Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Jan 19.
Published in final edited form as: J Phys Chem B. 2021 Aug 15;125(33):9517–9525. doi: 10.1021/acs.jpcb.1c05554

Application of 2D IR Bioimaging: Hyperspectral Images of Formalin-Fixed Pancreatic Tissues and Observation of Slow Protein Degradation

Sidney S Dicke 1, Ariel M Alperstein 2, Kathryn L Schueler 3, Donald S Stapleton 4, Shane P Simonett 5, Caitlyn R Fields 6, Farzaneh Chalyavi 7, Mark P Keller 8, Alan D Attie 9, Martin T Zanni 10
PMCID: PMC8769495  NIHMSID: NIHMS1745273  PMID: 34396779

Abstract

We used two-dimensional IR bioimaging to study the structural heterogeneity of formalin-fixed mouse pancreas. Images were generated from the hyperspectral data sets by plotting quantities associated with the amide I vibrational mode, which is created by the backbone carbonyl stretch. Images that measure the fundamental vibrational frequencies, cross peaks, and anharmonic shifts are presented. Histograms are generated for each quantity, providing averaged values and distributions around the mean that serve as metrics for protein structures. Images were generated from tissue that had been stored in a formalin fixation for 3, 8, and 48 weeks. Over this period, all three metrics show that that the β-sheet content of the samples increased, consistent with protein aggregation. Our results indicate that formalin fixation does not entirely arrest the degradation of a protein structure in pancreas tissue.

Graphical Abstract

graphic file with name nihms-1745273-f0001.jpg

INTRODUCTION

Many techniques exist for identifying and solving the structures of proteins in vitro, but it is extremely difficult to monitor and image the structures of proteins in tissues. There are imaging techniques such as antibodies, fluorescent dyes, and mass spectrometry from which structure is inferred.1-4 Electron scanning microscopy provides morphologies.5,6 There are also spectroscopies, like NMR and magnetic resonance imaging (MRI), that provide structural information but are difficult to apply on cellular length scales.7-9 Most related to the topic of this manuscript are Raman and infrared (Fourier transform infrared (FTIR)) imaging, which are techniques frequently used on tissues.10-18 They provide secondary structure information with spatial resolutions on nanometer to micrometer scales. They can be used in conjunction with vibrational dyes19-21 or applied label free.22 Most often, vibrational spectroscopies utilize the amide vibrational modes to monitor the secondary structure of proteins, along with the phosphodiester and methyl modes for DNA and lipids.14,23-25 FTIR imaging can distinguish tumorous brain regions,26,27 lesions in the aorta of cholesterol-fed rabbits,28 and bone mineralization in wild-type and density matrix protein-1 (DMP1) knockout mice.29 FTIR imaging is being developed for the analysis of tissue biopsies,25,30,31 among many other applications.13,27,32,33

Raman or IR images can be generated at a single well-defined frequency or hyperspectral data sets generated by measuring multiple frequencies or in a Fourier transform mode.13,26,34-36 A more recent hyperspectral vibrational imaging technique is two-dimensional (2D) IR microscopy, which is based on ultrafast 2D IR spectroscopy.37-40 2D IR spectroscopy uses a series of femtosecond pulses to generate multidimensional IR spectra that correlate vibrational modes through the physics of their vibrational couplings.41-45 2D IR spectroscopy is different than the analytical 2D correlation spectroscopy,46-48 which is a means of analyzing data sets. 2D IR spectroscopy provides information about protein environments through 2D line shapes, utilizes vibrational lifetimes to discriminate against solvent exposed and disordered proteins, and enhances resolution through off-diagonal cross peaks.44,49,50 2D IR microscopy can be implemented in a number of ways, including point-mapping37,39,51 and widefield approaches,38 but either way, 2D IR microscopy is a collection of 2D IR spectra that form a hyperspectral image. 2D IR microscopy is a new technique that has mostly been applied to model systems.37,39,40,52 The only biological systems studied so far are mouse, porcine, and human lens tissues and mouse kidney sections (stained with an IR active probe).39,40,52

In this work, we image the formalin-fixed tissue of mouse pancreas. We present images of diagonal peaks, cross peaks, and anharmonic shifts and use these metrics as probes of the protein structures within the sample. We also present histograms for each image displaying the range of values for each metric present in the sample. Interestingly, we identified the appearance of features within the pancreatic tissue that are typical of protein aggregation and found that these aggregates accumulate with time, even though the tissues are formalin-fixed.

