Raman microfluidic cell sorting technology controls the flow of cells within the flow channel, and the cells remain active in the microfluidic environment. As these cells traverse the Raman excitation spot, Raman signals are generated, facilitating sample dichotomies based on the signal (Wang and Yu, 2015; McIlvenna et al., 2016). Specifically, this entails determining whether the cells belong to the target cells. As shown in Figure 3A, McIlvenna et al. developed an automated RACS system that achieves reliable synchronization of continuous Raman signal acquisition, real-time identification, and cell sorting with a sorting accuracy of 96.3%. Central to the implementation of continuous flow sorting is the use of “pressure dividers” to eliminate flow fluctuations in the inspection area. During the sorting process, Raman spectra are continuously collected. If the spectrum meets the sorting criteria, the program triggers the pump to switch output pressures and direct the flow into the collection channel. The collection channel then opens within a preset delay time before the flow is directed back to the waste outlet. Utilizing a simple hydrodynamic focusing and switching mechanism, this system eliminates the dependence on cell size, medium conductivity, and refractive index in the “capture” approach and provides a significant advantage in separating cells from complex populations (McIlvenna et al., 2016).

The optical trap generated by optical tweezers plays a dual role in manipulating target cells within the microfluidic channel. First, it restricts the flow of target cells in the channel, facilitating prolonged spectral acquisition for improved spectra. Second, it allows for the manipulation of cell movement to a specified area post-spectral acquisition. Leveraging these features, Raman optical tweezers are effective in manipulating and isolating a single cell. However, it is crucial to note that the process of isolating a single cell is very time-consuming. Previous studies have reported obtaining only a single cell using this technique within a 15-min duration (Huang et al., 2009). Subsequent research endeavors have aimed at improving the throughput of this technique. Notably, specific model microbial cells have achieved sorting throughputs ranging from 3.3 to 8.3 cells/min. However, it is only for one type of cell and does not result in only one single cell (Lee et al., 2019).

The integration of droplet microfluidics with Raman cell sorting, using mutually incompatible two-phase flow to envelop cells and thereby preventing cross-contamination, represents the most efficient system for single-cell identification, sorting, and culture at present (Huys and Raes, 2018). In the RADS system, dielectrophoresis (DEP)-based droplet sorting was applied to isolate those containing target droplet cells. Notably, SCRS acquisition and DEP-based droplet sorting are controlled by multi-threading, i.e., SCRS acquisition is performed simultaneously during droplet sorting. This design further improves sorting efficiency. RADS significantly enhances the sorting throughput to over 100 cells/min (Wang et al., 2017). However, it is specific to a certain pattern of cells characterized by strong Raman signals and is not universally applicable. For instance, non-labeled Escherichia coli necessitates 3–5 s for the excitation of a more desirable spectrum.

The integration of laser-induced forward transfer technology with Raman sorting technology allows for the analysis of single-cell Raman spectra, followed by the use of laser technology for the ejection and collection of individual cells. This method measures and receives Raman spectra from single cells and enables the ejection sorting of target cells. Through subsequent single-cell whole-genome sequencing, it becomes possible to construct the relationship between single-cell genotypes and phenotypes. However, the use of laser technology poses a challenge as it can damage the cells, resulting in low genome coverage when dealing with a single cell. This technique proves highly suitable for microbial samples in complex environments like soil and sludge because there are no problems like clogging of microfluidic tubes. Additionally, it is well-suited for low-abundance samples. Despite its advantages, realizing the culture of the received single cells remains a challenging aspect of this approach (Song et al., 2017; Jing et al., 2018; Wang et al., 2020).

This paper considers factors such as cell damage probability, the need for labeling, and single-cell sorting with high detection accuracy. Considering these factors, this paper compares the traditional FACS technique with several of the most representative Raman cell sorting techniques, as shown in Table 1.

