Full-field supercritical angle fluorescencemicroscopy for live cell imaging

Thomas Barroca,1 Karla Balaa,1,2 Julie Delahaye,1 Sandrine Lévêque-Fort,2 and Emmanuel Fort1,*1Centre d’Imageries Plasmoniques Appliquées, Institut Langevin, Ecole Supérieure de Physique et de Chimie Industrielles

(ESPCI) ParisTech, CNRS UMR 7587, Université Paris Diderot, 10 rue Vauquelin, 75 231 Paris Cedex 05, France2Institut des Sciences Moléculaires d’Orsay and Centre de Photonique Biomédicale (CLUPS),

CNRS Université Paris-Sud 11, UMR 8214, F91405 Orsay Cedex, France*Corresponding author: emmanuel.fort@espci.fr

Received May 12, 2011; revised June 17, 2011; accepted July 11, 2011;posted July 12, 2011 (Doc. ID 147486); published August 5, 2011

We introduce a full-field fluorescence imaging technique with axial confinement of about 100nm at the sample/substrate interface. Contrary to standard surface imaging techniques, this confinement is obtained through emissionfiltering. This technique is based on supercritical emission selectivity. It can be implemented on any epifluores-cence microscope with a commercial high numerical aperture objective and offers a real-time surface imaging cap-ability. This technique is of particular interest for live cell membrane and adhesion studies. Using human embryonickidney cells, we show that one can observe simultaneously the surface and in-depth cell phenomena. © 2011Optical Society of AmericaOCIS codes: 170.2520, 260.6970, 220.0220, 240.0240.

Numerous cell mechanisms, like membrane traffickingand adhesion processes, are located in direct vicinityof the membrane [1]. The understanding of such pro-cesses is of crucial importance in many biomedicalissues. Because of the minute concentrations of biomo-lecules involved, fluorescence microscopy is the widelyfavored technique to investigate such systems. However,the diffraction-limited axial sectioning is not sufficient toobserve membrane processes separately from the innercell activity.In this context, the total internal reflection fluores-

cence (TIRF) configuration has emerged as a powerfulsurface imaging technique [2]. Axial sectioning isobtained by illuminating above the critical angle. It pro-duces an evanescent field that excites only fluorophoresin the direct vicinity of the interface. Excitation confine-ment enables one to observe minute concentrations offluorophores at the interface while reducing the back-ground noise from the inner part of the cell. The devel-opment of high numerical aperture immersion oilobjective lenses has popularized this technique in athrough-the-objective configuration. The sensitivity ofthis technique is mainly limited by the intrinsic light scat-tering into the cells, which eventually results in a loss ofexcitation light confinement [3].An alternative approach has been proposed to obtain

axial sectioning [4]. It takes advantage of the fluoro-phore/interface distance-dependent emission. When afluorophore in medium 1 (refractive index n1) is placedin the direct vicinity of the interface with medium 2(refractive index n2 > n1), part of the fluorescence isemitted above the critical angle θC ¼ arcsinðn1=n2Þ. Thissupercritical fluorescence is also called “forbidden light”since the Snell–Descartes refraction law does not allowsuch an emission [4]. Supercritical angle fluorescence(SAF) originates from the fact that evanescent compo-nents of the emission dipole can become propagativein medium 2. Thus, SAF contribution decreases sharplywith the fluorophore/interface distance d, contrary to theundercritical angle fluorescence (UAF) components that

remain constant [5]. In order to compare SAF with TIRF,we study the molecular detection efficiency (MDE) as afunction of the fluorophore/substrate distance d for bothtechniques. The MDE is given by [6]

MDEðdÞ ¼ ΓexcðdÞ × QY ðdÞ × MCEðdÞ; ð1Þ

where Γexc is the excitation rate, QY is the quantum yieldof the fluorophore, and MCE is the molecular collectionefficiency. For simplicity, we will assume that thefluorophore has QY ¼ 1. In TIRF, the variation of theMDE with d is dominated by the excitation profileΓTIRFexc ðdÞ ∝ expð−d=δÞ, where δ is the penetration depth

of the evanescent field. δ is a function of the wavelengthλ, the incident angle θinc, and the refractive indices n1 andn2. In contrast to TIRF, for the SAF technique, the MDE isdominated by MCEðdÞ while Γexc is roughly constant.Figure 1(a) shows the emission lobe in glass for a fluor-ophore placed at the interface (d ¼ 0). The SAF emissionappears in red and represents about 50% of the emissionin the glass [6]. The normalized MDE in TIRF for differentangles θinc and in SAF are represented Fig. 1(b). TheMDE in TIRF has been obtained multiplying the (UAFþSAF) MCE by the TIRF excitation profile ΓTIRF

exc . For SAF,

Fig. 1. (Color online) (a) Polar plots of the emission patternfor an isotropically oriented fluorophore positioned at the glass/water interface. (b) Normalized MDE versus fluorophore/interface distance for SAF and for various TIRF angles.

