Microfluidics for studying the deep underground biosphere: from applications to fundamentals

Abstract

In this review, selected examples are presented to demonstrate how microfluidic approaches can be utilized for investigating microbial life from deep geological environments, both from practical and fundamental perspectives. Beginning with the definition of the deep underground biosphere and the conventional experimental techniques employed for these studies, the use of microfluidic systems for accessing critical parameters of deep life in geological environments at the microscale is subsequently addressed (high pressure, high temperature, low volume). Microfluidics can simulate a range of environmental conditions on a chip, enabling rapid and comprehensive studies of microbial behavior and interactions in subsurface ecosystems, such as simulations of porous systems, interactions among microbes/microbes/minerals, and gradient cultivation. Transparent microreactors allow real-time, noninvasive analysis of microbial activities (microscopy, Raman spectroscopy, FTIR microspectroscopy, etc.), providing detailed insights into biogeochemical processes and facilitating pore-scale analysis. Finally, the current challenges and opportunities to expand the use of microfluidic methodologies for studying and monitoring the deep biosphere in real time under deep underground conditions are discussed.

Introduction

The depths of our planet, particularly in continental systems, have gained increasing attention, as they have been suggested to be the home of millions of undiscovered microbial organisms (Bar-On et al. 2018, Magnabosco et al. 2018). This intraterrestrial world has rarely been investigated, especially for technical and economic issues, although several national and international multidisciplinary programs have emerged to facilitate exploration and determine the impact of autochthonous microorganisms on deep geochemical cycles, such as the International Ocean Discovery Program, the International Continental Scientific Drilling Program or the Deep Carbon Observatory’s Census of Deep Life (Pedersen 2000, Fry et al. 2008, Mangelsdorf and Kallmeyer 2010, Edwards et al. 2012, Colwell and D'Hondt 2013, McMahon and Parnell 2014, Escudero et al. 2018).

The deep underground biosphere and its investigation. In addition to the absence of sunlight, the key parameter driving the definition of deep environments—to distinguish them from classical surface environments—is pressure (typically p > 10 MPa). The lithostatic pressure can exhibit strong lateral and depth heterogeneity, such as in three typical geological formations: (i) unmineable coal reservoirs, (ii) oil reservoirs, and (iii) deep saline aquifers (De Silva et al. 2015, Jayasekara et al. 2020, Liu and Grana 2020). Investigating the role of the deep biosphere in global cycles and its interaction with human activities is not straightforward. These environments are extremely difficult to access and require scientific drilling, the implementation of protocols to recover characteristic samples or sampling campaigns in mines such as those in South Africa or China (Takai et al. 2001, Zhang et al. 2005, Mangelsdorf and Kallmeyer 2010, Ranchou-Peyruse et al. 2023). The pioneers of high-pressure (HP) microbiology developed devices to sample and investigate deep-living organisms at the laboratory scale, including HP flow environments (Jannasch et al. 1996, Parkes et al. 2009). Such experimental breakthroughs have demonstrated the extended pressure and temperature limits of life (Yayanos et al. 1981, Takai et al. 2008, Zeng et al. 2009) and offered a better understanding of microbes’ adaptation to harsh environments, including biophysiological mechanisms (Cario et al. 2015, 2016, Amrani et al. 2016). However, the current equipment mostly uses large volumes and cannot easily accommodate in situ characterization techniques. Thus, Garel and colleagues (2019) developed an equipressure transfer procedure via HP bottles (stainless steel coated with PEEK), allowing subsampling to carry out a series of experiments. This means that conventional HP microbiology methods (phenotyping, characterization, etc.) are time-consuming and often favor well-known fast-growing organisms that outcompete others in culture. In addition, sample decompression, which is required for microbial characterization, is a common process that can skew community diversity and structure, and isolate physiology (Jayasekara et al. 2020). The main challenge in reproducing acceptable deep environmental conditions in the laboratory is therefore the need for suitable experimental tools in terms of the observation scale and pressure/temperature range.

