1. Introduction

Raman spectroscopy refers to scattered light from a sample that exhibits a frequency shift reflecting the energy of specific molecular vibrations within the sample of interest. In this manner, it provides a detailed chemical composition of the sample-a chemical fingerprint in essence. The technique has wide potential in the biomedical sciences as it may be applied to samples over a wide size range from single cells through to intact tissue. One of the major challenges of Raman spectroscopy is the extremely low intensity of the emitted signal. This, in addition to the fact that a Raman signal may be obtained from the local environment surrounding the sample, typically makes it difficult to discern the molecular signatures of interest. Thus, considerable effort has focussed on enhancing the ratio of signal to background noise. The Raman signal can be enhanced by increasing the time of acquisition (~minutes). However, in the case of live cells, long acquisition times can cause damage due to extended irradiation by the laser, exciting the Raman transitions, and also limit the number of cells that can be conveniently analysed.

The ability to hold the cell away from any surface without damage is desirable, to avoid any associated interference from it, and hence reduce integration times. Optical tweezers [1] uses the gradient force for trapping objects at mesoscopic size scales using only a few mW of trapping power. Researchers have combined optical tweezers and Raman spectroscopy (Raman Micro Spectroscopy) [2, 3]. These two techniques are compatible as the highly focussed trapping beam provides the necessary power for Raman excitation whilst simultaneously providing trapping of a single microparticle. The backscattered Raman signal may also be collected by the same high numerical aperture (NA) objective used for trapping. Other geometries have recently used separate beams for the trapping and Raman excitation [4, 5]. However, it is known that optical tweezers use a tightly focused light beam and, though powerful, such a system may result in undesirable cell damage (e.g. due to two-photon absorption) [6]. Additionally, single beam optical tweezers are not able to hold and move large cells at will, rather they may cause the cell to pivot or orient around the trap focal point or align to minimise energy in the applied laser field. Holographic optical tweezers may provide assistance to overcome these alignment issues [7].

In this paper, we demonstrate the use of a dual beam fibre optic trap to hold and manoeuvre large cells for the recording of Raman spectra. The novel combination that we realise allows us to record Raman data from large cell types in a trap and particularly to acquire Raman signals from the local parts of the cell in the trap. Additionally, we realise a microfluidic flow system where cells may pass through a dual beam trapping region and Raman acquisition may be performed.

Prior to the realisation of the single beam optical trap (optical tweezers) [1], the first actual trap demonstrated by Ashkin in 1971 employed two divergent counter-propagating beams [5]. This trap was revisited in 1993 by Constable and co-workers [8] who generated a version using two co-axial optical fibres. This dual beam trap has many advantages over the single beam optical tweezers that are important for the purposes described here. Firstly the trap does not require high numerical aperture optics and may be realised with weakly focused or divergent beams. In turn this indicates the prospect of reduced photodamage or any two-photon induced damage [6]. The dual beam fibre trap also encompasses another key attribute: it can hold large objects readily due to the large encatchment area and beam divergence. Finally due to its geometry it may be readily incorporated into a microfluidic flow system for cell analysis. Indeed recent studies have shown the use of a cell stretcher where a dual beam system may be used to elongate cells in a manner dictated by their internal elasticity: in turn this allows one to distinguish between normal and tumour cells [9].

2. Experimental section

A dual beam fibre optical trap was used for all experimental studies and placed above a Raman micro spectroscopy experimental arrangement. A schematic of the experimental arrangement can be seen in Fig. 1. An ytterbium fibre laser operating at 1070nm (IPG Photonics VLM-5-1070-LP, linearly polarized, coherence length~1mm) was used to operate the fiber trap. The light was coupled into two optical fibers via a λ/2 plate and a polarizing beam splitter. Separately both pairs of singlemode and multimode fibres were used in the experiments. The optical power emerging from each fiber could be adjusted with the λ/2 plate to ensure equal field intensity distribution. By carefully choosing the optical path difference in each fibre arm to be in excess of the laser coherence length, standing wave effects were avoided. One fiber (F1) was mounted on a cover slip which was in a fixed position above the Raman tweezers spectroscopy setup, the second fiber (F2) was mounted on a XYZ stage and could be aligned with F1 similar to a fiber to fiber pig tailing setup.

