February 15, 2014 / Vol. 39, No. 4 / OPTICS LETTERS
1053
OPTiSPIM: integrating optical projection tomography in light sheet microscopy extends specimen characterization to nonfluorescent contrasts Jürgen Mayer,1,2,5 Alexandre Robert-Moreno,1,2 Renzo Danuser,3 Jens V. Stein,3 James Sharpe,1,2,4 and Jim Swoger1,2,* 1
Centre for Genomic Regulation (CRG), Dr. Aiguader 88, 08003 Barcelona, Spain 2
Universitat Pompeu Fabra (UPF), Dr. Aiguader 88, 08003 Barcelona, Spain
3
Theodor Kocher Institute, University of Bern, Freiestrasse 1, 3012 Bern, Switzerland 4 Institució Catalana de Recerca i Estudis Avançats (ICREA), Pg. Lluís Companys 23, 08010 Barcelona, Spain 5 e-mail:
[email protected] *Corresponding author:
[email protected]
Received November 26, 2013; accepted January 3, 2014; posted January 16, 2014 (Doc. ID 201829); published February 13, 2014 Mesoscopic 3D imaging has become a widely used optical imaging technique to visualize intact biological specimens. Selective plane illumination microscopy (SPIM) visualizes samples up to a centimeter in size with micrometer resolution by 3D data stitching but is limited to fluorescent contrast. Optical projection tomography (OPT) works with fluorescent and nonfluorescent contrasts, but its resolution is limited in large samples. We present a hybrid setup (OPTiSPIM) combining the advantages of each technique. The combination of fluorescent and nonfluorescent high-resolution 3D data into integrated datasets enables a more extensive representation of mesoscopic biological samples. The modular concept of the OPTiSPIM facilitates incorporation of the transmission OPT modality into already established light sheet based imaging setups. © 2014 Optical Society of America OCIS codes: (110.6880) Three-dimensional image acquisition; (110.6955) Tomographic imaging; (170.2520) Fluorescence microscopy; (170.3010) Image reconstruction techniques. http://dx.doi.org/10.1364/OL.39.001053
Optical projection tomography (OPT) [1] and light sheet fluorescence microscopies (LSFMs) such as selective plane illumination microscopy (SPIM) [2] are optical 3D imaging techniques applicable to the same size regime, the mesoscopic scale, covering micrometers to millimeters. OPT is the optical analogue of x-ray computed tomography (CT) [3], while SPIM is a fluorescence imaging technique that uses a thin sheet of light for illumination to achieve optical sectioning. Unlike techniques such as mesoscopic fluorescence tomography [4], SPIM and OPT rely on nonscattered or weakly scattered photons for imaging, and thus chemical clearing [5,6] of the sample is sometimes necessary. The limitation to weakly scattering samples is compensated by the fact that OPT and SPIM can achieve high, diffraction-limited resolution even in relatively large samples [7,8]. In the past decade, LSFM has been used intensively to address questions in cell and developmental biology, neurology, and physiology. Even though LSFM has numerous variants (for a review see [9]), the basic principle is the same: laser light illuminates the sample orthogonal to the detection axis in a thin 2D sheet. This light sheet produces optical sectioning at the image plane, because only those in-focus fluorophores that are being imaged are excited. By sequentially moving the specimen through the light sheet the 3D distribution of the sample’s fluorescence is recorded. If the sample exceeds the field of view (FOV) of the imaging optics, it is straightforward to record 3D tiles and stitch them together in a post-imaging process. In this way, the resolution achieved and the size of the FOV are decoupled, and we achieve high resolution in large samples. Using LSFM techniques such as SPIM, however, it is not possible to image the nonfluorescent features of a 0146-9592/14/041053-04$15.00/0