METHODS

Housing and Animal Maintenance.

All studies performed on the mouse pancreas samples were preapproved by the Animal Care and Use Committees at UW-Madison (Protocol #A005821) and BYU (Protocol #17-1202). The pancreas samples analyzed here were exhumed from wild-type Friend Virus B/NIH Jackson (FVB/NJ) mice obtained from Jackson Laboratory (JAX stock no. 001800). FVB/NJ mice are a widely used model organism for transgenic insertions. The mice studied here were raised in a colony where they lived with 2–5 mice per cage in a temperature and humidity-controlled room with a 12 h light/dark cycle (6:00 am to 6:00 pm). After they were weaned, the mice were on a chow diet (Purina 5008) and sacrificed at 17 weeks of age.

Formalin-Fixed Pancreas Section.

Three mice were studied. A single wild-type mouse pancreas is presented in the main text, and similar results were observed for two additional wild-type mouse pancreas samples presented in Figure S1. The mice were euthanized via CO2 inhalation, and pancreas samples were extracted as described previously53 and immediately immersed in a 10% formalin in a phosphate-buffered saline (PBS) solution. The tissues were then incubated at 4 °C in the dark for 24 h (the sample is now considered formalin-fixed). Formalin-fixed samples were then placed in an aqueous solution, and the tissues were subjected to a progressive dehydration in an ethanol gradient prior to two final washes in 100% xylene solutions. Finally, dehydrated tissues were embedded in paraffin wax at 60 °C, shaped into a block format, and cooled to room temperature for slicing. The tissues are now described as “formalin-fixed and paraffin-embedded” (FFPE). The paraffin embedding was done by the veterinary medicine histology and the experimental animal pathology laboratories at UW-Madison. The pancreas removal and fixation, performed in our laboratory, takes 24 h, and the paraffin wax embedding process has an approximately two-week turnaround time; thus, after any routine realignments and calibrations of our optical setup we label our earliest time point as three weeks from when the sample was initially soaked in 10% formalin. FFPE tissues were stored in sealed plastic bags in the dark at room temperature between measurements. For the antibody staining, paraffin needs to be removed from the tissue slices. In our study, paraffin was not removed, because paraffin does not absorb in the amide-I region (as confirmed by measurements in regions of the slices that contained no tissue), and we wanted the tissue to be stable at room temperature for the duration of the experiment.

2D IR Bioimaging.

FTIR spectra of formalin in D2O and 10% formalin in PBS buffer are shown in the Figure S2. Ten micrometer slices of FFPE mouse pancreas tissue were mounted between two CaF2 windows. 2D IR spectra were collected as described previously.39,52 The sample was raster scanned in steps of 50 μm to cover an area of 1–5 mm with 31–91 points per axis. A dark blue pixel indicates either there is no tissue at that location or there is an excessive scatter of laser light that prevents that point from being measured. Spectra for these pixels are not included in the scatterplots nor histograms below.

RESULTS

2D IR bioimaging is a method of creating hyperspectral images from thousands of individual 2D IR spectra. 2D IR spectroscopy is analogous to NMR spectroscopy, except that it uses pulses of infrared light to measure vibrations rather than pulsed magnetic fields for nuclear spins. Images generated from 2D IR spectra use the amide I vibrational mode, which is dominated by the carbonyl stretch motions of the peptide backbone. The frequency of the amide I band depends on the protein secondary structure. A slice along the diagonal of the 2D IR plot is analogous to that of an FTIR spectrum, albeit with more well-defined peaks.54 Cross peaks appear off of the diagonal when vibrational modes are coupled to one another. More detailed information about the use of 2D IR spectroscopy to probe a protein secondary structure is covered in several recent reviews.41-43

In what follows, we show 2D IR bioimages collected on the same mouse pancreas sample at 3, 8, and 48 weeks after fixation. At each time point, we plot three different images created by analyzing the 2D IR spectra in three unique ways. The images visualize the frequencies, cross peaks, and anharmonicities, which are measures of the protein structure. We note that each of the 2D IR images have a spatial resolution of ~100 μm2 and contain signals from all proteins within the overlapping laser beams. The 2D IR images cannot identify single proteins in the tissue, for example, but are instead providing an assessment of global populations of a secondary structure. Spatial resolutions down to the diffraction limit of ~3–4 μm are possible.

