol. 81, No. 1, January 1, 2009
system studied above using Na2CO3 and CaCl2 solutions in deionized water. The presence of functional proteins that are normally in trace quantities, as well as ions and carbohydrates in the EP fluid, all play important roles in regulating crystal formation. Revealing the in vivo biomineralization processes remains a significant challenge, and the capability of the microfluidic system provides the opportunity for the rapid screening of the functional proteins that influence crystal formation, using small volumes. As proof of concept, EP protein extraction mixtures have been used directly to demonstrate on-chip screening. The reversible seal of the microfluidic chip enabled the retrieval of crystals formed for further off-chip characterization using SEM. As shown in Figure 5A, the presence of crude EP proteins in buffered solution results in the initial formation of branched crystals, similar to the morphology of aragonite.19 However, these crystals are not stable and subsequently turn into more stable ovoid crystals (Figure 5B and Figure 6A). This phenomenon was not observed when either Tris buffer (alone) or BSA was added. In both controls, only smooth cubic and spherical crystals are formed throughout the whole process, as shown in Figure 6B,C. Highresolution SEM analysis further revealed that the ovoid crystals had a rough surface (Figure 6B, insert) in contrast to the smooth surfaces of the cubic and spherical crystals formed in the two control experiments. It was found that the ovoid crystals, formed in the presence of EP proteins, have the same Raman spectra (Figure 6B). The strong single peak at 1086 cm-1 and two weak peaks at 710 and 281 cm-1 indicates calcite (Figure 4B). In contrast to the calcite formed in the control conditions, the background spectrum of ovoid-shaped crystals is higher, and the intensity ratios of the main peak (1086 cm-1) compared to the two weak peaks are lower. This may be a consequence of the scattering from the rough surface (Figure 6B) or the differences in density and internal structures of the crystals formed in the presence of the EP proteins. The cubic and spherical crystals formed in the controls are calcite and vaterite, respectively, as identified by their Raman spectra (Figure 6D,F).19 The vaterite formation is due to the presence of N-rich Tris buffer, similar to its formation using Ca(NO3)2.19 The crude extract of EP proteins demonstrates a significant influence on the morphology of crystal formation, whereas BSA has no effect.
Figure 6. Rapid screening of biomineralization on the chip: (A) In the presence of EP proteins, ovoid-shaped crystals are dominant. (B) The high-resolution SEM image reveals a rough organic coating on the ovoid-shaped crystals. The crystal has a calcite-like Raman spectra but with much higher background scattering. (C and E) On chip crystal formation in control experiments with nonfunctional protein BSA (C) and with pure Tris buffer (E). Cubic and spherical crystals are formed in both controls, as indicated by single head arrow and triple-head arrows in the images, respectively. (D) High-resolution image and Raman spectrum of the cubic crystals. Raman spectra show its polymorph as calcite. (F) High-resolution image and Raman spectrum of the spherical crystals. The Raman spectrum shows its polymorph of vaterite.
In Situ Optical and Raman Study on Chip. Crystal formation is a dynamic process involving crystal nucleation, precipitation, and growth.17 Induction time, the period between the mixing of solution and the initial formation of the crystal nuclei, is an important parameter in the crystal forming mechanism. With microfludics, induction time can be readily measured by time course recording (see Supporting Information). A clear interface forms upon the meeting of two solutions, defining the starting point. The induction time for the crystal formation in the presence of the EP extraction mixture is approximately 30% faster than for the formation in Tris buffer and for the formation in the BSA control, again indicating the influence of EP proteins on crystal formation.
Optical recording alone is not able to determine which crystal polymorph is present, since one polymorph can be represented by different gross morphologies (shown in both cubic and ovoid calcite). Monitoring crystal polymorph development in in vitro biomineralization experiments is crucial to establish the mechanisms involved. For this purpose, we integrated the microfluidic system with Raman spectroscopy to detect the polymorph changes on chip in the presence of EP proteins. Figure 7 shows a series of real-time Raman spectra taken every 40 s during the crystal formation in the presence of EP proteins. While no Raman peaks were present during the induction time, strong CaCO3 peaks appeared at 1086 cm-1 upon the initial formation of crystals. As the crystals grow, the intensity of this and the peaks at 281 cm-1 Analytical Chemistry, Vol. 81, No. 1, January 1, 2009
477
Figure 7. In situ Raman monitoring of on chip crystal formation in the presence of EP proteins. Consecutive spectra were collected every 40 s at the initial joining point of the two solutions, where crystals formed. The peak at 1086 cm-1 appeared upon the initial formation of crystals, and the intensity of this and the peak at 281 cm-1 increase with time (inserts A and B).
increase with time (Figure 7, inserts A and B), enabling polymorph formation to be monitored in real time. CONCLUSION Currently, there is significant impetus to understand the processes implicated in biomineralization and to unravel the complex involvement of a number of proteins essential to polymorph switching. A greater understanding of these processes will lead to advances in materials design where composites could be manufactured with predetermined physical and material properties. We have demonstrated a microfluidic approach to enable the rapid assessment of screening conditions. A reversibly sealed microfluidics device was developed to create designed concentration profiles, allowing a range of mineralization conditions to be investigated on-chip. This was used to investigate the influence of EP proteins on polymorph formation. We have also shown proof of principle measurements that in situ Raman is a valuable means
478
Analytical Chemistry, Vol. 81, No. 1, January 1, 2009
of interpreting polymorphs in real time. This combination of techniques provides a new tool to provide evidence for nacre formation and to advance our knowledge of organic-mineral interactions in biomineralization. These proof-of-concept measurements clearly show the advantages of adopting a microfluidics approach over traditional methods. The technique described requires minimal quantities of functional proteins yet produces high quality, rapid information across a range of experimental conditions, offering an ideal platform for investigations involving biomineral proteins. The ability to localize crystal formation permits in situ monitoring of kinetic processes, with the potential to reveal mechanisms and mechanistic information. In the future, we intend to incorporate micropatterning and self-assembled monolayers (SAMs) into the microfluidic structure to provide a pattern of sites for ordered nucleation in the carbonate system,20,21 thereby generating a nano/microengineered surface similar to that found in many biological shells. ACKNOWLEDGMENT H.Y. is supported by Royal Society of Edinburgh (RSE) as a RSE Personal Research Fellow. B.J. holds a Lord Kelvin/Adam Smith scholarship from the University of Glasgow. K.M., A.F., and M.C. gratefully acknowledge support from BBSRC Life Sciences Interface with the EPSRC (Grant BB/E025110/1). SUPPORTING INFORMATION AVAILABLE Video of the time course recording of crystal formation on a chip. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review September 18, 2008. Accepted October 22, 2008. AC801980B (20) Aizenberg, J. Adv. Mater. 2004, 16, 1295–1302. (21) Aizenberg, J.; Black, A.; Whitesides, G. M. Nature 1999, 398, 495–498.