August 15, 2011 / Vol. 36, No. 16 / OPTICS LETTERS 3051

0146-9592/11/163051-03$15.00/0 © 2011 Optical Society of America

the normalized MDE is simply given by the MCE. We cal-culate the MCE for UAF and SAF using the vectorialDebye integral model [7]. The calculated theoreticalpenetration depth in SAF is equal to 129 nm (withλ ¼ 593 nm, n1 ¼ 1:33, and n2 ¼ 1:51), which is the pene-tration depth of TIRF at 66°.Recently, an implementation of the SAF technique has

been proposed using a specifically designed objectivelens with a parabolic mirror to collect supercritical an-gles [8]. This objective lens has the advantage of simpli-city and low cost, which is of particular interest inbiosensing applications [9]. However, since imaging withthis objective involves objective/sample scanning, it is re-latively slow and limited to fixed biological samples.Here we propose an alternative configuration to obtain

directly full-field SAF (F2-SAF) images for real-time ima-ging of dynamic biological processes. This approach hasbeen introduced by Axelrod about ten years ago with a1.65 NA objective lens [10]. Acquired images showed qua-litative axial confinement and a loss of lateral resolution.In this Letter, we give quantitative analysis of the opticalcharacteristics of this technique and show that it opensthe path to biological applications using a standard im-mersion oil objective.The schematic of the experimental setup is repre-

sented in Fig. 2. We use a commercial invertedmicroscope (Nikon Eclipse Ti) with an apochromatobjective lens 1.49 NA 60× from Nikon and an electronmultiplying CCD (EM-CCD) camera (iXonþ, Andor Tech-nology). The illumination source is a standard fibred130W mercury lamp, which provides a homogeneous il-lumination on the sample. A two-lens system is includedin the optical path to image the back focal plane (BFP) ofthe objective on a conjugate plane. The relation betweenthe radius ρ in the BFP and the angle of emission θem isgiven by the Abbe condition: ρðθemÞ ¼ n2f sinðθemÞ,where f is the focal distance of an apochromat objectiveand n2 is the refractive index of immersion medium. Acircular mask in the shape of an opaque disk is placedon this conjugate plane to remove UAF. It is mountedon a micrometric XYZ translation stage that allows finepositioning. An additional removable Bertrand lens canbe added in the beam path to image directly the BFP

on the camera. This permits an easy centering and axialpositioning of the mask. An image of the BFP is shown inFig. 2 for a solution of Rhodamine (obtained with a Nikoncamera Reflex D700). UAF and SAF areas are easilyidentifiable.

For biological studies, it can be of great interest to ac-quire in parallel the standard epifluorescence image(UAFþ SAF) and the F2-SAF image. In order to obtaina simultaneous acquisition, a 20∶80 beam splitter cubehas been added to split the beam into a UAFþ SAF path(20%) and an F2-SAF path (80%) on two identical andsynchronized EM-CCD cameras.

In the following paragraph, we focus on the opticalperformances of this F2-SAF technique: the axial section-ing and the lateral resolution. We use a method intro-duced by Mattheyses and Axelrod [11] to measure thepenetration depth. This technique takes advantage ofthe intensity profile of 10 μm fluorescence beads to mea-sure the penetration depth. The measured values are160� 15 nm. This is in good agreement with the Debyemodel. Besides, the penetration depth is constant all overthe image.

The lateral resolution of the F2-SAF is measured by ac-quiring the two-dimensional (2D) point spread function(PSF) of 100 nm fluorescent latex beads (FluoSpherescarboxylate-modified microspheres, Invitrogen). Toreach a sufficient accuracy for the 2D PSF measurementson the camera, we have added a 6:7× afocal opticalsystem, composed of two doublet lenses of focal lengthF1 ¼ 30mm and F2 ¼ 200mm, between the microscopeand the camera. We thus obtain an oversampling of42 nm=pixel (far above the Nyquist criterion).

Figure 3(a) shows the theoretical normalized cross-sectional profile (dashed lines) of the PSF obtained withthe vectorial Debye integral model, which takes into ac-count the polarization [12]. It is in good agreement withthe experimental profiles (smooth lines) in UAFþ SAFand F2-SAF. In the case of a dipole perpendicular to thesurface, the emission is polarized radially with a mini-mum at the center and a larger profile. The dipole aver-age orientation has been used as a fitting parameter. Anearly isotropic distribution gives the best fit (with small

Fig. 2. (Color online) Schematic of the experimental setup formultimodal real-time imaging with two parallel channels for theepifluorescence and F2-SAF images. DM, dichroic mirror; L1,lens1; L2, lens2; L3, lens3; L4, lens4; Exc. filter, excitation filter;Em. filter, emission filter.

Fig. 3. (a) Normalized PSF profiles in UAFþ SAF and F2-SAF:model (dashed lines) and experiment (smooth lines). (b),(c) Experimental PSF images in UAFþ SAF and F2-SAF respec-tively; associated calculated images are shown in (d) and (e).