The rise of microfluidics. Microfluidics is a field of research that addresses the precise manipulation of fluids at small scales, typically from microliter to picoliter volumes. This interdisciplinary area combines principles from physics, engineering, chemistry, and biology to create devices known as “lab-on-a-chip” systems. Microfluidic devices utilize micron-sized channels, chambers, and valves to integrate various functionalities, such as mixing, separation, and analysis, into compact microscale platforms, typically of the size of a credit card (see Fig. 1). By enabling precise control over fluids at the microscale, microfluidics has revolutionized experimentation and analysis, opening new avenues for scientific discovery and technological innovation. Microfluidic tools have been extensively used over the past 20 years for microbiology (El-Ali et al. 2006, Duncombe et al. 2015), thermodynamics (Pinho et al. 2014), hydrodynamics (Zhang et al. 2018), materials synthesis (Marre and Jensen 2010), and chemistry (Hartman and Jensen 2009, Jensen et al. 2014). Microfluidic tools have inherent advantages, such as low sample consumption, faster equilibrium times, better control of the experimental conditions, flexibility in terms of designs for recreating complex geometries (Morais et al. 2020), and high-throughput capabilities. For example, we recently demonstrated the use of confined microfluidic chambers for HP microbial phenotyping (Cario et al. 2022) under anoxic conditions, whereas others used similar approaches to simulate and observe fluid mixtures in real time to understand precipitation mechanisms during pH variations in the case of hydrothermal vents on the ocean floor (Weingart et al. 2023). On the basis of these characteristics, microfluidics has largely been involved in accelerating discoveries in microbiology, although far focusing almost exclusively on biomedical and health applications along with investigations of human cells. These studies considered only pressure and temperature conditions close to ambient conditions with “conventional” laboratory model microorganism strains and were mostly interested in investigating fundamental microbial mechanisms and their responses to stimuli. Only a few reports exist on applications of microfluidics for environmental microbiology, such as planktonic and microalgal research (Girault et al. 2019, Wang et al. 2020), but again, these reports are close to ambient conditions. Nevertheless, given the flexibility provided by photolithography, it is rather straightforward to reproduce “micromodels” mimicking the porous geometry of deep underground environments, which has driven the use of microfluidics to investigate—at the laboratory scale—subsurface processes, particularly those linked with human activities [geological CO₂ storage, enhanced oil recovery (EOR), etc.] (Lifton 2016, Morais et al. 2016, Morais et al. 2020). The resulting microfluidic devices generally consist of 2D patterns extruded in the third shallower dimension on a substrate [glass, silicon, PMMA (poly(methyl methacrylate)] or cast on a mold [polydimethylsiloxane (PDMS)] and chemically or physically sealed with a transparent material (Morais et al. 2020). However, it is important to note that access to 3D micromodels is also possible owing to multiple etching steps, leading to multidepth geometrical profiles (Park et al. 2015). Additionally, minerals can also be inserted inside microchannel geometries to reproduce full 3D porous media (Bowden et al. 2016). These micromodels have demonstrated that they can be powerful tools to complement more conventional experimental methods, such as core flood experiments, particularly when they are compatible with experimental work performed under harsh conditions. As an illustration, Table 1 summarizes some advantages and limitations of conventional and microfluidic approaches for investigating the deep underground biosphere.

Hence, investigating the biological processes in micromodels enables real-time observations and monitoring of microorganism metabolism in realistic microporous media (Sun et al. 2011, Xu et al. 2013, Tang et al. 2014, Lee et al. 2017, Almeida et al. 2018).

On the basis of these advantages, through selected examples, current studies concerning the use of micromodels for studying the interactions between the deep biosphere and a selection of human activities in deep geological porous reservoirs are presented in this review. Then, several challenges and opportunities related to the utilization of microreactors for such studies are discussed. High-pressure and high-temperature transparent microfluidic systems are powerful tools whose use deserves to be developed to simulate exploitation scenarios and generate missing experimental data for large-scale modeling of the evolution of exploited reservoirs and address certain fundamental questions related to deep underground life (Fig. 1).

Microfluidics approaches for evaluating the impact of microbiology on deep underground anthropogenic activities

Micromodels have been extensively used to investigate several bioprocesses occurring in geological environments at the pore scale. These micromodels provide critical physicochemical information, which are some keys to driving the efficient injection and recovery of fluids, both in deep geological reservoirs and in geochemical reactions, reactive flows, or biofilm formation and monitoring.

We discuss hereafter some selected examples of the use of microfluidics and micromodels for accessing a deeper understanding of how the deep biosphere interacts with human exploitation of geology underground.

Biosurfactant generation influence on interfacial phenomena

Microbial enhanced oil recovery (MEOR) is a technology developed to accelerate and increase underground oil recovery via the use of microorganisms. A combination of various mechanisms improves the sweep and displacement efficiency. In the literature, the following mechanisms are commonly quoted and investigated:

• ‐ An improvement in mobility due to changes in oil and water/brine viscosities is mainly related to the dissolution of gases (CO₂, H₂, N₂, and CH₄) produced by microbial metabolism (Alkan et al. 2014, Sugai et al. 2014).

• ‐ An increase in microscopic sweep efficiency due to changes in interfacial tension (IFT) (Paulsen et al. 1999, Nourani et al. 2007, Armstrong and Wildenschild 2012a) and wettability (Al-Raoush 2009, Polson et al. 2010, Armstrong and Wildenschild 2012b) decreases capillary forces.

• ‐ Biofilm formation (bio-plugging) reduces the permeability in high-permeability zones or fractures (Jenneman et al. 1984, Gray et al. 2008, Kaster et al. 2012, Karambeigi et al. 2013) and causes the corrosion of pipelines (production of H₂S) in some cases (Zhou et al. 2022).

• ‐ The production of acids can lead to rock dissolution.

The change in wettability and the reduction in IFT caused by biosurfactants are key parameters for MEOR processes. They strongly affect the fluid displacement mechanisms at the microscale, and an increase of three orders of magnitude is necessary to displace oil from capillaries of 10–100 µm (Reed and Healy 1977, Kowalewski et al. 2006, Crescente et al. 2008, Karimi et al. 2012). Hence, microfluidics may help overcome the lack of microscale information. Indeed, over the past 30 years, microfluidics approaches have proven that they are indubitably suitable for studying wettability effects on drainage or imbibition for problematic EOR or geological CO₂ storage and sequestration investigations (Bora et al. 2000, Sohrabi et al. 2001, Nguyen et al. 2002, Jamaloei and Kharrat 2009, Lifton 2016, Gerami et al. 2018, Cao et al. 2019).

