Correlative Synchrotron Fourier Transform Infrared Spectroscopy and

Subscriber access provided by Stockholm University Library

Article

Correlative Synchrotron Fourier Transform Infrared Spectroscopy and Single Molecule Super Resolution Microscopy for the Detection of Composition and Ultrastructure Alterations in Single Cells Donna R Whelan, and Toby D M Bell ACS Chem. Biol., Just Accepted Manuscript • Publication Date (Web): 30 Sep 2015 Downloaded from http://pubs.acs.org on October 1, 2015

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

ACS Chemical Biology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Chemical Biology

Correlative Synchrotron Fourier Transform Infrared Spectroscopy and Single Molecule Super Resolution Microscopy for the Detection of Composition and Ultrastructure Alterations in Single Cells

Donna R. Whelan* & Toby D.M. Bell School of Chemistry, Monash University, Victoria, Australia, 3800. Current address: Department of Biochemistry and Molecular Pharmacology, New York University School of Medicine, New York, NY, USA, 10016 [email protected]

ACS Paragon Plus Environment

1


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 31

Abstract Single molecule localization microscopy (SMLM) and Synchrotron Fourier transform infrared (SFTIR) spectroscopy are two techniques capable of elucidating unique and valuable biological detail. SMLM provides images of the structures and distributions of targeted biomolecules at spatial resolutions up to an order of magnitude better than the diffraction limit, whereas IR spectroscopy objectively measures the holistic biochemistry of an entire sample thereby revealing any variations in overall composition. Both tools are currently applied extensively to detect cellular response to disease, chemical treatment and environmental change. Here, these two techniques have been applied correlatively at the single cell level to probe the biochemistry of common fixation methods and have detected various fixation-induced losses of biomolecular composition and cellular ultrastructure. Furthermore, by extensive honing and optimizing of fixation protocols, many fixation artifacts previously considered pervasive and regularly identified using IR spectroscopy and fluorescence techniques, have been avoided. Both paraformaldehyde and two-step glutaraldehyde fixation were identified as best preserving biochemistry for both SMLM and IR studies while other glutaraldehyde and methanol fixation protocols were demonstrated to cause significant biochemical changes and higher variability between samples. Moreover, the potential complementarity of the two techniques was strikingly demonstrated in the correlated detection of biochemical changes as well as in the detection of fixation-induced damage that was only revealed by one of the two techniques.


2

Page 3 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Single molecule localization microscopy (SMLM) is a super resolution fluorescence technique that allows spatial mapping of targeted in situ biomolecules at resolutions as good as 10 nm1. By isolating single emitters in time and fitting emission patterns to precisely localize them, SMLM research in the past decade has yielded structural and mechanistic insight into various subdiffraction structures. Recently we conducted an extensive and systematic investigation of the effects of chemical fixation on SMLM images and found sub-diffraction fixation-induced damage to the structures and distributions of biomolecules2. In particular we demonstrated that without optimization of preparation protocols and/or prior knowledge of the structure under investigation, fixation artifacts could be misleading. This is troubling because interpretation of SMLM images often relies on either visual examination or rudimentary analysis of the rendered list of molecular coordinates. Furthermore, without reference to known structures or live-cell data, there is no reassurance that the images themselves as biologically accurate.

In light of the speed with which SMLM techniques have come into widespread use, the tendency to use fixed cells, and the recent emphasis on the potential for misinterpretation of SMLM data3,

4

arising from, among other things, fixation damage5, 6, ongoing consideration of sample preparation effects on the underlying biochemistry of samples is critical. Moreover, because SMLM research often endeavours to elucidate unknown structures and thus cannot easily use known structural preservation to validate the sample preparation used, complementary methods are highly desirable. Use of live cell SMLM as a benchmark for fixed cell SMLM is precluded in most cases due to the general need for several minutes measurement to generate a single SMLM image.

Electron

microscopy (EM)7-9 and atomic force microscopy (AFM)4, 10, 11, among others12, have recently been highlighted as good correlative techniques for combination with SMLM, and are most often used to confirm observations made in SMLM or to provide structural context. However both suffer from requiring similar preparative demands as SMLM —namely a general lack of live-cell methodology


3


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 31

for direct comparison of samples before and after fixation—and are prone to failing to detect fixation artifacts 13, 14.