sample. This can be a limitation because these nonfluorescent features are important for many samples of biological interest. In developmental biology, for example, it is difficult to perform whole-mount fluorescent detection of mRNA in mouse models, if the samples are thick enough to require chemical clearing in BABB (a 1∶2 mixture of benzyl alcohol and benzyl benzoate). BABB renders the sample optically transparent but alcohol-soluble chromogens such as the fluorescent stain FastRed get dissolved during the required dehydration. Absorption contrast, however, can be easily obtained by in situ hybridization with the widely used stains NBT and BCIP [10]. These nonfluorescent 3D gene expression patterns obtained by in situ hybridization [1,11] can be imaged with OPT. As the optical analogue of CT, OPT uses 2D projections of a 3D object, collected from multiple orientations. These 2D projections are used to reconstruct the volumetric data, typically via a back-projection algorithm based on an inverse Radon transform. OPT does not work exclusively in nonfluorescent mode but operates in two different modes, fluorescence (emission) OPT and bright field (transmission) OPT [12]. It can therefore image 3D absorbing features in transmission mode and fluorescent features in fluorescence mode. However, unlike SPIM, with OPT it is nontrivial to decouple the resolution from the FOV and obtain high resolution in large samples. Once the specimen exceeds the FOV or the depth of field (DOF), data acquisition and processing are complicated by the assumptions underlying the reconstruction process. Each of these techniques has its advantages and disadvantages. OPT has the advantage of recording nonfluorescent features, while SPIM has the advantage of © 2014 Optical Society of America
1054
OPTICS LETTERS / Vol. 39, No. 4 / February 15, 2014
recording large samples at high-resolution. Therefore, we explored the possibility of constructing an instrument able to take advantage of the features of both techniques at the same time. Here we present such an instrument, the OPTiSPIM, that combines the advantages of transmission OPT with the high resolution of SPIM data. We achieve the combination in a hybrid configuration by partial sharing of the imaging optics for both OPT and SPIM. This is similar to the approach of [13]. However, our system includes optimized Köhler trans-illumination, acquisition of fluorescence OPT and SPIM in the same instrument (Fig. 2), as well as automated sample translation in three dimensions. The OPTiSPIM is composed of four main modules. The first two are the SPIM and transmission illumination arms. These modules are specific to each technique. The sample is rotated and positioned in 3D with micrometer precision using the third module, the micromanipulator. The detection path is the fourth module. These last two modules are common to both imaging techniques. All modules are controlled by custom software written in LabView (National Instruments) for fully automated 3D and time-lapse imaging. The setup is shown in Fig. 1. The first module, the SPIM illumination, combines the different laser lines (wavelengths) and controls excitation intensities by the acousto-optic tunable filter (AOTF). A cylindrical lens breaks the circular beam geometry to produce the 2D light sheet, similar to [14]. The laser beam is collimated in y and focused in z. The final lens is a microscope objective, and couples
Fig. 1. OPTiSPIM setup consists of four major modules: (1) SPIM illumination, (2) transmission illumination, (3) sample positioning, and (4) detection (including fluorescence OPT epiillumination). Module 1 forms the SPIM light sheet. AOTF, acousto-optic tunable filter; RSM, resonant scan mirror; SL, SPIM illumination lens; M, mirror; CL, cylindrical lens; HS, horizontal slit; VS, vertical slit; OL, objective lens. Module 2 trans-illuminates the sample in a Köhler configuration. TIL, transmission illumination lens; TI, transmission illumination iris. Module 3 manipulates the sample during the scan and supports it in the immersion medium. IMC, immersion medium chamber; S, sample. Module 4 images the sample onto the camera. DI, detection iris; EF, emission filter; TL, tube lens. OPT fluorescence excitation can be coupled in if needed. DM, dichroic mirror; FEX, fluorescence excitation lamp.