Bryant et al. studied the feasibility of enhancing oil recovery via the injection of microorganisms into a reservoir (Bryant and Douglas 1988, Bryant 1990, Buciak et al. 1996). These microorganisms provide amphiphilic molecules that act as biosurfactants, leading to the detachment of the oil from the reservoir rocks. Studies have shown that Bacillus licheniformis, which was isolated from a pristine oilfield, was the most effective (da Cunha et al. 2006). Three decades later, Afrapoli et al. explored the performance and mechanisms of model bacterial strain flooding at the pore scale (Afrapoli et al. 2011, 2012). They investigated the improvement in oil (dodecane) recovery essentially due to an alkane-oxidizing Rhodococcus sp. strain 094 in glass micromodels with three different configurations (the ratio of the pore diameter to the pore throat dimension) and different wettabilities (water wet or oil wet). The emulsification process occurs while the cells are growing exponentially and involves IFT reduction, wettability changes, and pore blocking mechanisms (Bredholt et al. 1998, Crescente et al. 2008). Microorganisms produce biosurfactants to access hydrophobic compounds, effectively reducing IFT, altering wettability, modifying flow patterns, and decreasing oil viscosity (Soudmand-asli et al. 2007). Indeed, a study with a glass micromodel showed that the use of a surfactant produced by B. subtilis led to additional oil recovery (Hadia et al. 2019). The presence of bacterial cells appears to amplify the oil recovery effect of biosurfactants, underscoring the increasing importance of interactions between microorganisms and their environment. White and colleagues published work on the interaction of a single oil droplet and bacterial strains via a microfluidic platform, ecology-on-a-chip (White et al. 2019). Owing to the layer-by-layer hydrophilic polyelectrolyte coating of the surface of the microchannel, an oil drop was immobilized for over a month. The authors observed the aggregate morphology as well as the interfacial response between the oil and the microbial environment. This approach represents a tremendous step forward in the matter of long-term observation of the MEOR process in microfluidics. While most MEOR studies in the literature are conducted under atmospheric pressure and oxic conditions, some strive to replicate actual environmental conditions more accurately. For example, Gaol and colleagues (2019) succeeded in investigating the pore-scale mechanisms of MEOR under relatively harsh processing conditions (37°C, 0.6 MPa). Nonetheless, these conditions are far from realistic reservoir conditions, typically between 6 MPa and 15 MPa and between 30°C and 80°C (Morais et al. 2016, 2020). The development of HP microfluidics could help address this gap (Marre and Jensen 2010). Moreover, microfluidic approaches combined with image reprocessing now make it possible to monitor the growth of microorganisms in real time while tracking their behavior in relation to the studied activity in general (Strobel et al. 2023) and their interaction with oil droplets in the particular case of MEOR (Gaol et al. 2021).

In addition to planktonic microorganisms, biofilm formation and monitoring are critical parameters to investigate when considering deep underground utilization since they can severely change some operating parameters, such as injectivity and reservoir rock stability. Thus, it is important to investigate the interactions between the considered substrates and the microorganisms. Microfluidics has been used to recreate several geometrical parameters from porous media while allowing in situ observations. These observations can provide deeper insight into the behavior of biofilms in confined geometries, as discussed in the next section.

Biofilm growth, metabolism, survival, and degradation in confined environments

In deep porous environments, microorganisms are known to grow mainly by attaching themselves to rock surfaces (Whitman et al. 1998, Griebler et al. 2002, McMahon and Parnell 2014, Bar-On et al. 2018). MEOR applications can create barriers through pore clogging, leading to a reduction in the permeability of the reservoir (Lappan and Fogler 1996, Kim and Fogler 2000). Conversely, investigations revealed that similar bioclogging mechanisms can contribute to decreases in porosity and permeability in geological CO₂ storage reservoirs (Thullner et al. 2002, Mitchell et al. 2009, Glatstein and Francisca 2014).

One of the pioneering works concerning biofilm processes at the microscale in porous media (Dunsmore et al. 2004) showed that, for the formation of a biofilm developed by cells of sulfate reducers (Desulfovibrio sp. EX265), a decrease in permeability was associated with biofilm growth. From that observation, correlations were established with the bioclogging of the pore space, which was previously detected at the core scale and field scale (Cunningham et al. 1991). Later, the same team coupled image analyses and chemical analyses to investigate the effects of nitrate treatment on oilfield microbial biofilms (Dunsmore et al. 2006). This time, they used glass capillary flow cells. They demonstrated the influence of nitrate treatment on sulfate reducer biofilms as well as changes in biofilm cell morphology. Additionally, chemical analyses revealed that the injection of nitrate ions into the flow cell changed the dominant metabolic process from sulfate to nitrate reduction. Another study including the use of a PDMS-based microfluidic channel revealed that the resistance of a biofilm developed by Pseudomonas aeruginosa was due mainly to the formation of stable extracellular polymeric substances, which provide a mechanical shield (Park et al. 2011).

Investigations at the microscale have been increasingly conducted to better understand how biofilms initiate their formation in different environments (Zhang et al. 2019a,b) and how biofilms impact mineral surface properties, geochemistry, and hydrodynamics in porous media (Liu et al. 2019). A recent study focused on sulfate reducer colonization in shale fractures, which promotes biosouring in oil and gas reservoirs (Zhou et al. 2022). In this context, a shale-based microfluidic flow cell reactor allows the quantification of both sulfate reduction rates and biomass growth while looking for biocide inhibition. This experiment revealed that biocides had little effect on biomass removal and opened shale fractures, which could contribute to the development of mitigation strategies for preventing environmental impacts on underground processes. As micromodels provide access to direct observations of biofilm evolution at the microscale (Lam et al. 2016, Yawata et al. 2016), they allow full visualization of the environmental effects of nutrient concentration, cells, surface properties (wettability, IFT) or shear stress on these biological formations (Skolimowski et al. 2010, Wilkins et al. 2014, Wang et al. 2018, Zhang et al. 2019). These microscale approaches are already used to study the properties of biofilms, such as stress/growth curves or deformation and cell detachment, through microrheological studies (Stoodley et al. 1999, 2002, Dunsmore et al. 2002, Klapper et al. 2002, Billings et al. 2015, Liu et al. 2023). Undeniably, these tools provide the ability to observe different phenomena in detail and collect quantitative data from the growth of the biofilm through the effects of starvation conditions (Kim and Fogler 2000), the estimation of the intrinsic permeability of the biofilm itself (Hassannayebi et al. 2021), or biodegradation in porous media at the pore scale (Vayenas et al. 2002).