Figure 1A shows a 2D IR spectrum of one location (~100 μm2) within a mouse pancreas measured three weeks following a formalin fixation. The 2D IR spectrum in Figure 1A is representative of the majority of locations measured from this sample. The main features are contained within black boxes. The fundamental vibrational modes (v = 0 to v = 1) appear along the diagonal, analogous to the peaks in an FTIR spectrum, at ωpump = ωprobe = 1635 cm−1 and ωpump = ωprobe = 1660 cm−1 (peaks 1 and 2). Each fundamental mode is accompanied by a 180° out-of-phase peak offset from the fundamental, which are the overtone peaks (v = 1 to v = 2). Peak 3 is the overtone feature for β-sheet vibrations. Peak 4 is the overtone feature for α-helix and random coil vibrations. In Figure 1B, a diagonal cut through the overtones is plotted that can be interpreted similarly to an FTIR spectrum, knowing that the intensities scale with the square instead of linearly with the absorption coefficient.39,54,55 We cut through the overtone because the tissue scatters laser light, which causes a background that interferes with the fundamental transition but not the overtone. The diagonal cut easily illustrates the relative intensities of the peaks at ωpump = 1635 and 1660 cm−1, which correspond to the peaks in boxes 1 and 2, respectively. In lens (unpublished result), liver, and heart tissues, formalin fixation has previously been reported to shift the amide I vibrational frequencies by ~4 cm−1 to higher values,23 because it is more nonpolar than water (several control spectra of preservation chemicals are given in Figure S2: formalin in D2O, formalin in PBS-D2O buffer, and formalin in PBS-H2O buffer). We find that subtracting 4 cm−1 from the two features in boxes 1 and 2 puts their frequencies at ~1631 and 1656 cm−1, which are typical frequencies measured in vitro for β-sheet and random coil secondary structures, respectively, consistent with the previously reported 4 cm−1 shift upon fixation.23

Figure 1.

Figure 1.

Open in a new tab

Hyperspectral images created using the intensities of the overtone absorptions. (A) Representative 2D IR spectrum. Peaks labeled by boxes 1 and 2 result from the v = 0 to v = 1 excitation and bleach for β-sheet (ωpump = ωprobe = 1635 cm−1) and random coil structures, respectively. Peaks labeled by boxes 3 and 4 result from the v = 1 to v = 2 excited state absorptions. (B) A slice through the overtone features taken from the 2D spectrum (A, thin line) with peaks 3 and 4 labeled. Hyperspectral images created by taking the ratio of the maximum intensity in boxes 3 and 4 for an image measured at (C) 3 weeks, (D) 8 weeks, and (E) 48 weeks after the formalin fixation. Pixels are spaced 50 μm apart. Scale bars represent 1.06, 0.5, and 0.25 mm, for total areas of 4.25, 3.1, and 1.6 mm2, respectively. (F–H) Histograms of the respective images.

Figure 1C-E shows images generated by plotting the ratio of the intensity at the maximum of peaks 3 and 4. Figure 1C-E were collected 3, 8, and 48 weeks following the formalin fixation, respectively (images are from the same slice of pancreas, but not the same position within the tissue. See Methods.). Figure 1F-H bins the values of each image into a histogram, again representing time points of 3, 8, and 48 weeks after the formalin fixation (histograms displayed below image containing the same data values). The average value in Figure 1C is 1.5, indicating that most positions have a β-sheet peak that is 50% larger than the random coil/α-helix peak, similar to the chosen representative spectra (Figure 1A,B). Some regions have much lower β-sheet content with a few locations having a very high percentage of β-sheet proteins. The tissue contains a variety of cell types and biological structures (hematoxylin and eosin stained section available in Figure S3) including acinar cells, α, β, and γ cells of the islets, blood vessels, and secretory ducts. In the images presented here, we are not scanning at a spatial resolution that can distinguish between cell types, but the heterogeneity of the images suggests that cell types might be distinguished with a higher resolution. The images collected at 8 and 48 weeks after the fixation and slicing, along with their respective histograms, reveal that the ratio of the intensities becomes larger with age, indicating that the amount of β-sheet secondary structure relative to the random coil/α-helix feature is more prominent. The average ratio is ~4 at 48 weeks with a larger distribution of values, indicating that the tissue is becoming more structurally heterogeneous. Each pixel in these images represents a measurement that encompasses ~100 μm2, and each pixel is spaced by 50 μm. Some of the images are more “pixelated” because a smaller area was measured and thus contains fewer individual measurements.