the excitation light into the sample. The resonant scan mirror (RSM) tilts the light sheet in the image plane of the detection module, to reduce shadowing artefacts [15]. Note that by removing the cylindrical lens, the RSM could in principle be used to implement a digital scanned light-sheet microscope (DSLM) [16] setup, although we have not optimized our system for operation in this mode. The second module comprises the transmission illumination optics. Spatially and temporally incoherent light illuminates the sample in a Köhler configuration, and is imaged on the camera as a shadow pattern [17]. The optics are designed such that the DOF covers approximately half the width of the sample, as images captured by OPT represent projections only of the portion of the sample residing within the DOF [12]. We use an iris placed directly behind the detection objective to reduce the numerical aperture (NA) of the detection and thereby control the DOF. In the sample positioning module, the sample is held in place, translated, and rotated with the same micromanipulators for both OPT and SPIM imaging. Translation can be performed along three orthogonal axes: those of detection, SPIM illumination, and the vertical (microtranslation stages M112.1DG, Physik Instrumente). Rotation is around the vertical axis (rotation stage M116.DGH, Physik Instrumente). The sample is attached to a holder connected to the rotation stage, which can be positioned using the three translation stages. A glass cuvette is used as the immersion chamber, such that the sample is submerged in the immersion fluid and supported from above. In this way, the chamber remains fixed while the sample can be scanned within it. The detection module is also shared by both techniques. A 4f telescope and the appropriate filters (for fluorescence imaging) image the specimen onto the camera. In OPT fluorescence mode, the light of a fluorescence lamp is coupled into the detection path (via a dichroic mirror), such that the detection objective lens can be used for both epi-fluorescence excitation and detection. Thus, during a fluorescence OPT scan, the entire sample is excited simultaneously by completely filling the back aperture of the objective with light. The choice of microscope objective lenses depends on the sample and immersion conditions—typically we use illumination objectives in the 2× to 10× range, and detection objectives from 5× to 20×. For the data sets presented in this Letter, a 2.5× air objective lens (Leica, NPLAN, NA 0.07, WD 11.2 mm) is used to illuminate the sample in SPIM mode (Fig. 1, OL1), resulting in an effective axial resolution of ∼10 μm. A 5 × N Plan Epi air objective (Leica, NPLAN EPI, NA 0.12, WD 14 mm) and a 12-bit cooled CCD camera (Hamamatsu ORCA-ER C4742-80) are used for detection. Since both techniques use the same detection optics, the lateral resolution is similar (the axial resolution in SPIM is typically worse than the lateral) [18]. Although air lenses were used for imaging in BABB immersion, aberrations were not significant for the conditions presented here. The similarities and differences between the two techniques are illustrated comparing a fluorescent OPT scan with a (fluorescent) SPIM scan (Fig. 2), both acquired in the OPTiSPIM. As a sample, we chose a chemically
February 15, 2014 / Vol. 39, No. 4 / OPTICS LETTERS
Fig. 2. Fluorescent OPT scan compared to (fluorescent) SPIM scan, both imaged in the OPTiSPIM. The high endothelial venules (HEV) of a murine lymph node are visualized. (a), (d) full field of view, (b), (e) lateral resolution, (c), (f) axial resolution. Scale bar: 200 μm, scale bar inset: 50 μm.
cleared [5] murine popliteal lymph node (LN). The high endothelial venules (HEV) of the LN are stained with Alexa 633-conjugated MECA-79, a monoclonal antibody. The HEV are of interest from an imaging point of view, as they constitute a network with structure at various scales. The SPIM scan has a slightly better lateral resolution [Figs. 2(d) and 2(e)], as the OPT reconstruction tends to smooth small features [Figs. 2(a) and 2(b)]. SPIM lateral resolution is ∼3 μm and OPT resolution is ∼4 μm, but both resolve the distribution of the HEV. Examining the SPIM dataset from the side (the same dataset projected orthogonal to the detection axis) reveals