Light scattering is an essential parameter in flow cytometry, facilitating functions such as size measurement, discrimination of cell types on the basis of shape and morphology, detection of fluorescence-negative cells, and gating of fluorescence measurements. Light scattering measurement of viruses is generally not feasible with current flow cytometers due to their small size. The problem is aggravated by the fact that the light scattering of particles in this size range (<200 nm) falls off with roughly the sixth power of their linear dimensions.
A new optical layout using darkfield illumination and detection has been developed. A 532-nm laser was used for excitation, and scattered light was collected with large aperture optics.
Light scattering histograms of polymer particles with diameters of 70–300 nm were recorded without gating by other parameters. By extrapolation, a detection limit of about 50 nm was obtained. Different species of virus with sizes of approximately 100 nm also were recorded.
Viral and similar submicroscopic particles are well outside the light scattering detection range of current flow cytometers. The (ungated) light scattering detection limit of commercial instruments is typically around 300–500 nm as obtained with polymer particles and considerably larger for cells such as bacteria and algae due to their high water content and correspondingly lower refractive index. Thus, there appears to be only a couple of reports of flow cytometric measurement of the light scattering of viruses. Hercher et al. (1), using a custom-built instrument, were able to measure T2 phage (∼100 nm), while reovirus, which is 60–80 nm (2), was just discernible from noise. It is not clear to which extent this noise was caused by debris in the sheath water. With a microscope-based instrument equipped with a 488-nm laser, I measured pox virus, which is among the very largest (i.e., roughly 250 nm) (2), and T4 phage (∼100 nm) (3). In subsequent studies on the measurement of the light scattering of viruses, the histograms were gated by fluorescence, and there was no evidence that the light scattering signal was significantly above the noise level (4). Clearly, there is a need for instruments with greater light scatter sensitivity to reach the size range of small bacteria and at least larger and midsize viruses. The early attempts in this regard (1, 3) and the present report have demonstrated that this is technically feasible.
Most viruses are smaller than 200 nm, with the majority near to or smaller than 100 nm (2). This may not seem so much lower than the detection limit of current flow cytometers. However, for particles with sizes well below the wavelength of the excitation light, the light scattering intensity falls off with nearly the sixth power of the linear dimension of the particles. Hence, moving the detection limit from 200 to 100 nm may require an increase of the signal-to-noise (S/N) ratio of a factor in the order of about 60. On the other hand, if sufficient sensitivity can be achieved, this strong size dependence implies a correspondingly high resolution with regard to size, so that, for example, a 10% difference in linear dimensions translates into an approximately 70% difference in light scattering, provided the particles have the same refractive index.
Some flow cytometers do have sufficient fluorescence sensitivity to detect viruses stained with highly fluorescent dyes, such as SYBR green, which bind to nucleic acids (4). However, a study of several species of marine viruses showed no linear relation between the fluorescence signal and the nucleic acid content, indicating a significant degree of non-stoichiometric and possibly nonspecific binding of the dye to capsid components (5). This conclusion is in accordance with the notorious problems of staining viruses with highly DNA-specific dyes such as Hoechst 33258 and 4′6-diamidino-2-phenylindole (Dapi). The reason for these problems may be that the capsid is impermeable to such dyes or the interior of the virus is so dry that there is not sufficient water to facilitate binding. Hence, light scattering, as an additional parameter independent of staining, may be a useful tool for discrimination and identification of viruses by flow cytometry.
Sheath water purity is critical in the measurement of submicroscopic particles. Water from standard purification systems contains significant amounts of debris in this size range, which may contaminate light scattering histograms to an unacceptable degree, even after filtration by standard 0.22-μm filters. I have solved this problem by means of a fluidics system in which the sheath water is re-circulated through multiple high-grade filters.