Fourier transform infrared (FTIR) spectroscopy is a potential complementary technique that can be applied non-invasively to live specimens and requires no significant deviation in SMLM sample preparation for correlative measurements after fixation (in contrast to sectioning or heavy-metal labelling for AFM/EM experiments)15. More importantly, IR spectroscopy yields a holistic readout of the entirety of the specimen and is capable of detecting, without bias, the biochemical composition of single cells and deviations in biochemistry as it is related to disease state16, cell lineage or fate17, and cell cycle phase18, 19.

The combination of techniques is shown to be particularly successful in the complementarity of their contrasting advantages and limitations, thus demonstrating the potential to be used correlatively in future applications. Far-field FTIR spectroscopy is an intrinsically non-invasive holistic technique but is limited to spatial resolutions of ~3-5 µm at best, although some recent 20, 21

theoretical advances

, indicate ways that this may be improved on in the future. In contrast,

SMLM can achieve spatial resolutions down to tens of nanometres but can only yield information regarding one or two specifically targeted molecules. One further interesting and perhaps telling point of difference is IR spectroscopy’s long history, continuously hindered by artifacts caused by hydration22, substrate use23 and sample morphology24, the complexity of data interpretation and the necessity for multivariate statistics15. This contrasts with the relatively recent arrival and rapid uptake of SMLM techniques with seemingly unlimited applications and thus far limited discussion of potential pitfalls and problems.

Here we have probed the biochemistry of individual cells by acquiring Synchrotron-source IR spectra of the same cells both pre- and post-fixation, after which the microtubule ultrastructure of


4

Page 5 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


these cells was imaged using the direct stochastic optical reconstruction microscopy (dSTORM) variant of SMLM25. The microtubule structure of the cell was targeted for SMLM because of its omnipresence through the cytoplasm of the cell during most phases of the cell cycle, and its polymeric nature. Thus, any observed damage to this network would likely indicate that the less robust free cytosolic molecules within the cell were also damaged, as well as other cytoskeletal structures such as actin. Immunostaining of tubulin, the monomeric form of microtubules, also allowed us to observe losses of cytosolic tubulin dimers from which we could infer overall losses of cytosolic components via membrane damage and washing. Cells were fixed with optimized protocols2 using three of the most common fixatives—paraformaldehyde (PFA), glutaraldeyde (GA) and methanol (MeOH)—and the biochemical changes produced by each of these fixatives elucidated. Aldehyde fixatives such as PFA and GA cross-link amine residues of proteins in order to preserve cellular structure by capturing other biomolecules in the meshwork of cross-linked protein26. MeOH fixation achieves sample preservation and protection via precipitation and aggregation of the biomolecules of the cell by removal of water and interruption of the native state hydrogen bonding26. While GA has been demonstrated to better cross-link proteins and optimally preserve the ultrastructure of many cytosolic biomolecules for EM experiments in the past27, PFA and MeOH have the advantage of being more straightforward to work with.

Changes to various fixation parameters including fixative concentration, incubation time, hydration state, and temperature were also investigated and considerable fixation damage identified when suboptimal methodologies were used. We further highlight the ability of IR spectroscopy to detect compositional changes caused by fixation that are not readily apparent in SMLM images of cellular ultrastructure. Such fixation damage would otherwise be almost impossible to detect when imaging an unknown structure or distribution but could nonetheless drastically affect conclusions drawn from the data. Similarly, the simplicity of interpreting SMLM images of targeted, known structures


5


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 31

such as microtubules is demonstrated to aid in understanding and relating correlated IR spectra to the holistic biochemistry.


6

Page 7 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Results Correlative S-FTIR/SMLM detects subtle fixation-induced changes to the cellular composition and ultrastructure. IR spectra and SMLM images from four single COS-7 cells are presented in Figure 1 and are representative of the full data set. IR spectra collected both before and after fixation with 3.7% PFA, two-step 0.4%/3% GA, 2% GA or -20°C MeOH are shown, as well as both overview and zoomed SMLM images of the fixed cytoskeletal microtubule (MT) architecture of the same cells. These fixatives were chosen because of their differing mechanisms and their widespread use in SMLM imaging. The protocols used here for correlative IR/SMLM were optimized for each of these fixatives in order to best-preserve the MT ultrastructure—and by inferred extension many other biomolecular structures and distributions—of the cell. The second derivative spectral region shown in Figure 1b (3000-2800 cm-1) is mostly associated with the CH2 and CH3 stretching absorptions of lipids while the region shown in Figure 1c (1300-950 cm-1) is primarily associated with backbone vibrations including νas(PO2-), νs(PO2-), ν(CO-O-C), ν(C-O-C) and ν(C-O) from nucleic acids and carbohydrates28.