the elongated ellipsoid shape of the small object in the center of the inset [Fig. 2(f)]. This ellipsoid shape is caused by the asymmetric point spread function in the SPIM scan. In contrast, the small object is roughly symmetrically shaped in the OPT reconstruction in Figs. 2(b) and 2(c). The isotropic OPT resolution intrinsically results from the reconstruction of projections taken over 360°. By computationally stitching the tiles of a SPIM scan in 3D, a global resolution of micrometers can be achieved in samples of several millimeters in size. The embryonic day 12.5 (E12.5) mouse head in Fig. 3(a), for example, has about 40× the volume of the LN shown in Fig. 2. Its structures are visualized (maximum intensity projection of two orthogonal views) by stitching 4 × 3 tiles of a 3D SPIM scan. The stitching was done semi-automatically using the open source software FIJI [19,20]. In the E12.5 mouse head, developing muscles and nerve precursors surround the eye. The craniofacial muscles are responsible for facial movements such as mastication, eye movement and facial expression. The sensory neurons in the face and the muscle-innervating neurons that will control the craniofacial muscles are apparent (for a review, see [21]). To see the spatial relationship between the eye, the controlling muscles, and their corresponding neurons, the E12.5 mouse head was prepared by double immunofluorescence labeling using desmin and Tuj1 antibodies. Tuj1 labels the neuronal cell bodies, dendrites and axons of developing neurons, including ganglion cells; desmin labels differentiating muscle. There is no fluorescent
1055
staining of the eyes themselves, but the pigments of the eye can be imaged with transmission OPT. This makes the E12.5 mouse head an ideal test sample for our OPTiSPIM approach, imaging fluorescently labeled tissues in the context of an absorbing feature. The OPTiSPIM allows us to visualize the 3D relationship between detailed fluorescent features (nerves and developing muscle) and the nonfluorescent eye pigmentation. Figures 3(c)–3(e) show maximum intensity projections of regions of interest of the E12.5 mouse head, along different axes. To provide a comparison for our OPTiSPIM, the E12.5 mouse head was imaged with a commercial OPT scanner (Bioptonics 3001). Figure 3(b) shows the data of the dashed yellow-boxed region of interest in Fig. 3(a) but imaged with the Bioptonics 3001 OPT scanner. Although it shows the same features as the SPIM data in Fig. 3(c), the resolution of the neuronal network is not high enough to resolve single smaller neuronal tubes, particularly in dense regions of the network. SPIM on the other hand achieves significantly higher resolution for the fluorescent features of the sample [Fig. 3(c)] but cannot image the eye pigmentation. To capture the multimodal 3D data set, we first scanned each eye in the transmission OPT mode of the OPTiSPIM. Then we imaged the nerves and the muscles of the entire E12.5 mouse head in SPIM mode. It was not necessary to remove the sample from the OPTiSPIM, or change the configuration of the detection optics during the acquisition of the images. As a result, the 3D datasets acquired
Fig. 3. (a) OPTiSPIM scan of an embryonic day 12.5 (E12.5) mouse head (maximum value projections), left: projection along the dorsoventral axis, right: along the lateral axis. Red: differentiating muscles, green: developing nerves (fluorescent SPIM scans). Blue: eye pigmentation (reconstructed transmission OPT scans). (b) Fluorescence OPT scan (Bioptonics 3001 scanner) compared with (c) SPIM scan of the dashed yellowboxed region of interest. (d) Dorsoventral projection and (e) lateral projection of a region of interest around the eye [orange boxes in (a)]. The differentiating extraocular muscles (red) start to surround the eye (arrowheads). [See also (Media 1).] Scale bar: (a) 1000 μm, (b)–(e) 400 μm.
1056
OPTICS LETTERS / Vol. 39, No. 4 / February 15, 2014
