The health of a eukaryotic cell depends on the proper functioning of its cell organelles. Characterizing these nanometer- to micrometer-scaled specialized subunits without disturbing the cell is challenging but can also provide valuable insights regarding the state of a cell. We show that objective-based scanning surface plasmon resonance microscopy can be used to analyze the refractive index of cell organelles quantitatively in a noninvasive and label-free manner with a lateral resolution at the diffraction limit.
© 2019 Optical Society of America
Surface plasmon resonance microscopy (SPRM) describes a broad field of different optical techniques as well as many different analysis approaches. It is based on the excitation of collective electron oscillations (plasmons) inside a thin metal layer, which can be, e.g., substrates for cell culture in order to study the cell–substrate interface or observe cellular dynamics. Illumination of the sample under the plasmon resonance angle can excite plasmons in the metal layer, causing a drop in the intensity of the reflected light. Analyzing the intensity as a function of the angle of incidence can provide insight into the sample composition, as the plasmon resonance angle and reflection profile strongly depend upon the environment of the metal layer.
Wide-field SPRM setups illuminate the surface under one tunable angle of incidence allowing for a qualitative observation of the sample in real time [1–13]. This technique has been shown to resolve cell adhesion sites [1,6] as well as subcellular structures , but it can also be used for specific quantitative analysis, e.g., for tracking of cell organelles in three dimensions by analyzing the intensity of the organelle appearing in the wide-field images over time . Here, the refractive index (RI) was not identified since wide-field SPRM does not allow for RI measurements.
Scanning SPRM setups focus the incident light spot-wise at the cell–substrate interface, resulting in a sample illumination with a broad spectrum of angles. The intensity profile of the reflected light is analyzed for each individual scanning point in an angle-resolved manner, allowing for a detailed quantitative analysis of the sample composition [1,5,6,13] while reaching a lateral resolution at the diffraction limit [5,6].
In previous work, we showed that the cytosolic RI can be extracted from reflectance profiles at each scanning point in an automated manner . This automated RI analysis improves the accuracy of the cell–substrate distance calculation tremendously. It was shown to be a reliable tool to quantitatively measure the temporal variations of the cytosolic RI at the cell–substrate interface of contracting cardiomyocytes. Since these measurements required capturing large numbers of frames per scanning point, the scanning resolution had to be chosen relatively low in order to measure the entire cell–substrate interface in a reasonable time frame, which in turn frustrated the resolution of cell organelles. In this work, we want to show for the first time, to the best of our knowledge, that SPRM is also suited to measure the RI of cell organelles in a label-free and noninvasive manner by increasing the density of scanning points at the interface. (The light intensity at the sample reaches at a wavelength of 632.8 nm. Around 20% is locally absorbed at the interface and dissipates to the surrounding as heat resulting in an increase of temperature directly at the interface of . Since we illuminate each scanning point for less than 10 ms at room temperature, this temperature increase is not expected to cause cell damage. Successive measurements of the same cells over a period of seven days have indeed shown no visible damage. We therefore assume the measurement to be noninvasive.)
Studying the RI of cell organelles can give important insights into their composition under physiological conditions as the parameter is closely related to the water content and the protein concentration. As the RI has been shown to indicate bacterial infections  as well as metabolistic states such as cell dormancy , the quantitative measurement of the RI is of huge interest regarding fundamental biological questions as well as label-free diagnostics.
There are several well-established techniques suitable to measure the RI of cells. In quantitative phase imaging (QPI) [16–18], the cell is illuminated with a plane wave introducing a phase shift, which depends on the cell’s RI. Analyzing this phase shift, one can determine the average RI of the cell. Even though this projection-based technique is not suited to resolve the RI distribution of complex structures, it gave valuable insights into the optical properties of the nucleus .
Optical diffraction tomography (ODT) is based on QPI and combines multiple measurements from various angles of incidence. Inverse scattering theory is applied to reconstruct the 3D RI distribution within cells  reaching a resolution of up to 166 nm . This technique has also been shown to resolve the RI changes induced by cell dormancy, compared to normal metabolic activity .
Recently, total internal reflection microscopy (TIR) has been demonstrated to resolve the RI at the cell–substrate interface of adherent cells at a lateral resolution of  by detecting the critical angle. Understanding the sample as a layer structure, the position of the critical angle is only influenced by the RI of the first and the last layer, which are penetrated by the evanescent field. Plotting the reflectance as a function of the angle, the critical angle appears as an edge in the profile and allows for the determination of the intracellular RI.