7


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 31

Fig 1. IR spectra and SMLM images from samples fixed using well-optimized protocols. (a) shows the IR spectra of five live cells as well as spectra of these same single cells after fixation with 3.7% PFA, two step GA, 2% GA, and MeOH. Enhanced sections of the calculated second derivative spectra are included for (b) the CH2/CH3 stretching region, which is associated with lipids, and (c) the backbone vibration region associated with nucleic acids and carbohydrates. (d-k) show overview and zoomed in SMLM images of the MT architecture of the same single cells. Scale bars are 5 μm (d,f,h,j) and 1 μm (e,g,i,k).


8

Page 9 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


While there are small differences between the live cells due to asynchronous cell phase18, more substantial spectral changes are observed upon fixation. In the lipid region, relatively intense welldefined peaks (detected as minima in the second derivative spectra) are present in the live cell spectra

at

νas(CH3)=2962±1

cm-1,

νas(CH2)=2924±1

cm-1,

νs(CH3)=2876±2

cm-1,

and

νs(CH2)=2852±2 cm-1 (errors from variation between all spectra collected)29. These peaks broaden and lose some intensity following both two-step GA and 3.7% PFA fixation however the second derivative spectra indicate only small changes in peak position (±3 cm-1). These spectral changes are likely a result of disordering of the lipid bilayer caused by the cross-linking of the aldehydes and consequent degradation of the hydrogen-bonding network30. Fixation in 2% GA causes significantly more biochemical damage than both the optimized double GA fixation and the PFA fixation as seen in the broadening of the four aliphatic stretching absorptions to the extent that in the second derivative it is difficult to identify peak positions. This is despite little perturbation to the MT architecture in the SMLM image and the general longstanding consensus that GA is the fixative of choice for cell structure preservation, including lipid droplets and membranes31. These changes are indicative of significant disruption to, and loss of, the lipid bilayer structure itself upon reaction with the GA. This has been reported previously when methanol washing was also employed but was observed here with only the extensive PBS washing required for immunostaining30. Finally, MeOH fixation causes the most extensive perturbation of the aliphatic CH2 and CH3 absorptions demonstrating a very significant loss of membrane and lipid composition. This has previously been assumed due to the precipitative nature of MeOH and inferred from fluorescence studies of lipid droplets32 however IR spectroscopy shows the true extent of this fixation damage. Surprisingly, dehydration of the cells in situ and without further washing or fixation appears to well-preserve lipid concentration and structure with hardly any loss of intensity observed in the spectra and only a 4 cm-1 red-shift of both CH3 absorptions. This is likely demonstrative of significant changes to the hydrogen bonding within the bilayer but no change in total bilayer mass; this makes sense


9


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 31

considering no washing of the sample was done after dehydration and is in contrast to methodologies which do require post-dehydration washing29.

In the nucleic acid/carbohydrate backbone absorption region of the spectra (Fig 1c) significant spectral changes are also observed and previous research has highlighted this region as being particularly prone to fixation-induced damage33. Both the underivatized and second derivative spectra of the live COS-7 cells show the characteristic B-DNA absorptions22 namely νs(PO2-) = 1089±1 cm-1 and the splitting of the νas(PO2-) into two troughs due to the overlap of the Amide III/lipid phosphate absorptions (ν=1247±1 cm-1) and the B-DNA phosphate absorption(νas(PO2-) = 1229±2 cm-1). Upon dehydration the B- to A-DNA transition is detected via a shift of the νas(PO2-) from 1229 to 1244±2 cm-1 and a severe loss of intensity in the νs(PO2-). These absorption changes are in good agreement with other IR spectroscopic investigations of mammalian cells and confirm that the spectral features in this region can be reliably interpreted as absorption bands with strong nucleic acid contributions and relatable to both conformation and concentration34.