with OPT and SPIM are self-registered, even though the contrasts are achieved in very different manners. This hybrid representation of fluorescent and nonfluorescent 3D structural data gives a more extensive picture of the developing tissues around the eye. A detailed representation of the eye in the context of the surrounding nerves and differentiating extraocular muscles is seen in Figs. 3(d) and 3(e) (Media 1). The different branches of the innervated muscle precursors (arrowheads) that will allow the animal to rotate its eyes within their sockets, and their relation to the eye itself, can be clearly discerned. In summary, the OPTiSPIM platform enables acquisition of two completely distinct 3D contrast mechanisms in one hybrid setup. The combination of OPT and SPIM allows fluorescent contrast imaging in mesoscopic biological samples with high resolution together with nonfluorescent contrast imaging of the same sample. This combination automatically provides registered datasets by preserving the same sample conditions in both modes and is a great advantage over combining data from two separate instruments. By complementing LSFM setups with the OPT modality, the imaging capabilities extend to a contrast mechanism that has previously not been accessible in this context. In addition, because of the common detection and sample manipulation modules, it should be straightforward to incorporate the OPT modality into many of the LSFM setups that have been described [7,9,13], particularly those designed to be capable of multiview imaging [22]. The combination of OPT data with multiview LSFM datasets can provide isotropic resolution globally (data not shown), overcoming the poor axial SPIM resolution shown in Fig. 2(f). Further enhancement to the OPTiSPIM reconstructions could be achieved by deconvolution [22], denoising, or frequencydistance relationship filtering [23]. In conclusion, the OPTiSPIM allows the combination of fluorescent and nonfluorescent 3D contrasts into integrated data sets generated by a single instrument that we believe will enable the investigation of questions in mesoscopic biological imaging. The authors thank Glenda Comai [Institut Pasteur, Paris, France] for providing and preparing the E12.5 mouse head, and they thank Xavier Diego [CRG, Barcelona, Spain] for revising and improving the manuscript. The research was funded in part by the VIBRANT project (No. 228933 of the FP7-NMP) and the Sinergia
project (CRII3_125477 of the Swiss National Science Foundation—SNSF). References 1. J. Sharpe, U. Ahlgren, P. Perry, B. Hill, A. Ross, J. HecksherSorensen, R. Baldock, and D. Davidson, Science 296, 541 (2002). 2. J. Huisken, J. Swoger, F. Del Bene, J. Wittbrodt, and E. H. Stelzer, Science 305, 1007 (2004). 3. J. R. Walls, J. G. Sled, J. Sharpe, and R. M. Henkelman, Phys. Med. Biol. 50, 4645 (2005). 4. C. Vinegoni, C. Pitsouli, D. Razansky, N. Perrimon, and V. Ntziachristos, Nat. Methods 5, 45 (2008). 5. L. Quintana and J. Sharpe, Cold Spring Harb. Protoc. 2011, 664 (2011). 6. H. Hama, H. Kurokawa, H. Kawano, R. Ando, T. Shimogori, H. Noda, K. Fukami, A. Sakaue-Sawano, and A. Miyawaki, Nat. Neurosci. 14, 1481 (2011). 7. J. Huisken and D. Y. Stainier, Development 136, 1963 (2009). 8. V. Ntziachristos, Nat. Methods 7, 603 (2010). 9. P. A. Santi, J. Histochem Cytochem. 59, 129 (2011). 10. M. Hargrave, J. Bowles, and P. Koopman, Methods Mol. Biol. 326, 103 (2006). 11. L. Quintana and J. Sharpe, Cold Spring Harb. Protoc. 2011, 586 (2011). 12. J. B. Pawley, Handbook of Biological Confocal Microscopy, 3rd ed. (Springer, 2006). 13. E. J. Gualda, T. Vale, P. Almada, J. A. Feijo, G. G. Martins, and N. Moreno, Nat. Methods 10, 599 (2013). 14. K. Greger, J. Swoger, and E. H. Stelzer, Rev. Sci. Instrum. 78, 023705 (2007). 15. J. Huisken and D. Y. Stainier, Opt. Lett. 32, 2608 (2007). 16. P. J. Keller, A. D. Schmidt, J. Wittbrodt, and E. H. Stelzer, Science 322, 1065 (2008). 17. J. Sharpe, Annu. Rev. Biomed. Eng. 6, 209 (2004). 18. C. J. Engelbrecht and E. H. Stelzer, Opt. Lett. 31, 1477 (2006). 19. S. Preibisch, S. Saalfeld, and P. Tomancak, Bioinformatics 25, 1463 (2009). 20. J. Schindelin, I. Arganda-Carreras, E. Frise, V. Kaynig, M. Longair, T. Pietzsch, S. Preibisch, C. Rueden, S. Saalfeld, B. Schmid, J. Y. Tinevez, D. J. White, V. Hartenstein, K. Eliceiri, P. Tomancak, and A. Cardona, Nat. Methods 9, 676 (2012). 21. R. Sambasivan, S. Kuratani, and S. Tajbakhsh, Development 138, 2401 (2011). 22. J. Swoger, P. Verveer, K. Greger, J. Huisken, and E. H. Stelzer, Opt. Express 15, 8029 (2007). 23. J. R. Walls, J. G. Sled, J. Sharpe, and R. M. Henkelman, Phys. Med. Biol. 52, 2775 (2007).