3052 OPTICS LETTERS / Vol. 36, No. 16 / August 15, 2011

enhanced orthogonal proportion of 4%). The experimen-tal images of a single nanobead are shown in Fig. 3(b) forUAFþ SAF and Fig. 3(c) for F2-SAF. The associated the-oretical images are shown in Figs. 3(d) and 3(e) respec-tively. As expected, we notice a loss of resolution on theF2-SAF PSF [13,14]. To evaluate the resolution, we com-pare the FWHM. The UAFþ SAF PSF is known to befitted accurately by a Gaussian[15]. The normalized in-tensity profile IðrÞ is given by IðrÞ ¼ expð− r2

2σ2Þ, wherethe standard deviation σ is the only fitting parameter. TheFWHM is given by W ¼ 2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 lnð2Þp

σ. The Gaussian curvefits accurately the experimental data. For UAFþ SAFcollection, we obtain the value Wepi ¼ 198 nm� 2 nm,which is in good agreement with the one expected fromthe theory: W ¼ 197 nm (σ ¼ 0:21 × λem=NA). For SAFcollection, we measure WF2

−SAF ¼ 347� 5nm (averageover 15 nanobeads). We also pay a particular attentionto the study of any possible off-axis modifications ofthe PSF due to the aberrations introduced by the objec-tive lens. The value of W is constant all over the image.We validate this full-field imaging technique by per-

forming real-time cell imaging. We use human embryonickidney (HEK 293) cells transfected with type I cannabi-noid receptors tagged with green fluorescent protein.Two cameras synchronized together with a time expo-sure of 300ms have been used to simultaneously obtainimages of the whole cell and of its membrane. Figure 4(a)shows HEK cells in UAFþ SAF and their membranes[Fig. 4(b)] in F2-SAF, where the fluorescence comingfrom the inner parts of the cells is clearly removed. Theseimages have been obtained with a simple white lamp. Asexpected, there is a slight loss of lateral resolution in theF2-SAF image; however, the quality of the image is satis-factory. By using this configuration, it is possible to moni-tor membrane trafficking in real time.In this Letter, we have shown that F2-SAF enables real-

time cell membrane imaging with a simple mask placed

in a conjugated plane of the BFP of a commercial objec-tive to remove UAF components. We observed a slightloss in the lateral resolution that is due to the largerPSF of the fluorophores with emission dipoles orthogo-nal to the surface. Nevertheless, the quality of the imagespermits us to easily resolve the structures of the mem-brane. Besides, image quality could be further improvedby using deconvolution techniques. The axial sectioningis of the order of 100 nm. Moreover, penetration depthand lateral resolution performances are maintainedthroughout the whole image. The F2-SAF technique al-lows one to perform the observation of the whole cell andits membrane simultaneously at low cost using a stan-dard epifluorescence microscope and an incoherentsource that provides an homogeneous illumination allover the sample. We are currently applying this techniqueto biomedical issues, in particular the study of dictyoste-lium motility and endocytosis tracking. This techniquecould also find applications in other field like liquid–liquid interface studies. Ongoing developments focuson alternative optical configurations using removablemasks with a single camera.

The authors thank C. Boccara and S. Grésillon for fruit-ful discussions, S. Lécart for cell samples, S. Sivankuttyfor helpful comments on this article. This work has beensupported by grants from Region Ile de France andC’Nano Ile de France.

References

1. M. S. Bretscher and C. Aguado-Velasco, Curr. Opin. CellBiol. 10, 537 (1998).

2. D. Axelrod, Traffic 2, 764 (2001).3. M. Oheim and F. Schapper, J. Phys. D 38, R185 (2005).4. T. Ruckstuhl, J. Enderlein, S. Jung, and S. Seeger, Anal.

Chem. 72, 2117 (2000).5. T. Ruckstuhl and D. Verdes, Opt. Express 12, 4246 (2004).6. E. Fort and S. Grésillon, J. Phys. D 41, 013001 (2008).7. B. Richards and E. Wolf, Proc. R. Soc. Lond. A 253,

358 (1959).8. T. Ruckstuhl, M. Rankl, and S. Seeger, Biosens. Bioelectron.

18, 1193 (2003).9. C. M. Winterflood, T. Ruckstuhl, D. Verdes, and S. Seeger,

Phys. Rev. Lett. 105, 108103 (2010).10. D. Axelrod, J. Biomed. Opt. 6, 6 (2001).11. A. L. Mattheyses, D. Axelrod, J. Biomed. Opt. 11, 014006

(2006).12. W. T. Tang, E. Chung, Y-H. Kim, P. T. C. So, and C. J. R.

Sheppard, Opt. Express 15, 4634 (2007).13. C. Rivolta, Appl. Opt. 25, 2404 (1986).14. V. N. Mahajan, J. Opt. Soc. Am. A 3, 470 (1986).15. B. Zhang, J. Zerubia, and J. C. Olivo-Marin, Appl. Opt. 46,

1819 (2007).

Fig. 4. Images of HEK cells in (a) UAFþ SAF mode and in(b) F2-SAF mode.

August 15, 2011 / Vol. 36, No. 16 / OPTICS LETTERS 3053