Metagenomic sequencing and sorting are increasingly powerful tools that enable direct analysis of microbiota at the genotype level (Albertsen et al., 2013; Nielsen et al., 2014; Jing et al., 2018). However, phenotypic information on microorganisms is inevitably missed. Raman spectroscopy has proven to be a useful method to provide comprehensive phenotypic information on single cells in a non-destructive manner. Metagenomics is a culture-independent approach to microbial research. It provides a comprehensive understanding of the diversity and potential functions of microorganisms in a sample. Single-cell genomics serves as a complementary strategy to metagenomics, enabling the establishment of genetic links among DNA sequences within single cells (Swan et al., 2011; Lasken, 2012). Raman cell sorting, combined with single-cell genomics, represents a potent approach for circumventing microbial culture and revealing the microbial “dark matter.” Although fluorescence-activated cell sorting has successfully sorted single cells for subsequent single-cell genomics studies (Rinke et al., 2013), Raman cell sorting presents a valuable alternative. This method sorts cells based on SCRS, where the spectra reflect the cell’s phenotype and label-free biochemical “fingerprint” information (Huang et al., 2015). Such information can link sorted cells to their phenotypic traits and ecological functions (Song et al., 2017).

The potential of Raman cell sorting combined with single-cell genomics has been demonstrated. Song et al. used a novel RACE method to sort single bacterial cells from Red Sea water samples based on SCRS. They showed the isolation of single cells containing carotenoids from Red Sea samples based on SCRS characterization. RACE-based single-cell genomics revealed putative new functional genes associated with carotenoid and isoprenoid biosynthesis, as well as previously unknown phototrophic microorganisms, including non-culturable Cyanobacteria spp. (Song et al., 2017). Jing et al. utilized SCRS as a biochemical map, and RACE was able to link cellular phenotypes to genotypes. Mini-metagenomic sequences from RACE can be used as a reference to de-construct a near-complete genome of key functional bacteria by binning shotgun metagenomic sequencing data. This novel approach reveals the role of marine “microbial dark matter” in global carbon cycling by linking uncultured Synechococcus spp. and Pelagibacter spp. to carbon fixation and in situ flux activities (Jing et al., 2018).

Raman cell sorting has been applied to single-cell genome sequencing. Li et al. developed a new method combining RACS, stable isotope probing (SIP), and genome-directed cultivation (GDC). For the identification, sorting, and culture of active toluene degraders from complex microbial communities in petroleum-contaminated soil, the single cells therein were sorted and isolated using RACS. They successfully assembled the genome of Pigmentiphaga based on the metagenomic sequencing of 13C-DNA and genomic sequencing of sorted cells. Additionally, the genotypes and phenotypes of this degrader were directly correlated at the single-cell level (Li et al., 2023). Li et al. used 13C-labeled phenanthrene as the target and developed a new method coupling MMI−SIP and RACS. This method is designed to identify active bacterial cells that degrade phenanthrene from wastewater contaminated with polycyclic aromatic hydrocarbons (PAHs). Moreover, this approach significantly enriched active phenanthrene degraders and successfully isolated representative single cells. Novosphingobium was confirmed to be the active phenanthrene degrader by amplicon sequencing analysis by SIP, the 13C shift of the single cell in Raman spectra, and the 16S rRNA gene from single-cell sequencing by RACS (Li et al., 2022).

Optimized Raman microscopy can overcome the weak signal of spontaneous Raman scattering and significantly reduce the acquisition time of spontaneous SCRS. The CRS technique, a nonlinear Raman approach, involves spatially overlapping two light beams with frequencies w_p and w_s, which are then converged on the sample through the objective lens. As a nonlinear optical process, it encompasses both coherent anti-Stokes Raman scattering (CARS) and SRS. These techniques arise from the nonlinear coherent Raman phenomenon, occurring when the frequency difference between the pump beam (w_p) and the Stokes beam (w_s) matches the Raman-active vibrational modes. They can achieve Raman signal enhancement in the 3rd–5th order (Folick et al., 2011; Zhang et al., 2014; Zhang et al., 2015c). Specifically, the CARS technique increases the velocity through a coherent radiation field, which makes the signal increased and highly directional (Evans et al., 2005; Cheng and Xie, 2015), and the SRS technique amplifies the signal through the presence of a local oscillator at the detection wavelength (Saar et al., 2010; Cheng and Xie, 2015).