Another interesting aspect of microfluidic approaches is the ability to decouple complex simultaneous effects such as the effect of nutrient flow velocity and the impact of shear stress on biofilm formation (Liu et al. 2019). Authors performed a hydrodynamic study on biofilm growth (from Thalassospira sp.) and experimented with biofilm detachment with a T-shaped microchannel. Direct observation through a glass microsystem allowed spatial localization of biofilm accumulation and evolution of the adhesive strength with various flow velocities and nutrient concentrations. This permitted (i) the determination of the conditions that lead to a plugging effect and (ii) the estimation of the capacity of biofilm formation to reduce the permeability of model deep underground porous media.

Sygouni and coauthors (Sygouni et al. 2016) studied the impact of CO₂ injections on a Pseudomonas putida biofilm in a micromodel. After inoculation in a glass/glass pore network and over the course of 18 days, nutrient broth was injected into the micromodel to induce the growth of the biofilm. A series of CO₂ injections were subsequently performed at different times. They observed a pH reduction immediately after each CO₂ injection, leading to cell stress and inducing partial detachment of the P. putida biofilm. However, this gas injection had only a temporary effect on the biofilm, as evidenced by the regrowth after the completion of each injection. As the nutrient flow compensates for the porous media and CO₂ is consumed by various buffering reactions, its concentration then decreases, and biofilm saturation progressively reaches its initial value. The same observation was made for the permeability, which recovered quickly from its initial low value (due to the presence of the biofilm at the beginning of the experiment). The study highlighted that biofilm age is crucial, with older biofilms being denser and more resistant to degradation.

In a study on MEOR efficiency with microorganisms and their associated bioproduced surfactants, biofilm formation was also observed (Armstrong and Wildenschild 2012a). In a 2D transparent porous medium, changes in the pore morphology through the propagation of the biofilm and the redirection of the flow were observed, creating a new preferential path pattern. This investigation confirmed the suitability of 2D porous media for studying the hydrodynamic impacts of biofilms (Aufrecht et al. 2019).

In addition to microbial production, the hydrodynamics of a whole geological reservoir can also be impacted by microbial activity. Indeed, chemical gradients and chemotaxis have been shown to influence the dispersion of degrading microorganisms within a porous matrix (Wang et al. 2012, 2015, de Anna et al. 2021).

Mineralization induced by microorganisms at the microscale

Biomineralization is yet another effect of deep underground microorganisms, especially in storage applications, that can affect the evolution of geological reservoirs, and it has been observed at the core and field scales that the precipitation of carbonates through biomineralization metabolism leads to clogging in porous reservoirs (Fujita et al. 2008, Dupraz et al. 2009a, Mitchell et al. 2010). Microfluidic technologies have the ability to perform biomineralization experiments in model porous media and allow direct observation of crystal growth during microbiologically induced calcite precipitation (MICP) processes (Singh et al. 2015, Wang et al. 2019a,b,2020). Several studies have shown that quantifying biomineral precipitation at the pore scale via image processing is possible (Kim et al. 2018, 2020, Xiao et al. 2021, 2024). These findings showed that characterizing biomineral precipitation in porous media is complex because of the involvement of multiphase reactive transport and interactive processes (Dejong et al. 2013, Hommel et al. 2016). For example, the nucleation and growth of crystals as well as the quantification of the crystal kinetics were performed in a Y-shaped sand-containing microchip. Such biomineralization process monitoring at the pore scale provides new insights for optimizing underground processes in geotechnical engineering, such as biocementation (Elmaloglou et al. 2020, Zheng et al. 2023). MICP experiments carried out under HP conditions (100 bar) with Sporosarcina pasteurii resulted in lower CaCO₃ crystallization yields than those at atmospheric pressure (Liu et al. 2023), possibly because of the lower activity of the considered model bacterium induced by HP conditions. For those interested in more information on MICP-based microfluidic techniques and modeling results, we recommend consulting the review by Xiao et al. (2023).

HP studies are still scarce, but integrating them with conventional HP microbiology tools could offer new insights and accelerate discoveries in this field, as discussed in the next section.