Figure 2A shows the same 2D IR spectrum as that of Figure 1A, but with the cross-peak region marked by a box, where the intensity was magnified by a factor of 3. Figure 2B was processed in the same manner as Figure 2A, but it is a spectrum from the tissue measured at eight weeks after the formalin fixation. Figure 2C is a close-up view of the cross-peak region that has the two spectra overlaid, thereby comparing the cross peaks at three and eight weeks after the fixation (black and red contours, respectively). The cross peaks appear at different frequencies in the two spectra. In the three-week sample, the cross peak is at ωpump = 1635 cm−1, ωprobe = 1697 cm−1 (peak 5). The eight-week sample has its cross peak at ωpump = 1629 cm−1 and ωprobe = 1700 cm−1 (peak 6).39 The cross peaks correlate coupled vibrational modes, indicating that the β-sheet in the eight-week sample has a 6 cm−1 lower frequency than at three weeks.

Figure 2.

Figure 2.

Open in a new tab

Hyperspectral images created from cross-peak intensities. 2D IR spectrum collected from the mouse pancreas (A) three weeks (peak 5) and (B) eight weeks (peak 6) after the formalin fixation with an arrow indicating the cross peak, scaled by a factor of 3. (C) An overlay of the two cross peaks. Peak 5 is displayed in black, and peak 6 is displayed in red. (D) A tissue image created by taking the ratio of the intensity at the peak 5 position compared to the peak 6 position for the three-week-old pancreas tissue. (E) The same ratio values in (D) represented by a histogram. (F) A tissue image created by taking the ratio of the intensity at the peak 5 position compared to the peak 6 position for the eight-week-old pancreas tissue. (G) The same ratio values in (F) represented by a histogram. Scale bars represent 1.06 and 0.5 mm, for total areas of 4.25 and 3.1 mm2 in (D, F), respectively.

The two cross-peak spectra in Figure 2C are representative spectra of the cross peaks in the three- and eight-week samples. Of the thousands of 2D IR spectra collected in these images, some have peak 5, some have peak 6, and many contain both. To visualize the heterogeneity, we show images next to their corresponding histograms in Figures 2D (image), 2E (corresponding histogram), 2F (image), and 2G (corresponding histogram). Because the cross peaks overlap with one another, we do not plot their absolute intensities but use the ratio between the two peaks instead. For the three-week tissue, the average ratio is ~0.5. A value of 0.5 means that the cross peak is solely created by peak 5. A value of 1 means that there are equal contributions from the peaks 5 and 6. For the eight-week tissue, there are many locations where the ratio is 1 or larger, indicating that the tissue at that position is dominated by peak 6.

We also analyzed the anharmonicities of the β-sheet and random coil/α-helix peaks. Figure 3A shows the same 2D IR spectrum from the three-week-old pancreas as in the previous figures, but it is marked with a thin horizontal cut that passes through peaks 1 and 3. The intensity along the horizontal cut is displayed in Figure 3B. The fundamental transition has a positive intensity, and the overtone is negative. To extract the anharmonic shift, we fit each cut to two Gaussians (dashed lines) whose frequency and widths are varied to best reproduce the spectrum (dashed black line). Details of the fitting method are given in Equation S1. The frequency difference is the anharmonic shift. Fits were performed individually for each 2D IR spectrum measured, creating the images and corresponding histograms shown in Figures 3C (image) and 3D (corresponding histogram) for the three-week tissue and Figures 3E (image) and 3F (corresponding histogram) for the eight-week tissue.39 A similar procedure was performed along the thin line passing through peaks 2 and 4 (Figure 3A) to produce the images and histograms for the anharmonic shift of the random coil/α-helix peaks. For the three-week tissue, Figure 3G is the image and 3H is the corresponding histogram, and for the eight-week tissue, 3I is the image and 3J is the corresponding histogram.

Figure 3.

Figure 3.