Upon fixation with GA (2% or two-step GA) or MeOH, the characteristic B-DNA absorptions observed in the live cell spectra shift towards absorption positions characteristic of A-like DNA or significant disordering and denaturing of the B-DNA (νas(PO2-) = 1246±2 cm-1, 1248±1 cm-1, and 1243±2 cm-1 for 2% GA, two-step GA and MeOH respectively)35. The νs(PO2-) also loses some intensity (νs(PO2-) = 1089 ±2 cm-1) in the spectra from all three fixation conditions. These spectral changes are not as severe as those induced upon sample dehydration where complete B- to A-DNA conformational change is induced, however they are indicative of a significant change in the overall DNA secondary structure. 3.7% PFA is also demonstrated to cause changes to the DNA secondary structure although not to the same extent as the GA and MeOH fixations: the νas(PO2-) shifts from 1229 cm-1 in the live cell spectrum to 1233±1 cm-1 and the loss of intensity of the νs(PO2-) is less severe. This indicates less deviation from the native-state B-DNA form in PFA fixed cells than in


10

Page 11 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


cells fixed using GA or MeOH. Previous research has shown that PFA denatures and can cleave DNA and RNA backbones36 and reduces overall solubility37, albeit on longer time scales and in tissue; these are mechanisms which might aid in better preserving holistic secondary conformation despite the primary structural damage.

Importantly, the varying degree of damage to the lipid structures as detected by IR spectroscopy is almost unobservable in the SMLM images in which the polymeric MT ultrastructure appears very well preserved. Closer consideration of these images does reveal some fixation effects, namely loss of cytosolic dimeric tubulin38 in both the simultaneously permeabilized two-step GA fixed and MeOH fixed cells. Dimeric tubulin is expected to manifest in SMLM images as single spots not associated with polymeric MTs and is often considered an unwanted artifact despite being representative of biological reality2. Loss of these dimer localizations is indicative of loss of cytosolic composition and, in this case, is due to damage to the membrane, insufficient crosslinking of free proteins in place, and the subsequent extensive washing steps required for fixation. In contrast with the two-step GA and MeOH fixed cells, both PFA and 2% GA fixation conserve cytosolic tubulin, either by cross-linking diffusing biomolecules into the meshwork of the fixed cell prior to permeabilization and washing31 or by better maintaining the lipid bilayer and preventing cytosolic loss.

To further probe the effects of fixation on IR spectra and to evaluate the variability and reproducibility of each fixation method Principal Component Analysis (PCA) was employed. By decomposing the spectral series into principal components (PCs) that describe, in decreasing magnitude, the variance between the spectra, the overall dimensionality of the data is significantly reduced allowing examination of many spectra simultaneously39. Figure 2A shows the PCA 3D Scores Plot of the spectra of 480 live cells from 20 experiments prior to fixation. No experimentbased clustering is apparent. This demonstrates, as expected, that there are no differences in the


11


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 31

spectra of single COS-7 cells based on any experimental or measurement conditions pre-fixation. Nonetheless, the spectra collected were not identical, hence the scattered distribution of the spectra in PC space. Examination of the PC1 and PC2 Loadings Plots (Figure 2D) shows the main sources for variance between these spectra to be related primarily to lipid (PC1) and nucleic acid (PC2) concentration. Once again, this is expected due to the asynchronicity of the cell cycle as well as other slight differences in cell biochemistry and has been identified previously18, 40, 41.

Fig 2(a). Principal Component Analysis (PCA) 3D Scores Plots (PC1-3) of the IR spectra of all cells prior to fixation and (b) after fixation. (c) shows the combined PCA of the live and fixed cells. Each spot represents a single spectrum with colour indicating a different experiment number (a) or fixation (b-c). (d-f) The PC1 and PC2 Loadings Plots attributed to each PCA (a-c respectively).


12

Page 13 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


In contrast to this, Figure 2B shows PCA of the spectra of the same cells after fixation and significant clustering is immediately identifiable. Along PC1, the MeOH fixed cells separate well from cells fixed using any of the aldehyde protocols while in PC2 space the GA protocols separate out well from the PFA fixation with the 2% GA fixed cell spectra clustering in more negative PC1 space than the two-step GA fixed cell spectra. The Loadings Plot for PC1 (Figure 2E) is dominated by sharp peaks in the lipid region (3000 – 2800 cm-1) with the vas(CH3)=2959 cm-1, vas(CH2)=2921 cm-1, and vs(CH2)=2852 cm-1 all attributable to the more negative PC1 space and thus indicative of higher lipid concentrations in the aldehyde fixed cells, in particular in the PFA and two-step GA fixed cells. This is well supported by the individual cell spectra.

