A new high-aperture glycerol immersion objective lens and its

Journal of Microscopy, Vol. 206, Pt 2 May 2002, pp. 146 151 Received 11 September 2001; accepted 15 January 2002 A new high-aperture glycerol immersion objective lens and its Blackwell Science Ltd application to 3D-fluorescence microscopy N. MARTINI, J. BEWERSDORF & S. W. HELL High Resolution Optical Microscopy Group, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany Key words. Aberrations, confocal, fluorescence, glycerol, objective lens, refractive index mismatch, resolution. Summary High-resolution light microscopy of glycerol-mounted biological specimens is performed almost exclusively with oil immersion lenses. The reason is that the index of refraction of the oil and the cover slip of ~1.51 is close to that of ~1.45 of the glycerol mountant, so that refractive index mismatchinduced spherical aberrations are tolerable to some extent. Here we report the application of novel cover glass-corrected glycerol immersion lenses of high numerical aperture (NA) and the avoidance of these aberrations. The new lenses feature a semi-aperture angle of 68.5, which is slightly larger than that of the diffraction-limited 1.4 NA oil immersion lenses. The glycerol lenses are corrected for a quartz cover glass of 220 µm thickness and for a 80% glycerol- water immersion solution. Featuring an aberration correction collar, the lens can adapt to glycerol concentrations ranging between 72% and 88%, to slight variations of the temperature, and to the cover glass thickness. As the refractive index mismatch-induced aberrations are particularly important to quantitative confocal fluorescence microscopy, we investigated the axial sectioning ability and the axial chromatic aberrations in such a microscope as well as the image brightness as a function of the penetration depth. Whereas there is a significant decrease in image brightness associated with oil immersion, this decrease is absent with the glycerol immersion system. In addition, we show directly the compression of the optic axis in the case of oil immersion and its absence in the glycerol system. The unique advantages of these new lenses in high-resolution microscopy with two coherently used opposing lenses, such as 4 Pi-microscopy, are discussed. Introduction In the last decade biological fluorescence microscopy has witnessed a strong tendency towards the imaging of live cells. Correspondence to: S. W. Hell. Fax: +49 551 201 1085; e-mail: shell@gwdg.de The increasing utilization of the green fluorescent protein and its mutants has further nurtured this trend. Because the refractive index (RI) n of live cellular material is close to that of water, high-resolution three-dimensional (3D) imaging of living cells enforced the development of high-aperture water immersion lenses. The reason is that with standard highaperture oil immersion lenses high-resolution studies of live cells can be performed only in the closest vicinity to the cover slip (< 5 µm); otherwise the discrepancy between the RI of water (n = 1.33) and that of the oil/glass system (n = 1.51) leads to a substantial spherical aberration. In confocal microscopy, this RI-induced spherical aberration is a common source of errors (Sheppard & Cogswell, 1991): the detected signal decreases with increasing penetration depth, the resolution degrades, and the recorded images appear axially elongated or compressed (Hell et al., 1993; Jacobsen & Hell, 1995; Török et al., 1997; Egner et al., 1998; Egner & Hell, 1999; Bucher et al., 2000). For the imaging of watery specimens cover glass-corrected water immersion lenses avoid RI-induced spherical aberration and enable deep penetration into a watery specimen. Not surprisingly, these lenses became available from virtually all manufacturers of high performance confocal microscopes. In spite of this development, to this end a large portion of investigated cells are still mounted in glycerol-based media featuring a RI of n ~ 1.45. A similar portion of samples is embedded in polyvinyl alcohol-based media whose RI of n ~ 1.38 1.40 often increases with time and is then closer to glycerol than to immersion oil or water. Usually oil immersion objective lenses have been used for imaging these specimens. Albeit much less severe than for a watery specimen, the residual RI mismatch still degrades the 3D-data when specimens thicker than about 10 20 µm are observed with NA = 1.3 1.4 oil lenses (Hell et al., 1993). Another area benefiting from reduced index mismatch is 4 Pi-confocal microscopy (Hell & Stelzer, 1992) and related techniques applying interfering beams of two objective lenses (Gustafsson et al., 1999). In these systems, the RI mismatch 2002 The Royal Microscopical Society