MATERIALS AND METHODS
A schematic drawing of the optical layout of the instrument is shown in Figure 1A. This optics system is essentially a darkfield configuration in which the excitation light beam is blocked by a field stop situated within the microscope objective that collects fluorescence and light scattering (6). The full numerical aperture (NA) of this oil-immersion lens (DApo 40 UV, Olympus, Tokyo, Japan) is 1.3, and that of the field stop corresponds to 0.42. Thus, the field stop reduces the light-collecting efficiency of this lens only by about 10%, i.e., in proportion to NA squared. The lens used to focus the excitation light on the flow chamber was a Nikon 20× with an NA of 0.4 (Nikon Corp., Kanagawa, Japan). By means of an external aperture, the effective NA of this lens was reduced to about 0.12, which implies that the minimum detection angle was increased to about 16 degrees (half-angle in water). This was done to suppress background caused primarily by imperfections and impurities on the surfaces of the flow chamber. Because most of these scattering objects are relatively large, scattering at low angles dominates much more than does scattering of submicroscopic objects so that an increase of the minimum detection angle leads to a significant increase of the S/N ratio.
Figure 1B shows the optical layout of the previous design (7), which has been commercialized by Skatron AS (Lier, Norway) and Bio-Rad (Hercules, CA). In this design the fluorescence is detected in epi-mode, whereas light scattering is detected within the darkfield created by a field stop situated within the oil-immersion objective, which brings the light into the excitation focus and collects the fluorescence. Whereas these two configurations have approximately the same effective aperture for fluorescence detection, the new configuration has much greater collection efficiency for light scattering, especially large-angle scattering, as shown in Table 1.
Table 1. Effective Collection Efficiencies, (NA)2, for LS1 and LS2 Light Scattering*
LS1, small angle; LS2, large angle; NA, numerical aperture.
The light source used for the present measurements was a 532-nm frequency-doubled YAG laser (JDS Uniphase Corp., San Jose, CA) with a rated power of 25 mW. A diode laser rated at 5 mW at 405 nm (Power Technology, Little Rock, AR) was also used.
The flow cytometer was based on a Bryte HS (Bio-Rad). It was equipped with a closed flow chamber with a rectangular cross section and operated at a flow velocity of 2.0 m/s. Measuring rate was typically about 1,000 s−1. The light scattering detector was a R928 photomultiplier tube (Hamamatsu, Shizuoka-ken, Japan). Fluorescence was not measured, and no gating of the light scattering measurement was applied.
Monodisperse, fluorescent polymer particles with diameters of 176, 264, and 306 nm were obtained from Molecular Probes (Eugene, OR). Non-fluorescent particles of 136, 109, and 74 nm were obtained from Dow Chemical (Midland, MI). These samples, and the last one in particular, are not very monodisperse with regard to size.
The sheath water has to be free from particles producing light scattering comparable to that of the samples. Distilled water filtered through standard 0.22- or 0.1-μm filters does not fulfill this condition. We have solved the problem by recycling the sheath water through a series of filters, as shown in Figure 2. The flow is driven by the level difference of the two 400-ml sheath reservoirs, which are situated about 0.4 m above and 0.6 m below the flow chamber, respectively. The lower reservoir has detectors for minimum and maximum fluid levels, which trigger a small membrane pump that brings the water back to the upper reservoir. Close to the inlet of the flow chamber, the water runs through a 0.1-μm MILLIPAK 20 filter (Millipore Corp., Bedford, MA). Before entering the lower reservoir, the water is filtered through a carbon filter (First Need Water purification system, General Ecology, Exton, PA) and another 0.1-μm filter (Polycap 75TF, Whatman International, Kent, England). The best results were obtained when the water was kept running 24 h a day. Under those conditions, an average of about 1.3 particles/s in the range of 60–100 nm was recorded.
Volumetric sample injection was used according to a system described previously (8).
Light scattering and its polarization for various particle diameters were calculated from Mie theory by means of MieCalc 1.4 software obtained from Dr. Bernhard Michel (Feucht-Moosbach, Germany).