The PC2 Loadings plot, on the other hand, is dominated by nucleic acid associated peaks, namely νas(PO2-)=1230 cm-1, νs(PO2-)=1089 cm-1 and ν(C-O)=1052 cm-1 which indicate higher concentrations of RNA/DNA or better B-DNA preservation in the spectra distributed in the negative PC2 space. While the MeOH fixed cells are distributed evenly across PC2, the GA and PFA fixation methods are better separated with PFA proving superior at nucleic acid structure and concentration preservation as is shown in the single cell spectra.

Combining both the spectra of the live cells and spectra taken after fixation, a further PCA was undertaken (Fig 2C). The Scores Plot demonstrates complete separation of the MeOH fixed cell spectra from the live cell spectra and very good separation of both the two-step and 2% GA fixed cell spectra from the live cell spectra. Notably some of the PFA fixed cell spectra appear to colocalize with the spectra of live cells indicating, at least in some cases, minimal spectral differences. The PC1 and PC2 Loadings Plots (Fig 2F) follow similar trends to those seen in the separation of different fixation protocols (Fig 2E) with lipid and nucleic acid absorptions being the main cause for variance. The variance in PC1 is demonstrated as being almost entirely due to changes in the relative lipid concentration, again with the νas(CH3), νas(CH2), and νs(CH2) the most prominent


13


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 31

spectral features. This indicates best lipid preservation, as expected, in the live cells, followed by the PFA-fixed cells and then the two-step GA, which is again seen to outperform 2% GA fixation. MeOH proves to be the most destructive to lipid content and structure. PC2 space is responsible for the separation of the DNA-preserving PFA fixation (positive PC2 space) from the DNA-degrading GA fixations (negative PC2 space) with νas(PO2-), νs(PO2-) and ν(C-O) all featuring prominently on the PC2 Loadings plot.

Interestingly, unlike in previous multivariate analysis of IR or Raman spectra of fixed cells33, 42-44, The PCA Scores plots show a large amount of overlap of the different fixative-based clusters. Use of smaller spectral regions such as the CH2/CH3 lipid vibrations or the nucleic acid associated νas(PO2-) was found to achieve better cluster separation, however because differentiation was not the objective of the research, PCA of the entire accessible spectral region was used to better demonstrate and probe the complex interplay between natural cell cycle variance and fixation effects. Overall, this PCA demonstrates that the spectral changes observed in the single cell spectra in Figure 1 are in good agreement with overall trends for the entire data set.

Sub-optimal fixation manifests more severe sub-diffraction SMLM and S-FTIR artifacts. Figure 3 shows the spectra and SMLM images of single cells fixed using fixation protocols deliberately altered to induce sub-diffraction fixation damage. It is worth noting that the zoomed out SMLM images presented are not immediately identifiable as depicting significant fixation damage, particularly in the cases of under and over-fixation (Fig 3D and F) and RT MeOH (Fig 3J) and it is only by zooming in that extensive, but oftentimes sub-diffraction, damage to the MT structures is detectable. In the case of both the 2% and 8% PFA fixed cells (Fig 3D-G), the polymeric structures appear to be either broken up or with much of the epitope sufficiently damaged to prevent immunostaining. 2% PFA fixation for insufficient time results in lengths (>50 nm) of preserved MTs punctuated by lengths of missing MTs, whereas the 8% PFA fixation leaves even smaller


14

Page 15 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


lengths (>20 nm) of MTs with some regions seemingly only populated by single localizations indicating large-scale destruction of the polymeric structure or the antibody epitope. In agreement with the SMLM depiction of widespread damage, the IR spectrum of the 8% PFA fixed cell shows an overall significant loss of cellular composition with broadening of both the Amide II and the nucleic acid backbone absorptions indicating severe denaturing and primary structure breakdown of both of these components. In contrast, and perhaps surprisingly, the IR spectrum of the under-fixed cell demonstrates the best preservation of the cellular composition and is comparable with the wellfixed 3.7% PFA fixed cell spectrum shown in Figure 1. Peaks remain sharp and almost as intense as in the live cells. This demonstrates a fixation artifact that is easily detected using SMLM through the targeting of a known structure but that would likely be missed using only IR.