Nevertheless, the application of both SRS and CAR entails the use of femtosecond lasers. Although this method drastically increases the number of photons interacting with a cell in a short duration, resulting in Raman signal amplification, it concurrently raises the likelihood of photodamage to the cell. This increased risk complicates the execution of experiments on living cells (Zhang et al., 2017; Suzuki et al., 2019). Compared to spontaneous Raman spectroscopy, which encompasses a wider range of vibrational spectra, the CRS technique covers a relatively small window of molecular vibrations (Song et al., 2016). The main limitation faced by the CRS technique is its narrow range of wave numbers.

The signal of Raman spectroscopy can be enhanced using SERS technology. Using surface plasmon resonance generated by nanoscale metal structures, such as metal nanoparticles, designing metal coatings with microstructures and nanostructures inside micro-orifices, the Raman spectral signal can be enhanced by 10⁶–10¹⁴ orders of magnitude, and a new Raman spectrum can be obtained (Chauvet et al., 2017). However, it is important to note that SERS is confined to a short range, which only has an extremely strong signal enhancement effect on chemicals within tens of nanometers close to the metal surface. In contrast, there is no significant enhancement effect on chemicals located far away from the metal surface, potentially introducing signal acquisition bias and a loss of representativeness of the Raman signal.

Research on the production of unlabeled compounds featuring strong Raman signals, known as biomarkers, is widespread. This method involves utilizing cell-intrinsic substances that are sensitive to Raman excitation and exhibit significant enhancement in a specific region of the Raman spectrum, aiming to achieve single-cell recognition and differentiation. Carotenoids (Jehlička et al., 2014; Song et al., 2017) and cytochrome c (Pätzold et al., 2008; Lee et al., 2019) are currently commonly used biomarkers. However, it is crucial to recognize that this technique can only work for specific organisms.

Due to its high sensitivity to minute alterations in the vibrational frequencies of chemical bonds, Raman spectroscopy can be effectively coupled with stable isotopes (SIPs). The combination of SIPs and Raman spectroscopy is sufficiently sensitive to allow for the semi-quantitative detection of ¹³C doping at the single-cell level (Li et al., 2012b).

Stable isotope labeling (SIL) involves introducing isotope-labeled nutrients into the cell culture medium. After the cells metabolize these nutrients, the molecular structure remains stable, but the corresponding Raman spectra produce a “red shift phenomenon.” This phenomenon serves not only as an indicator of functional microorganisms but also as an indicator of microbial activity (Li et al., 2019). When combined with sequencing technology, SIL enables an in-depth analysis of the metabolism and functions of single cells within complex samples. The isotopes commonly used for single-cell Raman spectroscopic characterization include ¹³C, ¹⁵N, and ²H (Wang et al., 2016).

Investigations using isotope-labeled substrates for the functional analysis of microbial communities have two significant limitations. Initially, isotope-labeled derivatives of numerous substrates of interest are prohibitively expensive and often not commercially accessible. Second, the introduction of labeled substrates modifies the composition of the natural substrate pool, potentially leading to alterations in the activities of community members (Wang et al., 2016).

Laser tweezer Raman spectroscopy (LTRS) stands out as a powerful, label-free technique for manipulating and analyzing individual cells or particles in aqueous solutions. As shown in Figure 7A, Lau et al. developed an optical tweezer-based Raman cell sorting system in a flowing state, where the laser serves as an excitation source, a trapping force, and a sorting switch concurrently. This system captures cells into the Raman acquisition region, effectively extending the Raman acquisition time and prolonging the total exposure time of the cells. Subsequently, optical tweezers are used to guide cells of interest into the collection channel (Lau et al., 2008).

Despite the slow capture speed and low throughput of the optical capture method known as optical tweezers, optical tweezers have been a method used to hold the position of cells during Raman detection (Xie et al., 2005b; McIlvenna et al., 2016). There is a high risk of photodamage to cells when using laser tweezers (Pilat et al., 2013). Currently, there are faster cell-trapping methods, such as DEP, which operates through electrical means. DEP can exert forces in the range of 0.1 nN–1 nN for particle manipulation, making it robust enough to capture single cells at high flow rates. Combining DEP with Raman spectroscopy facilitates label-free manipulation and cell identification, thereby contributing to an improvement in sorting throughput (Zhang et al., 2015b; Li and Anand, 2018). However, this method is based on low-conductivity, non-physiological buffers, the use of which may affect cell viability (Schroder et al., 2013).