A NEW HIGH-APERTURE GLYCEROL IMMERSION OBJECTIVE 147 encountered with oil immersion lenses results in an axial phase shift that depends linearly on the axial scanning position (Hell et al., 1997; Egner et al., 1998). In a recent pilot study, NA = 1.35 glycerol lenses have been developed. Here we report on the optical performance of such a lens in a standard confocal fluorescence microscopy setting. The glycerol lens The newly developed lens is a plan apochromatic glycerol objective (HCX PL APO 100 /1.35 GLYC CORR, Leica Microsystems, Wetzlar, Germany) with a specified NA of 1.35. With a RI of glycerol of n e = 1.451 (concentration 80%), the semi-aperture angle of 68.5 is slightly higher than the 67.3 of the widely employed 1.4 NA oil immersion lenses. Its correction collar buffers refractive indices ranging from n e ~ 1.44 to ~1.46, which corresponds to glycerol concentrations of about 72 88%. The best optical correction, however, is obtained around the midpoint, which can be adjusted easily by tuning the water concentration. The lenses were designed for quartz cover slips with a RI of n = 1.46, because of its proximity to that of glycerol. The cover glass thickness is defined as 220 ± 10 µm by design choice. The magnification of 100 and a free working distance of 100 µm are largely analogous to 1.4 NA oil immersion lenses; the same applies to the field flatness. Featuring a correction collar, the high-na lens is conceptually closer to high-aperture water immersion lenses than to the standard glycerol immersion lenses that have been available so far. These glycerol immersion lenses are optimized for UV illumination in photometry; glycerol is used because of its low absorption in the UV regime compared to immersion oil. Materials and methods The measurements presented in this paper were performed with the Leica confocal microscopes TCS SP1 and SP2. A 1.4 NA Leica oil immersion lens (HCX PL APO 100 /1.40 0.7 oil) served as reference for the oil immersion case. The reference measurements were carried out directly after the glycerol lens measurements; both were attached to the same nosepiece. The RI of the glycerol medium ranged between n e = 1.452 and 1.457. Four different kinds of sample were used. Reflecting cover slips A standard cover slip, as well as a quartz cover slip, were evaporated with silver and mounted with the clear surface towards the lens. The illumination wavelengths of 488, 543 and 633 nm were used simultaneously. The reflected light was chromatically separated and recorded with three photomultiplier tubes. The pinhole was set to its minimal value (< 0.1 Airy disks) and two-dimensional reflection images containing the optical axis (xz-images) were recorded. The difference in axial position of the signal maxima gave the axial chromatic aberrations of the system. As all wavelengths were recorded at the same time, sample movement artefacts were avoided. Fluorescent monomolecular layers To determine the z-response of the system, i.e. the axial response of an infinitely thin fluorescent layer, a standard cover slip as well as a quartz cover slip were coated with a monomolecular fluorescent polydiacethylen Langmuir Blodgett film (Schrader et al., 1998). The layers were excited at 488 nm and the fluorescence emission was collected between 550 nm and 650 nm. The standard and the quartz cover slip were placed on microscope slides with a drop of immersion oil and glycerol, respectively, in between. XZimages were recorded for two different pinhole diameters corresponding to 1.0 Airy disks and 0.5 Airy disks. Care was taken to avoid saturation of the fluorophore. The ideal collar position of the glycerol lens was found by maximizing the peak intensity of the fluorescence signal of the layer. The recorded images were averaged in the x-direction to improve the signal-to-noise ratio, which then enabled us to determine the z-response and to establish their full width at half maximum (FWHM) as a measure of the axial resolution of the system. Dye solution Rhodamine 6G was dissolved in glycerol at a concentration of 10 5 M and mounted with the cover slips as appropriate to form a fluorescent half-space or fluorescent sea. The thickness of the solution was elected to be larger than the free working distance of the objective lenses. XZ- intensity profiles of the half- space, i.e. the sea-response were recorded at an excitation wavelength of 488 nm and a detection band of 495 530 nm. The pinhole size was set to 1.0 Airy disks in diameter. It was ensured that the edge of the sample was still in the scan area and that the penetration depth did not exceed the free working distance of the objective lenses. The data were averaged in the x-direction to obtain the z-profile of a homogeneous dye solution. Glycerol-mounted biological specimen Caenorhabditis elegans strain PD4792 provided by the Caenorhabditis Genetics Center was grown at 20 C on nematode growth medium plates (Brenner, 1974) seeded with Escherichia coli OP 50. This strain expresses the EGFP driven by the myo-2 promoter. The sample preparation was based on the work of Finney & Ruvkun (1990). Adult worms were washed in physiological buffer 1 M9 and then chemically fixed with 3.5% paraformaldehyde (ph 6.8) for 5 min. After fixation they were extracted with 1% Triton X-100 in phosphate buffered saline (ph 6.9) for 20 min at room temperature to permeabilize the membrane. Next the C. elegans were rinsed and resuspended in distilled water. The specimens were attached to a standard