15


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 31

Fig 3. IR spectra and SMLM images from samples fixed using sub-optimal protocols. (a) shows the IR spectra of four live cells as well as spectra of these same single cells after fixation with 2% PFA for 6 minutes, 8% PFA for 12 minutes, with 3.7% PFA after sample dehydration, and RT MeOH. Second derivative spectra are included for (b) the CH2/CH3 stretching region and (c) the fingerprint region, which is associated with proteins, nucleic acids and carbohydrates. (d-k) show overview and zoomed in SMLM images of the MT architecture of the same single cells. Scale bars are 5 μm (d,f,h,j) and 1 μm (e,g,i,k). ACS Paragon Plus Environment

16

Page 17 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


The 3.7% PFA fixed cell that was allowed to dehydrate briefly prior to fixation (Fig 3h-i) resulted in a less intense IR spectrum similar to the 8% PFA fixed cell and in contrast to the cell dehydrated but unfixed (Figure 1) which maintained the majority of cellular composition. Loss of intensity here is again due to removal of cellular content during washing steps via the compromised cell membrane. The SMLM image demonstrates the extent of damage to the MT network of the cell with much of the polymeric and dimeric localizations replaced by unstructured large clusters.

RT MeOH fixation (Fig 3j-k), in contrast to the -20°C MeOH fixation (Fig 1) also caused significant damage to the MT structure, introducing non-native curvature to the MTs as well as causing an increase of the imaged MT width, possibly due to destruction of the polymer and displacement of tubulin. Similar to the 8% PFA fixed cell, the IR spectrum of the RT MeOH cell shows a large loss of the biomolecular composition when compared with the live cell spectra. This is a direct result of the fixative causing damage to the cell membrane, holistic ultrastructure and distribution, and then subsequent washing steps removing the precipitated cellular content. Conclusively, momentary dehydration, RT MEOH fixation, and PFA over-fixation all cause widespread damage to cells that is exacerbated by the washing, blocking, and immunostaining steps which follow.

Detected spectral differences between single cells within a sample and in replicates are due to both variability in fixation and cell cycle. The sample damage observed in response to deliberate dehydration for 20 seconds prior to fixation using both SMLM and IR spectroscopy (Fig 3) was further probed by fully aspirating the medium off a sample prior to fixation thus resulting in only momentary, partial dehydration prior to fixation. This resulted in extremely different sub-populations of cells as shown in representative SMLM images (Figure 4A and B) and IR spectra. The MT architecture is well preserved in the PFA fixed cell shown in Figure 4A and in good agreement with SMLM images of cells fixed without any


17


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 31

dehydration damage, thus implying that this particular cell was in an area of the sample that did not dehydrate to the extent required to induce significant damage. The IR spectrum also reflects good cell preservation with only small shifts in some of the absorptions, mainly the DNA conformation markers. However the MT architecture of the cell shown in Figure 4B is drastically altered with little evidence of the native polymeric structure remaining. The IR spectrum also reflects this: while the live cell spectrum of this cell is almost identical to the live cell spectrum of the cell shown in Figure 4A, the fixed cell spectrum of the cell in Figure 4B has lost significant intensity in many regions, in particular the nucleic acid and carbohydrate backbone region as was depicted in Figure 3 in response to dehydration of the sample. As discussed, this is indicative of complete degradation of the B-DNA structure into disordered and A-like conformations as well as a general loss of cell composition. Both the IR and SMLM clearly show these fixation effects and highlight the possibility of large differences between single cells within a sample due to fixation errors. The detection of these artifacts, however, is highly dependent on external comparison. In the case of IR, the spectrum itself is only clearly artifact-affected because it is being directly compared to live-cell spectra and spectra of cells fixed using superior protocols; the SMLM images are only demonstrably artifactual because the structure of MTs is well known. In combination with small unavoidable differences between replicates (SI Figure 1), such variations must be considered.


18

Page 19 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Fig. 4A-C. SMLM images and IR spectra of two single cells from the same experimental preparation prior to, and post PFA fixation following partial sample dehydration for 20 seconds. The IR spectra of the two cells pre-fixation (solid traces, black trace is the cell shown in A, red trace is the cell shown in B) are almost identical whereas after fixation (dashed traces) the spectra are quite different. (D-F) SMLM images and IR spectra of two single cells from the same experimental preparation prior to, and post two-step GA fixation. The IR depicts significant differences between the two cells pre-fixation (solid traces, black trace is the cell shown in D, red trace is the cell shown in F). Similar variations in the IR spectra are preserved after fixation (dashed traces). Scale bars all = 2 μm.