As shown in Figure 7B, Zhang et al. developed a pioneering Raman cell sorting platform that merges positive dielectrophoresis (pDEP) with RACS. By periodically applying and turning off the pDEP field on the electrode array, the platform achieves precise capture, sorting, and delivery of single cells to the Raman detection region. Raman spectra are subsequently measured one by one. In addition, the capture time of this platform is adjustable. As a proof of concept, the platform successfully sorted carotenoid-producing and non-carotenoid-producing yeast cells. The characteristic peaks in the Raman spectra of carotenoid-producing yeast cells were markedly distinct from those of their non-carotenoid-producing counterparts, resulting in an 8-fold enrichment. This outcome serves as further evidence that the method effectively addresses the issue of weak spontaneous Raman signals. Additionally, the method exhibited notable high-throughput capabilities, capturing and characterizing at least 577 cells within a 540-s timeframe (Zhang et al., 2015b; Li and Anand, 2018).

In the absence of “cell capture,” achieving precise focusing of flowing cells onto the excitation spot becomes challenging, especially in wide channels. Hydrodynamic focusing offers a solution that is independent of the physical properties of the cells and the medium, making it universally applicable to a broad spectrum of biological systems. As shown in Figure 7C, Lyu et al. developed a 3D hydrodynamically focused microfluidic system for fully automated, continuously Raman-activated cell sorting (3D-RACS). This system incorporates a 3D-printed detection chamber (1 mm³) integrating a PDMS-based sorting unit, optical sensors, and an in-line collection module. It is able to precisely localize cells in the detection chamber and perform Raman measurements, effectively eliminating spectral interferences from the device material. As a proof-of-concept, Raman-activated sorting of a mixture of Chlorella vulgaris and E. coli was performed with a purity of 92.0% and a throughput of 310 cells/min. This capture-free system exhibits significant potential for high throughput (Lyu et al., 2020).

As shown in Figure 7D, Hiramatsu et al. used an acoustofluidic focusing microfluidic chip to tightly focus each cell in a high-speed flow, ensuring stable and efficient acquisition of Raman spectra. An acoustic standing wave, generated by a piezoelectric transducer (PZT), plays a crucial role in achieving accurate and consistent measurements in the context of a high-speed flow. This standing wave focuses cells onto a standing wave node positioned at the center of the microchannel. In this manner, all cells are tightly focused in a single stream, facilitating optical interrogation at the center of the microchannel (Hiramatsu et al., 2019).

Capturing and localizing cells precisely within the Raman detection area facilitates prolonged signal acquisition. This capability extends the total exposure time of the cells, consequently enhancing the signal-to-noise ratio of the Raman spectra. Historically, many microfluidic systems have predominantly used optical or electrical techniques, such as optical tweezers and dielectrophoresis (Zhang et al., 2015c), for accomplishing “cell capture.” Typically, Raman cell sorting techniques based on cell capture and localization methods are very common. However, the capture rate used by the above methods may have imposed a limitation on the detection throughput. In addition, capture efficiency depends on variables such as cell size, medium conductivity, and refractive index (Lyu et al., 2020).

Some methods can similarly focus cells precisely at the excitation spot position, allowing the cells to receive sufficient total cell exposure time in the Raman detection region, enhancing the spontaneous Raman signal, and improving the signal-to-noise ratio of the spectrum. Hydrodynamic focusing and acoustic focusing are common ways of precisely focusing cells in the flow state at the excitation spot position and detecting Raman spectra (flow Raman). Moreover, the integration of acoustic focusing and flow focusing is a common practice, further enhancing the precision of cell localization in the Raman detection region (Nitta et al., 2020).

Meanwhile, optical tweezers or dielectrophoresis methods often utilize hydrodynamic focusing or acoustic focusing to pre-focus flowing cells before cell capture and localization in order to capture and localize the cells in the Raman detection region more accurately, for example, by pre-focusing cells to generate a narrow stream before their capture (utilizing optical tweezers) (Lau et al., 2008). Hydrodynamic focusing typically yields elevated cell velocities, diminishing the efficiency of cell capture. In contrast, acoustofluidic focusing focuses cells at velocities conducive to effective capture (Wang et al., 2023).