A Raman system, based around a Nikon TE 2000-U microscope, is introduced from below enabling spectroscopic studies of samples held in the fibre optical light force trap. Briefly, a temperature stabilised diode laser operating at 785nm (a circularised laser diode, Sanyo DL-7140-201s up to 80mW power) is expanded and introduced, via a holographic notch filter (HNF, Tydex notch-4), into an inverted optical tweezers geometry and passed to the sample via a ×100 NA 1.25 oil immersion objective or a ×50 NA 0.9 oil immersion objective. The backscattered Raman light is collected by the same objective and passed through the HNF. The Raman signal is then imaged onto a confocal aperture and passed through a short pass filter, with a cut off at 1060nm, in order to suppress any scattered radiation from the fibre trap laser. Finally, the Raman signal is reflected by the dichroic mirror (DM) and imaged, using a lens (f=80mm), onto the spectrograph (Triax 550 Jobin Yvon). This spectrograph is equipped with a CCD camera (Symphony OE STE Jobin Yvon) for detection of the Raman spectrum. This system results in a spectral resolution of approximately 2cm^-1. For the cellular based studies primary human keratinocyte (PHK), Human promyelocytic leukemia (HL60) cells and Human cervical squamous carcinoma (SiHa) cells were used and were prepared as follows: Primary (normal) human keratinocytes (PHKs) from neonatal foreskin (Cambrex) were cultured in keratinocyte growth medium-2 (KGM-2) containing the supplied supplements (PromoCell) as described previously [10]. The HL60 human promyelocytic leukemia cell line was grown in RPMI 1640 medium containing 10% fetal calf serum (FCS) [11]. The SiHa cells are a cell line derived from carcinoma of uterus, I [12] and grown in Dulbecco’s Modified Eagle’s Medium containing 10% fetal calf serum (FCS) and 1% Penicillin, Streptomycin and Glutamine (PSG).

 

Fig. 1. Schematic of the Experimental arrangement. PBSC: Polarising Beam Splitting Cube; HWP: Half Wave Plate; F1, F2: Fibre one and two; HNF: Holographic Notch Filter; DM: Dichroic Mirror; CA: Confocal Aperture and BEF: Band Edge Filter.

Download Full Size | PDF

3. Experimental results

The use of a fibre optical light force trap opens up the possibility of stably trapping large objects with reduced power densities and without the need for holographic elements or multiple trap sites. As an example, Fig. 2 shows a fibre optical light force trap, with multimode fibres, being employed to trap a 100 micron object using 800 mW of light travelling in each fibre, resulting in an output irradiance of 5.2×10⁸ W/m².

 

Fig. 2. A 100 micron polymer sphere trapped in a fibre optical light force trap, viewed from below. The fibre trap uses 62.5/125µm (core size/cladding size) multimode fibre, a trapping power of 800mW in each fibre arm and a fibre separation of 240µm.

Download Full Size | PDF

Additionally, we retain the advantages that simple Raman tweezers provides for smaller samples such as the ability to move particles away from interfering surfaces such as coverslips. Notably, with the use of the fibre trap as the primary trapping mechanism, we now also have local control over where in the particle we collect data for the Raman spectra. To demonstrate the ability and usefulness of this technique a 50 micron polystyrene sphere was trapped in the fibre optical light force trap, using a power setting of 40 mW emanating from each singlemode fibre (irradiance 4.6×10⁹W/m²) and a fibre separation of 175µm. Using the mechanical motion of the sample stage, upon which the fibres are placed, the particle was then scanned across the laser exciting the Raman transitions. The particle was scanned perpendicularly to the fibre trap trapping axis, in the y direction as shown in Fig. 1, in 5 micron intervals measured with the use of a calibration grid. The intensity of the 1000 cm^-1 benzene ring breathing mode in polymer [13] was monitored as the particle was scanned to demonstrate the system’s ability to measure the depth of the particle at each point. The intensity of the laser, exciting the Raman transitions, was set to 30 mW (irradiance of 4x10¹⁰ W/m²) and focused onto the sample with a Nikon x50 N.A.=0.9 oil immersion objective, the Raman signal was then integrated for two seconds. The results can be seen in Fig. 3.

 

Fig. 3. Left Chart showing the variation in Raman intensity, of the 1000cm^-1 benzene ring breathing mode of polystyrene, as the microsphere is scanned across the Raman excitation laser. Right hand chart shows a fit of the data obeying a Lorentian relationship.