Variability within a sample of 30-40 cells fixed in situ on a single CaF2 window is also detected in the PCA (Figure 2) in that no two cell spectra are identical and are, indeed, spread in PC space. However these spectral variations are on a much smaller scale than those caused by sample dehydration prior to fixation (Fig 4A-C) and is primarily due to the asynchronicty of the cells in culture18, 41. Such variations are difficult to identify because the changes induced by fixation in the ACS Paragon Plus Environment

19


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 31

relative lipid and nucleic acid concentration as highlighted in Figures 1 and 2 are very similar to the detected changes to biochemistry based on cell cycle. Figure 4 shows the results from two single cells from the same experiment fixed using two-step GA. The SMLM images show well-preserved polymeric MT structures with little dimeric tubulin present and, based on these images, the cells would be considered comparable. However the second derivative IR spectra of the cells before fixation (solid traces) demonstrate significant differences. Based on previous work18 one cell (black trace) is likely only a few hours pre- or post-mitosis and thus has a very high relative concentration of DNA and phospholipid as detected in the splitting of the νas(PO2-)=1224 cm-1 and 1239 cm-1 and the intensity of the ester carbonyl ν(C=O)=1742 cm-1 and RNA ribose ring vibration ν=1125 cm-1 45

. The other cell (red trace) is likely late G1/early-S phase and has a relatively high

glycogen/carbohydrate concentration as indicated by the intense ν(C-O)=1050 cm-1 and the lack of νas(PO2-) splitting.

Surprisingly many of these differences are maintained upon fixation indicating that the previously demonstrated sensitivity of live cell IR spectroscopy to two-hour differences in cell phase is also possible in fixed samples18. While the nucleic acid absorptions lose intensity and the splitting of the νas(PO2-) is lessened upon fixation, a distinctive difference in this second derivative spectra of the two cells is evident with the late G1 phase cell νas(PO2-) significantly blue-shifted from the M-phase cell (1245 cm-1 and 1231 cm-1 respectively). Similarly the differences between cell spectra intensities of the ester ν(C=O) (1742 cm-1), DNA ν(C=O) (1718 cm-1), lipid associated δs(CH2) (1464 cm-1) and the fatty acid/protein associated ν(COO-) (1401 cm-1)46 are also maintained after fixation.


20

Page 21 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Discussion Of the fixation protocols investigated 3.7% PFA fixation is demonstrably the more straightforward and predictable method for holistic compositional preservation as well as conservation of the ultrastructure of MTs in the cytoplasm. This is in good agreement with previous IR studies28 as well as research demonstrating good preservation for many antigens for immunolabeling47. GA fixation, in particular the optimized two-step GA approach, was again shown to preserve MT architecture even better than PFA. This is in line with EM and SMLM studies that have investigated the effects of PFA and GA fixation on actin, another cytoskeletal structure, and also found that PFA, but not GA, introduces sub-diffraction artifacts48. Permeabilization was observed to strongly influence the ratio of polymeric to dimeric tubulin localizations and demands further investigation combining SMLM and IR with assays for cytosolic content post-fixation. In conjunction with the superior cytoskeletal preservation observed in SMLM images of GA fixed cells, IR spectroscopy also highlighted the potential for two-step GA fixation to preserve many aspects of cellular composition just as well as PFA.

Furthermore, we have demonstrated that even small differences in application of the same fixative can change the extent of sample preservation. This was emphasised with both the different GA methodologies and consequently different IR and SMLM images, and demonstrates the validity and importance of the ongoing debate regarding the structure and mechanism of GA fixation49. Similarly, fixation with -20°C MeOH and RT MeOH gave drastically different images and spectra.

Interestingly, previous work37, 50 has emphasised the combination of cross-linking and hydrolysis cleavage of both PFA bridges and the DNA/RNA backbone and cautioned against use of PFA in experiments that require high fidelity maintenance of DNA or RNA. No evidence of a significant amount of these effects was observed in the IR or SMLM data. On the contrary, PFA was primarily identified as superior to GA in maintaining some degree of B-like DNA conformation upon


21


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 22 of 31

fixation. It is possible that previous work has exposed samples to PFA for longer time periods that have exacerbated the extent of hydrolysis reactions leading to large-scale damage. Due to time constraints IR spectra were taken of the fixed cells within a few hours of the fixation process and so the full extent of any slow or reversing reactions may not have occurred. Nonetheless, in the 8% PFA over-fixed sample significantly more nucleic acid damage is detected in the IR spectrum than in the 3.7% fixed sample. It is also possible that direct damage to nucleic acids, even on a relatively large scale, is too subtle an event to be detected using IR spectroscopy, particularly in amongst more significant changes to the overall composition of the cell. Perhaps also indicative of our fixation methodologies being overall less harsh than those used by others is that no spectral signatures of the fixatives themselves were detected in either the spectra or the PCA as has been observed before51.