The detection of the RI in SPRM is based on a similar principle: the shape of the plasmonic reflectance curve is decisively influenced by the RI of the last layer within the evanescent field. Even though the reflectance profile is more complex compared to TIR, the edge in the reflectance profile representing the critical angle can be detected automatically and the RI determined . In contrast to TIR, scanning SPRM profits from its high spatial resolution of facilitated by the excitation of localized plasmons  and therefore allows for the resolution of cell organelles.
In this work, we used an objective-based SPRM setup, which is shown schematically in Fig. 1. It requires the use of specific samples consisting of high-index coverslips coated with a thin chromium adhesion layer () and a gold layer (), which are glued to Petri dishes and used as substrates for cell culture. These samples are placed on top of a high-NA total internal reflection fluorescence (TIRF) objective () and illuminated spot-wise using radially polarized He–Ne laser light. The incident light is focused at the glass–gold interface where it is mostly reflected, while a fraction of the light is absorbed due to the excitation of surface plasmons inside the gold layer.
Imaging the back focal plane (BFP) [see Fig. 1(c)] allows us to extract the reflectance as a function of the angle of incidence at the gold–glass interface [see Fig. 1(d)] using the Abbe sine condition . Using the transform matrix method, we model the sample as a multilayer system and fit a reflection curve to the data [1,6]. We previously showed that the position of the critical angle depends only upon the RI of the first and the last layer within the evanescent field while it is independent of the number and properties of intermediate layers. Therefore, it suffices to detect the position of , which is represented by a prominent edge in the reflectance curve [see Fig. 1(d)].
In the angle-resolved analysis, it is assumed that the glass–gold interface is illuminated with -polarized light. For objective-based scanning SPRM, this can be realized using a radially polarized laser beam illuminating the BFP . In order to analyze the data correctly, we need to take into account that the radial polarizer, which was used in these experiments (ZPol), consists of four phase plates and does therefore not provide perfect radial polarization. Considering only the light passing through the central parts of the four phase plates [see Fig. 1(c), illustrated by the white frame], the illumination of the sample can be approximated to be of the required scheme. Therefore, the reflectance curves are extracted only from these small slices of the BFP images.
In a first step of the data analysis, we determine the mean reflectance curve by averaging the curves extracted from the BFP images taken at the individual scanning points. We describe the sample as a multilayer system (see Table 1) and fit the curve to the data by varying the layer properties. After this initial fit, the individual BFP images from the measurement are analyzed separately. Therefore, all the layer properties are kept constant except for the RI of the last layer. During this procedure, the fitting area is limited to a narrow angle spectrum () around typical values found for (assuming a RI of the cytosol between 1.34 and 1.42). Since is independent of the number of intermediate layers, this method gives the RI of the last layer within the evanescent field. This method has been validated by measuring the RI of DMSO–water mixtures in different ratios corresponding to the range of .
Here, we demonstrate that this technique is also applicable to detect the RI of cell organelles in livings cells. For this purpose, we searched for cells showing pronounced structures in phase contrast and bright-field images and scanned the respective cells using the SPRM. Afterwards, we determined the RI values of the last layer within the evanescent field at each scanning point and plotted the determined values as RI profiles (see Fig. 2).
The phase contrast image of the cortical neuron in Fig. 2(a) reveals the position of its nucleus appearing as a relatively dark area in the center of the cell. A RI scan of this same cell [see Fig. 2(b)] shows the adherent areas of the cell as relatively bright areas in the scan compared to the surrounding culture media. The area corresponding to the position of the nucleus is assigned a lower RI than the surrounding cytoplasm.
We can therefore confirm the heavily debated [24–26] findings by Schürmann et al. as well as Steelman et al., which are based on quantitative phase microscopy, that the nucleus has a lower RI than the cytoplasm [17,27] by an independent method.
A second SPRM scan performed over a scanning grid with 190 nm spacing demonstrates that the technique is also capable of resolving small cell organelles. The bright-field image of the neuron shown in Fig. 2(c) reveals the positions of many small organelles represented by point-like structures, which are either brighter or darker than their surroundings. The same structures reappear in the RI scan [see Fig. 2(d)].