Download Full Size | PDF

From Fig. 3 we see that as the depth of the particle decreases, due to its spherical nature, the Raman intensity falls off proportionately with the depth as we move from the centre of the sphere, showing that this method allows us to gain sensitive localized information from large objects. We also observe a small unexpected dip in intensity at 45µm into the scan, we believe this is due to a defect in this microsphere and serves to further show the sensitivity of this technique. Although not demonstrated here the use of a tight confocal arrangement and movement of the sample stage in the z axis would permit scanning in three dimensions. To further supplement this data we used the diode laser, exciting the Raman transitions, to trap and load the fibre trap with particles of different materials. Polymer particles (5µm) and ethylene dimethacrylate (EDMA) particles (7µm) were loaded into the fibre trap creating a colloidal conglomerate of these objects. Once in the fibre trap, the objects may be moved by varying the intensity of the optical field from each fiber permiting us to record the local Raman signal from this sphere amalgamation giving chemical information from the trapped particles. The trapping and Raman excitation powers as well as the fibre type used are the same as those described for the experiment with the 50 micron sphere and the fibre face separation was 65µm. The results can be seen in Fig. 4.

 

Fig. 4. A conglomerate of spheres in the fibre trap created from two 7µm EDMA spheres and two 5µm polymer spheres. They are shown with their measured Raman spectra demonstrating the ability of this technique to gain localized sensitive chemical information.

Download Full Size | PDF

Figure 4 shows that the fibre trap allows us to not only store multiple particles, but also to perform localized Raman spectroscopy and distinguish easily different types of particle purely by adjusting the laser power or by moving the fibre positions, relative to the Raman excitation laser beam, via mechanical movement of the sample stage.

We have also recorded local Raman spectra from single large isolated cells in this manner, as shown in Fig. 5. We trapped a primary human keratinocyte (PHK) cell and were able to move the cell around in the trap by varying the laser power in the fibres and/or the fibre positions. PHKs are normal cells that form the lining of the skin. They are approximately 30µm in diameter and have relatively small nuclei, with abundant cytoplasm, giving them a low nucleo-cytoplasmic ratio. This presents a significant challenge to trapping technologies and demonstrates clearly the utility of the approach described here. In order to confirm this we attempted to tweeze HL60 cells and SiHa cells with a single beam trap formed with a Nikon ×100 NA 1.25 oil immersion objective and a power of 80mW (irradiance of 1×10¹¹ W/m²) at 1070nm, a comparable power to that utilized by the fibre trap. We found that with this large power of 80mW it was possible to tweeze the smaller HL60 cells, approximately 8 microns in diameter; however there was little control over the trapping position and such high power increases the risk of two photon damage cellular damage [6]. We found that the orientation of the cells in the single beam trap was not stable. Two trapping sites were observed on the cell, the nucleus and the membrane; this is most likely due to the non uniform cell density. Over periods of a few seconds the ‘stable’ trap site would often oscillate between the membrane and the nucleus. We then tried to tweeze the large SiHa cells and found it was only possible to tweeze the cells in the x,y plane for a time before they attached to the glass slide. Although we were able to tweeze these cells to an extent, the high power density at the focus of the single beam tweezer poses a serious risk of photodamage. We note that holographic optical tweezers may be used to hold large cells, by the use of multiple trap sites, for Raman examination [7]. However both a single beam trap and multiple high N.A. trap sites created by holographic optical tweezers may initiate trapping of individual intra cellular components [14] that could affect the cell cycle dynamics as the Raman spectrum is acquired. The dual beam trap due to its low power density and optical distribution (divergent trapping beams) does not result in such effects on the cell and thus potentially offers a significant advantage as well as experimental simplicity. Using the dual beam trap we can ensure trapping stability over long periods of time and keep the cells away from any surfaces denying the cell any opportunity to form attachments thus making long term studies on a single cell viable [15].