148 N. MARTINI ET AL. glass cover slip using poly l-lysin. By adding 20 µl of glycerol the sample was mounted between the standard cover slip and a quartz cover slip, so that the sample could be viewed consecutively with both objective lenses upon reversion. The images were recorded 6 weeks after preparation so ensuring that the glycerol was equally distributed throughout the sample. Stacks of xy-images of 25 µm 25 µm in size with an axial distance of 1.14 µm were recorded with each objective lens. Excitation of EGFP was performed at 488 nm and fluorescence was registered between 500 nm and 530 nm. The detection pinhole diameter corresponded to 1.0 Airy disks. Results Axial resolution and axial chromatic aberrations Intensity [a.u.] 1.0 0.8 0.6 0.4 0.2 Glyc / 0.5 Airy Oil / 0.5 Airy Glyc / 1.0 Airy Oil / 1.0 Airy Fig. 2. Axial chromatic aberrations. Relative axial offset of the reflection maxima for the NA = 1.4 oil immersion and the novel NA = 1.35 glycerol immersion lens at 488, 543 and 633 nm. First, we established the FWHM of the z-responses shown in Fig. 1 gained by the monomolecular fluorescent layers. For a pinhole diameter corresponding to one Airy disk we found an axial FWHM of 610 ± 20 and 640 ± 20 nm for the glycerol and the oil immersion lens, respectively. Reducing the pinhole diameter down to 0.5 of an Airy disk produced a slightly narrower z-response of 460 ± 20 nm and 510 ± 20 nm, respectively. The comparison of the experimental results proves that the axial resolution of the two lenses is only marginally different. It is well known that because of manufacturing tolerances, the FWHM of z-responses of individual lenses of a particular kind vary by about 50 nm. The corresponding theoretical FWHM, as calculated by a vectorial theory (Richards & Wolf, 1959), yielded for both the oil and the glycerol lens 600 nm and 460 nm, for 1.0 Airy disks and 0.5 Airy disks, respectively, which is in good agreement with the experimental values. We note that RI mismatch aberrations are excluded in these measurements because the fluorescent layer was immediately below the cover slip. Hence, the experiment effectively reveals that the axial resolution of the glycerol lens is of the order of a highly corrected 1.4 NA oil immersion lens. The axial chromatic aberrations were extracted from xz-images of the reflective cover slips. The relative axial position of the reflection maximum reveals the axial chromatic aberrations. Figure 2 shows that for the oil and the glycerol lens the magnitude of the longitudinal chromatic shift is of the same order and smaller than 250 nm. The curves represent a square fit, which is a reasonable assumption for highly corrected lenses. Because of tolerances in the manufacturing process, the curves may be shifted to longer or shorter wavelengths for each individual lens. 0.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 z [µm] Fig. 1. Experimental z-response of a scanning confocal fluorescence microscope to an axially ultrathin fluorescent plane for the NA = 1.4 oil immersion and the NA = 1.35 glycerol immersion lens. The z-responses shown as solid lines were recorded for a detection pinhole diameter corresponding to half of that of the Airy disk. The dashed curves are the full Airy disk counterparts. The excitation wavelength was 488 nm; the FWHM is a measure of the axial sectioning capability of the microscope and is established to be around 500 nm for both systems. Refractive index mismatch induced signal decrease By scanning the fluorescent glycerol sea in the axial direction we investigated the decrease of the signal in the confocal detector as a function of the axial penetration depth. As anticipated from the theory and experiments (Hell et al., 1993; Egner et al., 1998), in the case of the oil immersion lens the spherical aberration introduced by the RI mismatch leads to a marked decrease in image brightness with increasing penetration depth. The reason is that the signal detected in the confocal detector is largely proportional to the square of the intensity point spread function of the lens (Wilson &