Challenges, opportunities, and limitations

Many microbiology laboratories have equipment for handling strictly anaerobic microorganisms (anaerobic chambers) and culturing them under controlled atmospheres at near-atmospheric pressure. However, one of the main difficulties faced by microbiologists specializing in deep environments is working safely under HP conditions, especially when working with flammable (CH₄, H₂), toxic and corrosive gases (CO₂, H₂S). Many microbiology laboratories cannot accommodate classical HP reactors (hundreds of milliliters to a few liters), mainly because of space, safety (explosion, gas leaks), and cost. Microfluidic tools can help overcome these constraints along with the possibility of performing experiments with high reproducibility given the improved control of the operating parameters compared with large-scale experiments and the ability to work at the scale of the microenvironments in which the microorganisms evolve. Microfluidic cells can be manipulated inside an anaerobic chamber to introduce anaerobic microorganisms. The cells are then taken out to be connected to the HP pump containing the strict volume of fluids required for the experiment. In addition, it is possible to screen various conditions simultaneously in a single experiment, generating high throughput with a low sample volume (Tang et al. 2014, Xu et al. 2017, Gantz et al. 2022). For example, the possibility of observing the effects of many chemical and physical variables simultaneously has not yet been fully exploited and could extend the scope of possibilities. Moreover, the ability to control gradients and dissociate the effect control separately for each experimental parameter at very small scales will allow strong multiplication of intercomparable results in a single experiment, as recently demonstrated for microbial phenotyping under pressure (Cario et al. 2022). Finally, the use of microfluidics lends itself naturally and easily to the study of organisms evolving in strictly anoxic conditions since such experiments can be perfectly well controlled (Mohr et al. 2010, Dickson 2019, Cario et al. 2022).

Fast screening challenge under harsh conditions at the microscale

On-chip multiple experimental conditions. It is possible to generate temperature gradients in microreactors to study, e.g. the optimal culture conditions of a microbial strain, understand the behavior of a microbial community, or even perform the isolation of microorganisms from complex communities originating from the deep subsurface. Concentration gradients can also be easily generated on a chip to evaluate, e.g. their influence on biofilm growth, attachment, and interaction with mineral surfaces (Jeong et al. 2014, Yawata et al. 2014, Zhang et al. 2019). These gradient approaches can also be applied to other physicochemical parameters, such as salinity, pH, and molecules of interest (chemotaxis, toxicity), hence providing ways to perform multiple experimental conditions in a single experiment. Such methodologies can therefore speed up the rate at which deep underground-related experiments can be performed.

High-pressure/high-temperature microfluidics. Conventional transparent micromodels cannot withstand HP conditions. However, it is possible to work at high pressures and temperatures with micromodels whose channels are etched in mechanically resistant materials such as silicon, onto which borosilicate glass slides are bonded (Trachsel et al. 2008, Marre and Jensen 2010). The use of such microfluidic reactors for microbiology studies under harsh conditions has already been demonstrated (Cario et al. 2022), but these require reflected light microscopy, which is not ideal for microbiology. However, such a setup precludes that the use of transmission optical microscopy would be efficient when coupled with transmitted light microscopy for optimum observation of microorganisms. While glass‒glass microreactors (Dietrich et al. 2005) have demonstrated HP capabilities up to 20 MPa or more (Tiggelaar et al. 2007), they are not able to cover a large range of conditions representative of the deep subsurface and are only transparent in the visible range. However, it is now possible to produce full sapphire microreactors capable of working at pressures of up to 80 MPa (Marre et al. 2021) while exhibiting transparency from UV to mid-IR, opening opportunities for further investigations under larger parameter window conditions. Hence, HP-compatible microfluidics tools exist that can be used in combination with more conventional HP reactors for studying the deep underground biosphere. Both experimental scales are equally important for investigating multiparameter biogeochemical processes from the pore scale to the core scale and further at the reservoir scale. Similarly, while microfluidics allows fast parameter screening, conventional larger scale reactors can generally achieve better quantitative fluid analysis because of the large processed volumes.

In situ live characterization of microbial activities in microfluidics. Working with transparent micromodels at a wide range of wavelengths opens up a wide range of possibilities for real-time analyses, starting with all types of microscopy: confocal, fluorescence, Raman spectroscopy (Liu et al. 2012, Ochoa-Vazquez et al. 2019), Fourier transform infrared (FTIR) microspectroscopy (Perro et al. 2016), and X-rays for analysis (Beuvier et al. 2015) and imaging (Morais et al. 2023). The combined use of such characterization techniques allows access to critical information on local biogeochemical processes and microbial metabolism without any external perturbation of the studied reactive medium. For example, Raman spectroscopy can be used to selectively identify different strains without any fluorescent labeling (Dhankhar et al. 2021, Lister et al. 2022, Rebrosova et al. 2022, Shakeel et al. 2022). Similarly, live imaging of living cells can be obtained by coupling FTIR microspectroscopy with microfluidics (Vaccari et al. 2012, Loutherback et al. 2016). The advantage of FTIR over Raman is that imaging can be performed (in contrast to dot-by-dot mapping), thus providing instant localized characterization. For example, Birarda and colleagues (2016) developed a low-cost microfluidic platform (IR-Live) compatible with synchrotron FTIR imaging to study living eukaryotic cells. On the basis of the specific vibrational footprint of molecules, they were able to identify and localize proteins and lipids without any preliminary labeling. This technique is perfectly well suited for monitoring microbial activities in biofilms (Holman et al. 2009, Pousti et al. 2018) and could also be coupled with Raman spectroscopy, as proven concepts have already been developed with other methodologies (Muhamadali et al. 2015, Rohman et al. 2019, Gieroba et al. 2020, Lima et al. 2022) and could easily be adapted to microfluidic experimentation.