Only light scattering was measured in these experiments. Thus, no other parameters were used for gating the light scattering measurement. Only one light scattering detector was used, covering the full range of scattering angles collected by the oil-immersion objective, i.e., approximately 16–70 degrees.
Light scattering histograms of various particle sizes are shown in Figure 3. In this case the S/N for 109-nm particles is about 36, whereas that for 74-nm particles is reduced to about 10. Thus, the coefficient of variation (CV) of the respective histogram peaks, which is much larger than 1/36 and 1/10, respectively, represents the true size variation of the particles. Extrapolation of the data to an S/N of 1, assuming S ∝ d6, yields a minimum detectable size corresponding to dmin ≅ 50 nm.
The noise was due almost exclusively to the background light from scattering by components of the flow chamber and detection optics. Hence, it depends on the cleanness of the flow cell, the internal scattering of the oil-immersion lens, the width of the laser focus, and the setting of the excitation- and measuring slits. The histogram in Figure 3C was obtained with a somewhat lower S/N than those shown in Figures 3A and 3B due to the use of a larger measuring slit, under which conditions the measurements are not so susceptible to variations in the position of the sample flow.
Raman scattering from the sheath water amounted to less than 10% of the background as demonstrated by comparing the background measured with and without a narrow-band 532-nm filter in front of the photomultiplier detector and correcting for the absorption of the filter at 532 nm. No better results were obtained with the 405-nm laser. Apparently, the increase in scattering intensity by moving from 532 to 405 nm, i.e., a factor of about 1.7, is more than offset by the fivefold higher intensity at the longer wavelength.
Figure 4 shows a size calibration curve as obtained by running the various particle samples under identical conditions. It can be seen that the signal does not fall off with the sixth power of particle diameter as might have been expected from the theory of light scattering of submicroscopic particles (9). The reason for this can be found in the fact that, even for particles in the 100-nm range, the light scattering is not isotropic (Fig. 6A). Thus, the smaller the particle, the larger the proportion of the total scattering at higher scattering angles. Because scattering at the lower angles, i.e., less than 16 degrees, was not detected (see above), and the ratio between the scattering at high and low angles increases with decreasing particle size (Fig. 6A), the resulting signal will fall off with something less than the total signal, which is closer to being proportional to d6.
Figure 5 shows examples of histograms of two different species of virus, cytomegalovirus (CMV) and a marine species Chrysochromulina ericina virus (CeV). CMV is an intermediate-size virus (capsid ∼ 100 nm), with a lipid-containing envelope having a diameter of 150–200 nm, found in a significant proportion of adults. It can cause a dangerous infection in infants and immunosuppressed patients. CeV is somewhat smaller. The CV of the CMV peak corresponds to approximately 2.8% in the linear dimension, whereas that for CeV is approximately 3.9%.
Taking the refractive index of protein and DNA to be comparable to that of polystyrene, the present results show that viruses with capsids larger than 60 nm can be measured with fair precision using the present instrument. It is commonly stated that the scattering of particles significantly smaller than the wavelength of light is essentially isotropic. However, calculations based on the Mie theory (9) show that this is not quite so for particles in the present size range. Thus, as seen from Figure 6A, the scattering intensity of 200-nm particles varies by a factor of approximately 6.5 between 0 and 90 degrees, whereas that for 100-nm particles has a variation that is reduced to approximately 3.8. On the other hand, the angular variation of light scattering of spherical particles is independent of their refractive index (9). In principle, therefore, the ratio between the scattering intensities at two different scattering angles can be used to determine the absolute size of submicroscopic particles, provided the first diffraction minimum is beyond the largest scattering angle measured, as is the case with the present instrument for particle diameters smaller than 200 nm. In the present case, however, where the scattered light is collected over a wide angular range to optimize sensitivity, this approach is hardly feasible.