For the localized Raman experiment, 40mW of power from each singlemode fibre (irradiance of 4.6x10⁹ W/m²) with a fibre face separation of 85µm was used to trap the cell. A 20 mW laser beam, for Raman excitation (irradiance of 2.5x10¹⁰), was passed through a Nikon x100 NA=1.25 EPLAN oil immersion objective to the sample. The higher power objective was selected in this case to give a tightly focused beam hence a small examination area. In order to avoid interference from surrounding areas a 100 micron diameter pinhole was placed in the image plane to form a confocal arrangement that corresponds to a 1 micron aperture in the sample plane. The signal was then integrated for two minutes. By using this method we recorded separate spectra from the membrane, cytoplasm and the nucleus. Five spectra were taken at each point, summed together and smoothed, using the method of adjacent averaging, to produce the data shown in Fig. 5. Accompanying Fig. 5 we also present a tentative band assignment [16, 17, 18] as shown in Fig. 6.

 

Fig. 5. Raman Spectra obtained from 3 different positions within a PHK cell: Nucleus (A), Cytoplasm (B) and Membrane (C). The actual laser position during the excitation is also shown by the letters in the top left diagram. The top right hand diagram shows spectra taken from the nucleus of a cell in the fibre optical trap (1) and absorbed onto a glass coverslip (2). The use of the fibre optical trap reduces the background allowing us to discern the Raman features in more detail.

Download Full Size | PDF

 

Fig. 6. Band Assignment for Raman Spectra of PHK. Abbreviations:P-Protein, Tyr-Tyrosine, T-Thymine, G-Guanine, A-Adenine, C-Cytosine, bk-DNA sugar-phosphate Back Bone, Trp-Tryptophan, CY-Cytoplasm, NC-Nucleus, M-Membrane.

Download Full Size | PDF

Observing the spectra in Fig. 5 it is immediately noticeable that the three spectra are visually quite different. If we first consider the spectra pertaining to the nucleus and cytoplasm we can see that strong peaks corresponding to DNA sugar-phosphate backbone and bases A,T,G and C (785,830,895,1048,1093) are markedly reduced in the cytoplasm spectra. This is to be expected as the nucleus contains DNA in highest densities. However these peaks do remain visible in the cytoplasm spectra as the bases A, C and G but excluding thymine which form DNA also form RNA which we would expect to be present in significant quantities in the cytoplasm. Perhaps one of the most interesting aspects of this comparison is the study of the peaks attributed solely to Thymine and Deoxyribose. Thymine is a DNA base that is replaced by Uracil when RNA is formed and transferred to the cytoplasm, in a similar manner Deoxyribose has a hydroxyl group added to it to form the sugar backbone section in RNA. If we examine the peak at 751cm^-1, attributed solely to thymine, we see that it almost completely disappears in the cytoplasm spectra when compared with that of the nucleus; it is likely that the remaining small peak, in the cytoplasm spectra, is due to the small amount of DNA present in the cytoplasm. Similarly the two peaks attributed solely to Deoxyribose, at 895 and 1048 cm^-1, are affected in the same manner with comparatively large depletions in their intensities. It is also interesting to note that, as expected, peaks pertaining to proteins and amino acids (643, 852, 939 and 1660^-1) remain prevalent in the cytoplasm spectra and, in the case of the 1609 cm^-1 peak, display a small enhancement. Finally, in this comparison we also may not have expected the large Lipid peak at 1451 cm^-1 in the cytoplasm spectra. However the geometry of the Raman system means that in order to access the cytoplasm region, despite the small confocal aperture of 1µm in the examination plane, the laser will also pass through the membrane which will contribute to the total spectra collected. It is also possible that the laser beam, used for Raman excitation, may act as a single beam optical trap opening up the possibility of trapping vesicles in the probe during the excitation and collection of the Raman spectra [14]. If we look at the spectra for the membrane we can again see that, in comparison to the other two spectra, the Raman peaks are much reduced. However, we may have expected that the spectra would only contain the large Lipid peak at 1451 cm^-1 along with a few peaks relevant to proteins and amino acids that may be eimbedded in the membrane. Although this is generally the case we cannot fail to notice contributions to the spectra from peaks pertaining to DNA/RNA. This is likely to be due to interference from the surrounding cytoplasm.