A NEW HIGH-APERTURE GLYCEROL IMMERSION OBJECTIVE 149 Intensity [a.u.] 1.0 0.8 0.6 0.4 0.2 0.0 Glyc Oil 0 20 40 60 80 100 z [µm] Fig. 3. Experimental sea-response. Fluorescence signal as a function of the axial penetration depth, determined through a uniformly fluorescent half-space of a glycerol-based mountant. In the case of the glycerol lens the fluorescence signal remains constant with increasing axial depth; the confocal images from within the specimen feature the same brightness as those acquired close to the cover slip. This is not so with the oil immersion lens where the fluorescence signal drops significantly with increasing penetration depth. Sheppard, 1984). As the aberrations lead to an increase of side-lobes at the expense of the main focal maximum, squaring of the PSF lowers the signal of the outer parts and leads to a reduction of the total signal detected in the confocal pinhole (Hell et al., 1993). At an axial depth of ~30 µm the signal is reduced by 25% in the experiment, and at 80 µm the signal is down to almost the half of that found immediately beneath the cover slip, as can be seen in Fig. 3. For smaller pinhole diameters this effect is even more pronounced. In contrast, the successful compensation of the RI mismatchinduced aberrations with the glycerol immersion lens yields a constant signal throughout the axial scanning range of 80 µm. Refractive index mismatch-induced focal shift and axial scaling Another effect of RI mismatch-induced aberrations is a slight axial shift of the focus, away from the geometrical focal point (Hell et al., 1993; Egner et al., 1998). The actual focus position in the axial direction differs by a factor β from its nominal counterpart, which is the focal position that one would assume in the absence of a RI mismatch. It is known from theory and experiment that in the case of the 1.4 NA oil immersion lens focusing into glycerol, β = 0.92, so that the object appears thicker by 9% (Hell et al., 1993; Egner et al., 1998). In the case of the glycerol lens, the axial elongation should be absent (β = 1). This different behaviour was indeed found in the experiment shown in Fig. 4, where we depicted two series of xy-images of the same C. elegans sample. The image stacks were recorded with identical mechanical axial displacements of the sample of z = 1.14 µm. As the sample was sandwiched between a standard cover slip and the quartz cover slip, the sample was turned upside down when switching between the two lenses. Hence, the different origins of the z-axis in Fig. 4. The latter also reveals that, in the case of the oil immersion lens one more xy-image was required to image the whole sample three-dimensionally, despite the fact that the mechanical axial steps were identical. In accordance with the theoretical expectation, the approximately 10 µm thick specimen is erroneously represented to be ~1 µm thicker, so that for quantitative analysis the 3D-image requires re-scaling. In the case of the glycerol lens the mechanical steps correspond to the optical steps in the 3D-image stack. Discussion and conclusion In addition to its obvious advantages to confocal fluorescence microscopy, the glycerol objective lens is of particular relevance to 4 Pi-confocal microscopy of fixed specimens (Hell et al., 1997; Nagorni & Hell, 1998) and to related concepts using two opposing lenses, such as the I 5 M (Gustafsson et al., 1999; Nagorni & Hell, 2001). In the case of the 4 Pi- confocal microscope, the spherical aberration induced by the RI mismatch between the oil/glass system and the glycerol mountant limits the object thickness to 10 20 µm. Moreover, the RI mismatch causes a linear shift in the relative phase between the interfering spherical wavefronts upon axial movement of the sample, which has to be actively compensated during axial scanning. These precautions are not needed with glycerol lenses. We note for completeness, that the advent of highly corrected water immersion lenses recently enabled and simplified 4 Pi-confocal microscopy of (live) watery specimens similarly to how the glycerol lenses are expected to simplify the conditions for imaging glycerol-mounted specimens. In I 5 M only high- aperture oil immersion lenses have been used so far, so there is a requirement for raising the RI of the specimen as close as possible to 1.5 (Gustafsson et al., 1999). By eliminating this necessity, glycerol immersion lenses should be of even greater value to I 5 M, than to 4 Pi-confocal microscopy. In summary, we reported on the axial imaging properties of a novel high-na glycerol immersion lens. The lens is designed to work with quartz cover slips and features a correction collar for spherical aberration compensation. By establishing the achievable axial sectioning capability of this lens in confocal fluorescence microscopy we showed that its optical performance is comparable to that of commonly used high-na oil immersion lenses. A distinct advantage over the latter is that when imaging glycerol-mounted specimens, RI mismatch-induced problems such as signal loss at deep penetration, geometrical distortion and loss of resolution are virtually eliminated.