By extracting the data of two single scan lines [indicated by the dashed lines in Fig. 2(e)], the high resolution of the technique becomes evident. The intracellular RI along these lines shows strong variations . The two highlighted maxima in Figs. 2(f) and 2(g) show the position and RI value of two cell organelles of different sizes that have been labeled with D, B [see Fig. 2(c)]. While the maximum associated with organelle D appears to span three scanning points, organelle B shows a distinct maximum limited to one scanning point. The high resolution observed in these measurements is consistent with the measurements shown in Ref. . Here, Watanabe et al. demonstrated explicitly that objective-based scanning SPRM reaches the diffraction limit (corresponding to ) using an identical optical setup.
Generally, the RI values measured for the cytoplasm are in good agreement with the literature [17,20,27]. Sporadically, in very small areas, we find values in the RI profiles that are lower than the RI of water . This occurs if the reflection curve shows noise around the critical angle in the data of a scanning point. In such a case, the automated fitting routine underestimates the critical angle, resulting in an underestimation of the RI [see Fig. 3(b)]. This problem is likely to be solved by the use of a high-quality radial polarizer that allows for the data processing of the entire BFP image and therefore reduces the noise in the reflection curves.
In order to estimate the accuracy of the RI measurement, the reflection curve of each scanning point would have to be evaluated individually. Assuming perfect, noise-free data, the accuracy is limited by the pixel density and could potentially reach with the current setup .
For primary cortical rat neurons, RI profiles that clearly show the positions of the cell organelles as shown here (see Fig. 2) are found less frequently compared to rather homogeneous RI profiles. One possible explanation could be that the organelles are often spatially separated from the cell membrane and are therefore not penetrated by the evanescent field. The actin cortex of the cell could induce such a spatial separation, but due to the small number of studies [28,29], its thickness is still unknown for most cell types. For HeLa cells, its thickness was determined to be . Assuming a cortex thickness in the same order of magnitude for neurons, it is not surprising that the organelles are observed in few measurements as the expected thickness and penetration depth ( ) are in the same order of magnitude.
We have demonstrated that scanning SPRM can be used to analyze the RI of cell organelles at the cell–substrate interface in a label-free and noninvasive manner. The measured RI profiles have been correlated with bright field as well as phase contrast images. We could confirm the heavily debated finding that the RI of the nucleus is lower than the RI of the surrounding cytosol.
Since objective-based scanning SPRM has a great potential regarding the quantitative characterization of the cytosolic RI with a high accuracy and a lateral resolution at the diffraction limit, we believe that this technique will provide deep insights into a large variety of biological questions and also serve as a label-free marker in the diagnosis. The suitability of the technique for analyses of dynamics and optical density of cell organelles is obvious. But the technique also promises deeper insights into the metabolism of cells, such as cell dormancy, a state in which the water content of the cell is greatly reduced reflecting in an increased cytosolic RI.
We thank professor Dr. Andreas Offenhäusser and Daniel Haarhoff for fruitful discussions and proofreading of the Letter. We thank Michael Prömpers for the sample production and Bettina Breuer for the neuron isolation. H. H. developed the data acquisition software in LabVIEW and developed the data analysis software in MATLAB allowing for the local RI measurement. E. K. designed and carried out the experiments, analyzed the data, and wrote the Letter.