Accessing realistic geometries and mineralogy on a chip

Microfluidic experiments performed to study applications related to the utilization of deep underground provide excellent control and monitoring of operating parameters such as fluid velocity and composition, temperature, and pressure (Morais et al. 2016). Additionally, micromodels enable laboratory simulations of the microporosity conditions characteristic of porous rocks encountered in deep environments, such as aquifers or hydrocarbon reservoirs. Owing to the development of microfabrication techniques, all types of designs can be created, providing a flexible way to meet the needs of mimicking realistic environments that are site-specific (see Section 2 from Morais et al. 2020). Most of the microreactors developed thus far for deep underground studies can be seen as 2D porous structures, which are perfectly suitable as “micromodels” to describe and study the biophysico-chemical mechanisms occurring inside; however, the third dimension is very small, and some researchers often raise concerns about the applicability of the results obtained in 2D compared with the real 3D world, which includes reactive fluid flow, dissolution/precipitation mechanisms, biofilm development, etc. Recent developments have allowed the access of 2.5- or 3D structures on a chip, which help account for vertical flows (similar to core flood experiments), but in a controlled manner owing to microfabrication protocols (Park et al. 2015). Of course, not all microbiologists have the know-how and equipment to produce such micromodels. However, an increasing number of companies are offering products that can be adapted for sale to microbiologists and do not necessarily require advanced competencies. When studying microbial processes in porous media, surface chemistry is equally critical to account for since those interfaces are known to regulate mainly the exchanges and compositions of the fluids that surround the microbial cells. Hence, the use or inclusion of actual rocks and mineral surfaces that perfectly mimic the underground pore system is essential to obtain materials as close as possible to the fluid‒mineral surfaces encountered in the target environments. Recent developments in microfabrication and surface chemistry in microreactors promise to reach this goal. The microporous network can now be reproduced in actual rocks (Hauge et al. 2016, Singh et al. 2017). The surface properties of micromodels can thus be modified and tuned with great flexibility, e.g. by functionalization with molecules (Song and Kovscek 2015) or minerals such as crystallized calcite and others (Yoon et al. 2012) or by directly fabricating micromodels “out-of-the-rock” (Song et al. 2014), which can exhibit pressure resistance up to 100 bar (Porter et al. 2015).

Some examples of current scientific challenges

Microfluidic tools have the potential to revolutionize a wide range of research into microporous ecosystems, whether terrestrial or intraterrestrial. High pressure plays an obvious role in the physiology, morphology, and behavior of prokaryotes (Abe 2007). Thus, an increase in pressure has already been shown to be responsible for the decrease in acetate production to the benefit of formate by Clostridium ljungdahlii (Oswald et al. 2018). HP conditions have repeatedly been shown to produce deformations associated with cell division problems (Welch et al. 1993, Ishii et al. 2004). The use of transparent HP micromodels would enable these adaptations to be observed in real time. The following questions and current challenges are just a few examples of the approaches that could benefit from microfluidic technologies to study life from the molecules to microbial communities and societal applications, although this list is by no means exhaustive.

Pore-scale view of geological CO₂ and H₂ storage. CO₂, as a waste, and H₂, as a strategic energy source, are key molecules to be stored deep underground, benefitting from the large volume offered by such an environment (Haddad et al. 2022). However, both molecules can directly interact with deep environments, either through geochemical or biogeochemical reactions. This means that for such applications, it is important to learn from the benefit of scientific feedback concerning what could be stored at each specific selected geological site and what could be the result of such storage (Ranchou-Peyruse et al. 2019). While local measurements can be made concerning site-specific properties in terms of geophysical and geochemical properties, performing realistic storage tests at the field scale is quite difficult, both for technical and economic reasons. Hence, laboratory-scale approaches could help obtain insights into further industrial strategies. In that context, microfluidic experiments exhibit several interesting specificities, which could be complementary to conventional core flood experiments, such as direct in situ visualization and biogeochemical characterization of the ongoing phenomena with relatively simple laboratory equipment. Some papers concerning geological CO₂ storage (Morais et al. 2016, 2020) and, more recently, H₂ storage (Liu et al. 2023, Lysyy et al. 2023, 2024) have been published, resulting in a pore-scale view and investigations of the geochemical and biochemical effects, particularly the effects of deep underground microorganisms. Similarly, another study reported the use of microfluidic approaches for the simulation of alkaline vents (Weingart et al. 2023). In all these studies, the interest of coupling conventional and microfluidic approaches to investigate energy or waste underground storage is evident since the obtained data can be combined for a multiscale view of the mechanisms. From that point, typical geochemical reactions such as carbonation or serpentinization could be studied on a chip. Indeed, in the presence of water and at specific temperatures, iron-bearing minerals can lead to geochemical reactions along with the production of H₂. Olivine, greigite, magnetite, and awaruite can act as catalysts for the subsequent formation of simple carbon-containing molecules (acetate, formate, pyruvate, and methanol) (Preiner et al. 2018, 2020). These simple molecules can in turn feed entire microbial ecosystems and are at the origin of the subsurface lithoautotrophic microbial ecosystems concept for subsurface lithoautotrophic microbial ecosystems on the basis of H₂ and CO₂ (Stevens and McKinley 1995, Fry et al. 1997, Kotelnikova and Pedersen 1997, Chapelle et al. 2002, Haveman and Pedersen 2002, Takai et al. 2003, Lin et al. 2005, Basso et al. 2009, Crespo-Medina et al. 2014). Moreover, such ecosystems can interact strongly with the injection of gas feedstocks, which are either anthropogenic waste (CO₂) or energetic resources (natural gas, H₂, etc.). Hence, microfluidics approaches can help mimic the geological environment at the laboratory scale to evaluate the impact of such injection on the local biosphere at the pore scale (consumption, reservoir stability, biofilm formation, variation in injectivity, process integrity, etc.).