Another approach could be to measure two polarization components of the scattered light, provided the excitation light is polarized, as in the present case. Thus, the angular dependence of the polarization is a function of particle size but independent of refractive index (9). It turns out, however, that in the present size range, i.e., smaller than 200 nm, the ratio between Pperp and Pparl is essentially independent of particle size, as shown by the example in Figure 6B. With the present type of instrument, therefore, there is no additional information to be gained by measuring the light scattering (LS) at two different directions, e.g., forward and right angles, or the polarization of the scattered light. Hence, particle size can be determined only by means of calibration data, as in Figure 4, provided the refractive index is known. The calibration curve should then be shifted according to the difference in refractive index, m, between the calibration particle and the sample (9):
The photomultiplier high voltages used in the present measurements did not exceed 400 V, which is to say that the gain of the detector was far from being fully exploited, i.e., by a factor of about 1,000. This means that the detection limit was determined primarily by the level of optical background, i.e., light scattered off various parts of the flow cell and detection optics and/or laser noise (ripple). To determine the contribution from laser noise, Nl, in the present measurements a separate set of data was recorded, based on the fact that, whereas photon noise, Np, increases with the square root of the excitation intensity (10), iex, laser noise is proportional to iex. Hence, the noise-to-signal (N/S) ratio, is given by equation 2:
where p, s, and l are constants. According to this equation, a straight line should result when (N/S)2 is plotted against iex−1, as confirmed by Figure 7. The laser noise, i.e., the value of (l/s)2, can be determined from the intersection of this line with the (N/S)2 axis. Analysis of the data in Figure 7 shows that the background due to laser noise was less than 2% of the total noise under the conditions used in the present experiments. Hence, photon noise associated with the background of scattered light is the dominating factor in limiting sensitivity in this case, which is to say that there should be a potential for further improvement of the light scattering sensitivity by reducing this background. In practice, however, this potential may be quite limited, at least with the present type of instrument. Thus, it was found that a significant percentage, approximately 20%, of the background stems from the microscope objective collecting the light. Reducing this part of the background is no trivial task because the high-quality objective used in this instrument represents state-of-the-art technology with regard to suppressing internal light scattering. Most of the remaining 80% comes from the flow chamber. Because the detection limit is proportional to the square root of the background (10), a reduction of the background by, say, 50% would improve the detection limit only by about 30%, which translates into about 5% in terms of particle diameter for particles smaller than 100 nm.
Alternatively, one could apply a higher excitation intensity. However, because the detection limit is inversely proportional to the square root of the intensity (10), there is not much to gain by this approach. Moreover, if fluorescence measurement were included, a significant increase of excitation intensity might cause problems with fluorescence bleaching.
Another approach to increasing S/N is to reduce flow velocity. In this way the amount of light reaching the sample can be increased almost indefinitely, at least in theory. For example, a reduction of flow velocity by a factor of 100 should increase sensitivity by a factor of 10, which should give a reduction by of about 33% in the smallest particle diameter that can be detected. In practice, reduction of flow velocity is accompanied by a wider sample flow and therefore a need for wider excitation focus with corresponding reduction of intensity. Further, the maximum measuring rate must be reduced to avoid unacceptable frequency of coincidences. Moreover, a reduction of flow velocity leads to the same increase of dye bleaching as a corresponding increase of excitation intensity.
Hence, in the present experiments, we may have come close to what is practically possible with detection of submicroscopic particles when using the present type of optical system.
The data suggest that it is possible to measure viral and other submicroscopic particles larger than 60 nm by means of the optical system described above. This detection limit is determined by the level of photon noise, which is due primarily to a background of scattering from optics and flow chamber. Although this background may be somewhat reduced, the resulting reduction of the detection limit will be relatively modest due to the fact that the light scattering signal falls off with almost the sixth power of the particle diameter.
I am indebted to Halvor Rollag and Gunnar Bratbak for kindly providing CMV and CeV, respectively, and to Olav Kaalhus for carrying out the linear regression fitting shown in Figure 7.