1. K.-F. Giebel, C. Bechinger, S. Herminghaus, M. Riedel, P. Leiderer, U. Weiland, and M. Bastmeyer, Biophys. J. 76, 509 (1999). [CrossRef]
2. B. Huang, F. Yu, and R. N. Zare, Anal. Chem. 79, 2979 (2007). [CrossRef]
3. S. Ae Kim, K. Min Byun, J. Lee, J. Hoon Kim, D.-G. Albert Kim, H. Baac, M. L. Shuler, and S. J. Kim, Opt. Lett. 33, 914 (2008). [CrossRef]
4. W. Wang, Y. Yang, S. Wang, V. J. Nagaraj, Q. Liu, J. Wu, and N. Tao, Nat. Chem. 4, 846 (2012). [CrossRef]
5. K. Watanabe, K. Matsuura, F. Kawata, K. Nagata, J. Ning, and H. Kano, Biomed. Opt. Express 3, 354 (2012). [CrossRef]
6. K. Toma, H. Kano, and A. Offenhäusser, ACS Nano 8, 12612 (2014). [CrossRef]
7. A. W. Peterson, M. Halter, A. Tona, and A. L. Plant, BMC Cell Biol. 15, 35 (2014). [CrossRef]
8. K. Syal, W. Wang, X. Shan, S. Wang, H.-Y. Chen, and N. Tao, Biosens. Bioelectron. 63, 131 (2015). [CrossRef]
9. Y. Yang, H. Yu, X. Shan, W. Wang, X. Liu, S. Wang, and N. Tao, Small 11, 2878 (2015). [CrossRef]
10. C.-T. Yang, R. Méjard, H. J. Griesser, P. O. Bagnaninchi, and B. Thierry, Anal. Chem. 87, 1456 (2015). [CrossRef]
11. F. Zhang, S. Wang, L. Yin, Y. Yang, Y. Guan, W. Wang, H. Xu, and N. Tao, Anal. Chem. 87, 9960 (2015). [CrossRef]
12. X.-W. Liu, Y. Yang, W. Wang, S. Wang, M. Gao, J. Wu, and N. Tao, Angew. Chem. Int. Ed. Engl. 56, 8855 (2017). [CrossRef]
13. E. Kreysing, H. Hassani, N. Hampe, and A. Offenhäusser, ACS Nano 12, 8934 (2018). [CrossRef]
14. A. E. Ekpenyong, S. M. Man, S. Achouri, C. E. Bryant, J. Guck, and K. J. Chalut, J. Biophoton. 6, 393 (2013). [CrossRef]
15. S. Abuhattum, K. Kim, T. M. Franzmann, A. Eßlinger, D. Midtvedt, R. Schlüßler, S. Möllmert, H.-S. Kuan, S. Alberti, V. Zaburdaev, and J. Guck, Front. Phys. 6, 388876 (2018). [CrossRef]
16. M. Schürmann, J. Scholze, P. Müller, C. J. Chan, A. E. Ekpenyong, K. J. Chalut, and J. Guck, in Methods in Cell Biology, E. K. Paluch, ed., Vol. 125 of Biophysical Methods in Cell Biology (Academic, 2015), Chap. 9, pp. 143.
17. M. Schürmann, J. Scholze, P. Müller, J. Guck, and C. J. Chan, J. Biophoton. 9, 1068 (2016). [CrossRef]
18. P. Müller, M. Schürmann, S. Girardo, G. Cojoc, and J. Guck, Opt. Express 26, 10729 (2018). [CrossRef]
19. Y. Sung, W. Choi, C. Fang-Yen, K. Badizadegan, R. R. Dasari, and M. S. Feld, Opt. Express 17, 266 (2009). [CrossRef]
20. G. Kim, M. Lee, S. Youn, E. Lee, D. Kwon, J. Shin, S. Lee, Y. S. Lee, and Y. Park, Sci. Rep. 8, 9192 (2018). [CrossRef]
21. K. P. Bohannon, R. W. Holz, and D. Axelrod, Microsc. Microanal. 23, 978 (2017). [CrossRef]
22. K. Watanabe, N. Horiguchi, and H. Kano, Appl. Opt. 46, 4985(2007). [CrossRef]
23. M. Bass, C. DeCusatis, J. M. Enoch, V. Lakshminarayanan, G. Li, C. MacDonald, and V. N. Mahajan, Handbook of Optics (McGraw-Hill, 2010).
24. M. A. Yurkin, J. Biophoton. 11, e201800033 (2018). [CrossRef]
25. P. Müller and J. Guck, J. Biophoton. 11, e201800095 (2018). [CrossRef]
26. Z. A. Steelman, W. J. Eldridge, and A. Wax, J. Biophoton. 11, e201800091 (2018). [CrossRef]
27. Z. A. Steelman, W. J. Eldridge, J. B. Weintraub, and A. Wax, J. Biophoton. 10, 1714 (2017). [CrossRef]
28. A. G. Clark, K. Dierkes, and E. K. Paluch, Biophys. J. 105, 570 (2013). [CrossRef]
29. M. P. Clausen, H. Colin-York, F. Schneider, C. Eggeling, and M. Fritzsche, J. Phys. D 50, 064002 (2017). [CrossRef]