Microfluidics fast screening approaches for database generation for numerical modeling and scale-up. Another gap is the need to experimentally assess the scaling-up effect between micromodels and larger and more traditional set-ups. In fact, the numerous studies presented here aimed to connect the conclusions from microfluidic studies with other observations in the laboratory or even at the reservoir scale. Modeling can partially address that need but does not satisfy the necessary experimental data and demonstrations. This is where microfluidics approaches can play a critical role. Indeed, one of the major interests with microfluidic experimentation concerns the ability to perform several experiments in a single run. This method has already been applied to crystallization (Bhattacharya et al. 2020, Chauhan et al. 2023), drug screening (Cui and Wang 2019, Sun et al. 2019), organic chemistry, synthetic and structural biology (Hansen and Quake 2003, Kwon et al. 2023) and microorganism cultivation (Watterson et al. 2020, Cario et al. 2022).

In addition, numerous replicates can be realized on the same system, making it possible to envision more robust approaches supported by statistical processing while working under exemplary safety conditions. Hence, microfluidics has the ability to perform fast screening for generating multiple data, which could be implemented in machine-learning approaches to predict biogeochemical reactivity, permeability variation, or biofilm development. All these aspects can further be used in numerical modeling from the pore scale to the reservoir scale, although the scale-up process is still not straightforward. To evaluate separate parameters to assess thermodynamics and kinetics, micromodels (on-chip porous media) can also be used to evaluate the impact of the deep biosphere on geological processes. This can be achieved by designing and microfabricating microreactors that mimic specific storage locations and/or different scenarios. For example, transparent HP micromodel approaches coupled with X-ray imaging characterization, such as X-ray laminography (Morais et al. 2023), would enable a more detailed study of geochemical reactions under conditions simulating deep environments (pressure and temperature), providing additional information to more conventional methodologies based on core flooding coupled with X-ray computed tomography (X-ray CT) (Davit et al. 2011, Minto et al. 2017).

Microbial interactions (microbe–mineral surface and microbe–microbial interactions). Within the deep subsurface, microorganisms interact with their environment from both biological (cell-to-cell) and mineral (attachment and colonization) points of view. Getting access to such interactions at small scales can be performed with microfluidics. A recent study investigated the effects of exposure to short, low-carbon-weight compounds on biofilm formation on mica surfaces (Nuppunen-Puputti et al. 2023). The effects of the conditions tested (addition of methane, methanol, and acetate) on the taxonomic diversity of the communities were then analyzed via high-throughput sequencing approaches performed on both water and rock samples. A complementary culturing approach using a mica-filled micromodel (Bhattacharjee et al. 2022) enabled real-time observation of the formation and/or disintegration of the biofilm formed as a function of the molecules injected to quantify microorganisms or to monitor the behavior of sessile and planktonic microorganisms. Similarly, competition or cooperation phenomena can be studied throughout incubation, depending on the parameters applied (substrate concentrations, flow rates, etc.). For example, Kubik and Holden (2023) studied the competition between different hydrogenotrophs belonging to three different species. The aim of this study was to understand the effects of H₂ concentrations on the metabolism of these microorganisms, particularly to demonstrate that these microorganisms redirected their redox reactions from CH₄ and H₂S production to biomass production at lower H₂ concentrations. Axenic cultures and batch cocultures were generated at various incubation temperatures. Using existing microfluidic tools, it would be possible to create a temperature gradient (Mao et al. 2002) along with cocultivation on a chip to observe the cell distribution and interactions at a small scale (different cells can be easily identified in cocultures when an autofluorescent methanogenic archaeal strain (i.e. cofactor F420) and a bacterial strain are mixed).

Evolution of deep underground microbial communities. Omics approaches have been used to gain a more in-depth understanding of the functions of microbial communities in the deep subsurface but have also led to new questions. For example, in the past, any presence of aerobic microorganisms in anoxic deep continental environments was clearly identified as contamination during sampling, which quickly led to a debate. Recently, omics approaches have revealed the presence of active aerobic organisms (Ruff et al. 2023), even those known only as oxygenic phototrophs (Puente-Sánchez et al. 2018), which has led to hypotheses on the presence of O₂ in these environments (Gutsalo 1971, Winograd and Robertson 1982, Kadnikov et al. 2018). Several biogeochemical explanations have been proposed to explain the presence of this electron acceptor, ranging from radiolysis to the dismutation of various molecules (Coates and Achenbach 2004, Ettwig et al. 2010, Heck et al. 2010). In the case of cyanobacteria found in deep continental environments, H₂ consumption via their hydrogenases was hypothesized. Omic approaches are very powerful and informative, but they do not exempt the microbiologist from testing the hypotheses formulated by cultural approaches that clearly demonstrate hypothetical metabolisms and calculate yields to understand the impact on the functioning of the microbial ecosystem. The rapid development of microfluidic tools for isolating and screening prokaryotes in recent years (Yin et al. 2022, Wan et al. 2023) raises hopes for the transfer of such technologies to the deep biosphere community. Indeed, microbial isolation could be carried out directly on water samples from aquifers or faults and would consume only small volumes (a few microliters) of these often precious and rare samples.

Current limitations of microfluidic approaches for studying the deep underground biosphere

While microfluidics approaches can help address several scientific questions and complement conventional equipment and strategies to investigate the deep underground biosphere, some limitations specific to such methodologies exist. We discuss hereafter some of them along with remediation strategies.