The most extreme fixation effects were reflected in both the SMLM and IR of cells purposefully fixed with substandard protocols. In particular, extensive damage to cell membranes by dehydration, MeOH or strong cross-linking fixation was observed to cause extensive cytosolic losses as well as potentially further losses from the nucleus or organelles depending on the extent of membrane damage. This has been reported previously6 and would explain the excessive losses of overall IR spectrum intensity for these fixations. The presence of these artifacts and the clear potential for small changes in fixation methodology to cause them in both the SMLM images and the IR spectra emphasises the importance of optimization and standardization of fixation techniques. Without live-cell or well-fixed reference spectra the IR spectra of MeOH and 8% PFA fixed cells would not immediately be identifiable as erroneous despite the cells under investigation being significantly damaged by the fixation. Previously it has been argued that fixation induced effects are inherently reproduced sample-wide and are not significant enough to confound diseasestate changes or other differences being examined43,

44

. However, we have shown here that

differences between experiments arise even when the same optimized fixation protocol is carried


22

Page 23 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


out by the same individual, on the same day, in the same lab, thus demonstrating the irreproducibility of fixation due to the sheer number of variables at play. While these differences are indeed relatively small and would often be less than those caused by disease, environment, etc., they limit the sensitivity of such experiments in the same manner as the within-sample variation of unsynchronised samples imposed by the cell cycle. More importantly, the various effects of different fixation protocols are shown here to be of a larger magnitude than cell cycle and replication associated variability and thus further limiting to the sensitivity of IR studies that do not carefully maintain fixation procedure across all samples. Clearly, the interplay between the actual biochemistry detected by IR spectroscopy and the confounding cooperative effects of morphology, fixation and cell cycle must always be considered.

Similarly, the potential for artifacts in SMLM imaging necessitates careful consideration when using the technique and the ability to correlate with holistic IR spectroscopic measurements holds promise. This is particularly important because even when targeting a well-known cytosolic architecture such as MTs many types of damage to the overall cell are undetectable as was highlighted here by the persistent detection of fixation effects via holistic IR spectroscopy, even when using optimized protocols. These effects are not important when considering the MT architecture but must be appreciated when examining other structures, especially those that are not well understood prior to SMLM imaging.

In light of the pervasive and difficult to detect artifacts detailed herein, the correlation of these techniques offers a means to detect and avoid many fixation artifacts. Correlative IR/SMLM studies are expected to become significantly more approachable with new benchtop light-sources that are brighter than traditional Globars and thus deliver better signal-to-noise faster allowing for nonsynchrotron live-cell experiments52, 53. Furthermore, robust water absorption correction54 will allow for examination of the Amide I band and better elucidation of changes to protein composition.


23


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 24 of 31

Similarly, use of a substrate that can be used for both fluorescence and IR measurements will allow for better SMLM imaging conditions, localization precisions and spatial resolutions, as well as open the door for 3D-SMLM experiments.

By probing the effects of several fixatives and fixation parameters using two very different techniques we have demonstrated that in all cases, fixation causes detectable changes to the biochemistry of cells. That this occurs is hardly unexpected however artifacts have a long history of being overlooked. Thus we emphasize the importance and potential problems associated with preparative artifacts and highlight IR spectroscopy and SMLM imaging as correlative complementary methods which can be used to investigate, monitor and minimize preparative artifacts. Moreover, we demonstrate that the combination of these techniques could further allow investigation of complex biochemical changes as they relate to various other research.


24

Page 25 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Methods Cell Culture COS-7 cells were routinely cultured at 37°C in 5% CO2 in Dulbecco’s modified Eagle’s medium (Sigma). Cells were seeded to a final confluency of

Correlative Synchrotron Fourier Transform Infrared Spectroscopy and

Recommend Documents