We can quantify the differences in these spectra statistically with the use of Principal Component Analysis (PCA). This is a statistical technique that uses multivariant analysis to discriminate between data sets, by examining the spectral components with maximum variation [19]. These variations between the spectra are known as principle components and by comparing the original spectra with the principle components we may separate the data sets. Five spectra were taken at each location shown above and were then subjected to PCA. The results can be seen in Fig. 7. By using only two principal components we are able to show a large discrimination between the spectra taken at the three separate locations. Previous studies investigating localized Raman signals within cellular structures have required the cell to be fixed or adhered to a surface [20] or the use of exceptionally low Raman excitation powers to avoid interference with the trapping mechanism [7]. These results, however, demonstrate the ability of this system to obtain localized sensitive chemical information from cells stably trapped, in the fibre trap, away from interfering surfaces.

 

Fig. 7. Chart displaying the results of the principle component analysis for 5 spectra taken at each of the 3 different locations within the cell. This demonstrates further how we may gain an insight into the intra cellular makeup.

Download Full Size | PDF

4. Microfluidic Raman acquisition

One area where the applications of the fibre optical light force traps show great promise is that of microfluidics. The non contact large trapping area characteristic of fibre traps coupled with their ability to easily trap large biological cells makes them ideal for integration with microfluidic flow channels. Notably this has been realised using the cell stretcher where cell elasticity has been exploited to distinguish normal cells from malignant cells [9]. The incorporation of Raman Micro spectroscopy into microfluidic systems has been previously demonstrated but limited to a small cell held in a trap and monitored whilst fluid was flowed through the system [21], our system differs as we allow the cells to flow through the system. The inclusion of the combination of Raman Micro spectroscopy and fibre optical trapping into the microfluidic environment may greatly expand the potential of this technique allowing the trapping, analysis and release of many cells.

To demonstrate the potential of this technique for incorporation with microfluidics, we developed a microfludic flow system using the dual beam fiber trap to hold particles and cells with an orthogonally placed objective to excite and then record Raman spectra. The flow system consisted of a square capillary with an inner diameter of 80µm between two reservoirs resulting in a flow driven by simple capillary action. Initially, a solution of 10µm polymer particles were flowed through the capillary tube at a volume flow rate of 1.2 nl/s. Against the outside wall of the capillary tube, the two single mode fibers were placed opposing each other, resulting in a fibre face separation of 160µm, as can be seen in Fig. 8.The power from the fibers was set to 80mW emanating from each fibre (irradiance of 4.6×10⁹W/m²), the increase is due to the capillary walls scattering some of the incident light and to hold the particles stably against the flow. A 50mW Raman examination beam (irradiance of 6.3×10¹⁰W/m²) was then introduced from below using a Nikon x50 NA=0.9 oil immersion with an excitation time of 5 seconds. The spectra obtained from a polymer microsphere trapped inside the capillary can also be seen in Fig. 8. We were additionally able to implement cell flow through the system at a volume flow rate of 40 pl/s. Again, a Raman beam was introduced, as for the polymer, and an acquisition time of 60 seconds was used for cell studies. Despite the higher laser power from the fibre trap the cells did not exhibit any damage during the trapping. The spectra from HL60 cells, a human promyelocytic leukaemia cell line, are shown in Fig 8. The Raman peaks can be assigned according to Fig. 6. Notably, the acquisition time was 60 seconds and with possible use of surface enhanced Raman spectroscopy techniques [22] or resonance Raman spectroscopy techniques [23] we may reduce the acquisition time and thus truly turn this into a higher throughput technology opening up the possibilities of integrating this technique with established optical technologies such as optical sorting [24]. This result confirms both the practicality and potential for dual beam fibre traps coupled with Raman spectroscopy for use in microfluidic and lab-on-a-chip environments.

 

Fig. 8. Fibre optical trapping and Raman examination of a 10 micron sphere and a HL60 cell in a microfluidic flow constructed from a square capillary tube, the horizontal object in the figure, with two fibers placed orthogonally against the capillary walls forming a dual beam trap with the laser, exciting the Raman transitions, introduced from below the trap.

Download Full Size | PDF

5. Conclusions

We have demonstrated the use of a dual beam fibre trap combined with Raman micro-spectroscopy. In contrast to standard optical tweezers, the dual beam trap can hold and manoeuvre large cells with low probabilities of photodamage and we can record local Raman signals from a large trapped cell with ease. We have shown localised Raman signals from PHK cells, a previously unstudied cell type, and performed detailed statistical analysis. Finally, we have shown an implementation of a Raman micro-spectroscopy system capable of handling large cells with a microfluidic flow and have recorded spectra from polymer microspheres and HL60 cells.