150 N. MARTINI ET AL. OIL z = 13.68 µm z = 12.54 µm z =... x y GLYC z = 0.00 µm z = 1.14 µm z = 2.28 µm z =... z =... z = 2.28 µm z = 1.14 µm z = 0.00 µm 10 µm z =... z = 12.54 µm Fig. 4. Series of xy-images (3D-stacks) of the pharyngeal muscle of a C. elegans recorded with the oil immersion lens (yellow background) and its glycerol counterpart (blue background). Both stacks were recorded with equidistant axial mechanical displacement but, featuring an additional XY-image, the oil immersion stack incorrectly indicates a ~10% thicker sample. Because of the aberrations involved with recording a glycerol-mounted specimen with a 1.4 NA oil immersion lens, the optical axial distance is reduced by 10% with respect to that of the mechanical displacement. In the case of the glycerol immersion lens the mechanical and optical axial step intervals are equal. The sample was located between a standard and a quartz cover glass and therefore had to be reverted for viewing with different objective lenses. The positions of the lenses relative to the sample are shown schematically. The given numbers represent the mechanical displacements of the sample for each slice. Acknowledgements We gratefully acknowledge useful discussions with P. Euteneuer and R. Krüger of Leica Microsystems Wetzlar GmbH, Wetzlar, Germany. The nematode strain used in this work was provided by the Caenorhabditis Genetics Center funded by the NIH National Center for Research Resources (NCRR). References Brenner, S. (1974) The genetics of Caenorhabditis elegans. Genetics, 77, 71 94. Bucher, D., Scholz, M., Stetter, M., Obermayer, K. & Pflüger, H.-J. (2000) Correction methods for three-dimensional reconstructions from confocal images: I. tissue shrinking and axial scaling. J. Neurosci. Meth, 100, 135 143. Egner, A. & Hell, S.W. (1999) Equivalence of the Huygens-Fresnel and Debye approach for the calculation of high aperture point-spread-functions in the presence of refractive index mismatch. J. Microsc. 193, 244 249. Egner, A., Schrader, M. & Hell, S.W. (1998) Refractive index mismatch induced intensity and phase variations in fluorescence confocal, multiphoton and 4Pi-microscopy. Opt. Commun. 153, 211 217. Finney, M. & Ruvkun, G. (1990) The unc-86 gene product couples cell lineage and cell identity in C. elegans. Cell, 63, 895 905. Gustafsson, M.G.L., Agard, D.A. & Sedat, J.W. (1999) I 5M: 3D widefield light microscopy with better than 100 nm axial resolution. J. Microsc. 195, 10 16. Hell, S.W., Reiner, G., Cremer, C. & Stelzer, E.H.K. (1993) Aberrations in confocal fluorescence microscopy induced by mismatches in refractive index. J. Microsc. 169, 391 405. Hell, S.W., Schrader, M. & van der Voort, H.T.M. (1997) Far-field fluorescence microscopy with three-dimensional resolution in the 100 nm range. J. Microsc. 185, 1 5.

A NEW HIGH-APERTURE GLYCEROL IMMERSION OBJECTIVE 151 Hell, S. & Stelzer, E.H.K. (1992) Properties of a 4Pi-confocal fluorescence microscope. J. Opt. Soc. Am. A, 9, 2159 2166. acobsen, H. & Hell, S.W. (1995) Effect of refractive index on the imaging of a confocal fluorescence microscope employing high aperture lenses. Bioimaging, 3, 39 47. Nagorni, M. & Hell, S.W. (1998) 4Pi-confocal microscopy provides threedimensional images of the microtubule network with 100 to 150-nm resolution. J. Struct. Biol. 123, 236 247. Nagorni, M. & Hell, S.W. (2001) Coherent use of opposing lenses for axial resolution increase in fluorescence microscopy. I. Comparative study of concepts. J. Opt. Soc. Am. A, 18, 36 48. Richards, B. & Wolf, E. (1959) Electromagnetic diffraction in opticel systems II. Structure of the image field in an aplanatic system. Proc. Roy. Soc. Lond. A, 253, 358 379. Schrader, M., Hofmann, U.G. & Hell, S.W. (1998) Ultrathin fluorescent layers for monitoring the axial resolution in confocal and two-photon fluorescence microscopy. J. Microsc. 191, 135 140. Sheppard, C.J.R. & Cogswell, C.J. (1991) Effects of aberrating layers and tube length on confocal imaging properties. Optik, 87, 34 38. Török, P., Hewlett, S.J. & Varga, P. (1997) The role of specimen-induced spherical aberration in confocal microscopy. J. Microsc. 188, 158 172. Wilson, T. & Sheppard, C.J.R. (1984) Theory and Practice of Scanning Optical Microscopy. Academic Press, New York.