The difficulties in accessing microreactor technologies for new users and accessing adapted HP equipment. Microreactor technology, especially for polymer microfluidics, has become more accessible over 15 years (microfabrication kits are now commercially available), although they do not fully meet deep biosphere research needs. While robust materials such as glass or silicon require expertise, clean rooms are not always necessary—a fume hood suffices. The complexity of photolithography and etching may deter newcomers, but robust commercial microfluidic devices are available, simplifying the process. Several commercial companies now sell robust microfluidics devices in glass or silicon Pyrex to bypass the need to develop complex and expensive in-house microfabrication processes along with all the required equipment to proceed in a user-friendly way with microfluidic studies (Micronit, Little Things Factory, etc.). The overall equipment cost for performing microbial studies in a microreactor is, of course, slightly greater than that of conventional microbial incubation methods (with the exception of pressurized cultivation), but such approaches can (i) yield information that cannot be obtained otherwise (single-cell measurement, bioclogging in pore structures, etc.) and (ii) greatly reduce the experimental time owing to microfluidic fast-screening strategies, thus reducing the overall cost in terms of consumables, e.g.

Microfluidic anoxic experiments. Most of the studies dealing with the deep underground biosphere require anoxic environments to be representative of realistic conditions. As mentioned earlier, conventional polymer microreactors cannot be considered for such conditions given that oxygen can diffuse through these materials. Hence, inorganic microfabrication materials are needed, which increases the microfabrication protocol needed unless these materials are ordered from specialized companies. However, in addition to this limitation, anoxic experiments on a chip are similar to those on larger-scale reactors. This means that a glove or an intergas ramp box is generally needed to ensure the elimination of O₂ when the starting culture medium is prepared.

The implementation of characterization techniques in microreactors. While spectroscopy techniques can be easily implemented on a chip (Raman, FTIR, fluorescence, optical observations, etc.) and are relatively convenient for studying microbiological processes, they depend on the microreactor material, particularly in terms of transparency. Therefore, it is important to pay attention to the choice of construction materials. As mentioned earlier, X-ray analysis and imaging can also be used in association with microreactors, providing ways to characterize fluid flows and geochemistry in porous media. Spectroscopy techniques (Raman, FTIR, fluorescence) can be used directly inside microreactors without disturbing the system. Measurements of gases and metabolites in microscale volumes have already been reported using confocal Raman spectroscopy to quantitatively monitor the solubility of gases such as dissolved CO₂ and CH₄, electrode array detectors, or fluorescence detection (Kraly et al. 2009, Liu et al. 2012). These are powerful approaches for performing direct biochemical imaging (especially when considering porous media), providing important spatial information on microbial activities (Marcsisin et al. 2012, Loutherback et al. 2016, Perro et al. 2016). Additionally, they could also possibly be used for strain identification and quantification on the basis of specific biomolecular fingerprints without staining. However, compared with conventional larger-scale volume strategies, semiquantitative analyses are generally still considered. Hence, both methodologies need to be considered in parallel. In situ sequencing is still a tricky problem with microfluidic methodologies given the small sample volume considered, from picoliter (a drop) to few microliters (a chip). Hence, microreactors need to be coupled with milliliter-scale reactors to guarantee sufficient biomass for further analysis. However, recent trends in single-cell RNA/DNA sequencing (Zhou et al. 2021, Clark et al. 2023) could lead to the development of online microfluidics systems in the near future.

Conclusions and perspectives

While microfluidic tools have revolutionized molecular biology and health care in recent years, the possibilities of applying these approaches to the study of the deep biosphere have barely scratched the surface. However, these tools have long been used in the oil industry and in gas storage in the broadest sense. Despite undeniable advantages over large-scale or core-scale methodologies, the main criticism that can be made to microfluidics is the problem of scaling up the results obtained with a micromodel to model a reservoir as a whole. For deep subsurface microbiologists, however, these reservations are not very important in view of the advantages of such approaches, such as (i) the ability to operate in a large range of experimental conditions (p, T, concentration, etc.), representative of the deep underground environment; (ii) the safety related to microscale experimentation; (iii) the possibility of finely tuning the design of the microreactor from very simple micromodels to the complex permeability and porosity of certain rocks; (iv) the implementation of a wide range of noninvasive in situ characterization techniques to monitor the growth and behavior of microorganisms, the composition of fluids, etc.; (v) the possibility of performing numerous replicates on the same micromodel, making it possible to confirm the results obtained via statistical approaches; and (vi) the ability to perform fast screening through the utilization of on-chip gradients (T, pH, concentration, etc.) for the exploration of multiple conditions in a single run and (vii) the isolation of microorganisms while maintaining pressure conditions and consuming minimal volumes of formation water.

Author contributions

Sandy Morais (Writing – original draft, Writing – review & editing), Emeline Vidal (Writing – original draft, Writing – review & editing), Anaïs Cario (Conceptualization, Resources, Writing - original draft, Writing – review & editing), Samuel Marre (Conceptualization, Resources, Writing – original draft, Writing – review & editing), and Anthony Ranchou-Peyruse (Conceptualization, Resources, Writing – original draft, Writing – review & editing)

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as potential conflicts of interest.

Funding

Funding for this work was provided by the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation program (Grant Agreement No. 725100, project Big Mac). This research also received funding support from the French National Research Agency (ANR) with the projects ODEVIE (ANR-21-CE01-0018) and HOT DOG (ANR-22-CE02-0017). Additional funding was obtained from the French National Research Agency under France 2030 (project MICROFLUIDICS) bearing the reference ANR 22 EXOR 0013.

References