Open in a new tab

Hyperspectral images created from anharmonic shifts. (A) A 2D IR spectrum of pancreas tissue three weeks after a fixation with two cuts displayed (black lines) at 1635 and 1660 cm−1, respectively. The horizontal cuts pass through an overtone and a fundamental vibration at their respective pump frequencies. (B) Slice at 1635 cm−1. Positive and negative Gaussian fits (dashed red and blue lines, respectively) and the sum of the two fits (dashed black line) fit to the β-sheet pump slice from the spectrum displayed in (A). (C) is an image created from the anharmonic shift value of the β-sheet cut taken in 50 μm steps across a mouse pancreas three weeks after fixation, and (D) is the corresponding histogram representing these values. (E) Image created from the anharmonic shift value of the β-sheet cut taken in 50 μm steps across a mouse pancreas eight weeks after fixation; (F) is the corresponding histogram representing these values. (G) Image created from the anharmonic shift value of the random coil/α-helix cut taken in 50 μm steps across a mouse pancreas three weeks after fixation; (H) is the corresponding histogram representing these values. (I) Image created from the anharmonic shift value of the random coil/α-helix cut taken in 50 μm steps across a mouse pancreas eight weeks after fixation; (J) is the corresponding histogram representing these values. (Histogram axis held between 15 and 25 for consistency with the images.)

The distribution of values is narrow in the three-week tissue, with an average value of 19 cm−1 for the β-sheet mode and 22 cm−1 for the random coil/α-helix mode. The anharmonic shift depends on the extent that the vibrational modes are delocalized; the larger the delocalization, the smaller the anharmonic shift. Typically, β-sheets have vibrational modes delocalized over 5–15 amide bonds,56-58 whereas the vibrational modes of α-helix are approximately five amide bonds, and random coils are delocalized over a maximum of two amino acids.54 The images and histograms presented here are consistent with those physics, giving us confidence in the consistency of the fits. We do note that these values are larger than the anharmonic shift of a single amide group, which is between 12 and 15 cm−1,59,60 which we attribute to the fact that these tissues contain many proteins, and so the extracted values are highly averaged quantities.

For the eight-week tissue, the average anharmonic shift of the β-sheet mode is 18 cm−1, and the random coil/α-helix mode is 24 cm−1. Thus, the average β-sheet anharmonic shift is 2 cm−1 smaller at eight than three weeks, and the average random coil/α-helix mode is 2 cm−1 larger. A smaller anharmonic shift indicates a larger delocalization, consistent with the formation of larger β-sheets. The larger anharmonic shift is consistent with a higher population of random coil structures. The two observations taken together, larger β-sheets and more random coil, is consistent with a protein aggregation. The distributions of the anharmonic shift at eight weeks is also larger than at three weeks, indicating a greater degree of variation in the secondary structure, as would be expected for an aggregation.

Images of the cross peaks and anharmonic shifts are not shown for the 48-week tissue because excessive laser scatter prevented an accurate analysis of the cross peaks and anharmonic shifts. An analysis of the overtone peaks was still possible because those features are large and lie away from the scatter. We found, in general, that older tissues scattered the laser light more strongly than younger tissues. That qualitative observation is also consistent with a protein aggregation; in several studies on cataract tissues, laser scatter correlated with aggregation.39,52

DISCUSSION

The above images spatially map three quantities that reflect protein secondary structure: frequencies, cross peaks, and anharmonic shifts. Data collected on tissues 3, 8, and 48 weeks after a formalin fixation show that all three quantities change, indicating that the protein structures are slowly altered within the tissue over the course of the year. The 8 and 48 week images are more pixelated but still contain over 950 individual spectra. Moreover, multiple locations across the pancreas were measured, and the observations were confirmed in two other mice (see the Supporting Information). Thus, we conclude that degradation occurs uniformly throughout the tissue.

The peak associated with β-sheets becomes larger, the lower-frequency cross peak gains intensity, the anharmonic shift in the β-sheet region becomes smaller, and the anharmonic shift of the random coil/α-helix region becomes larger. These observations are consistent with a decrease in a native β-sheet and α-helical structure and the increase in random coil structures and non-native β-sheets. We know that it is non-native β-sheet structures, because the cross peak that forms is at a lower frequency. We also see an increased laser scatter in aged samples, indicating that the tissues are more spatially heterogeneous.

From these observations, we conclude that the proteins within the tissue are slowly aggregating. Amyloid fibrils are a common type of aggregate, but their frequencies are 10–20 cm−1 lower than that of native β-sheets.39,52,55,57 The frequency difference we observe here is only a few wave-numbers. The anharmonic shift of amyloid β-sheets can be as small as 5 cm−1,61 which also does not agree with the data here. Thus, we assign the features here to a small or amorphous protein aggregation.

Formalin is used to fix tissues because it cross-links proteins and other biological molecules to create a gel-like state.62 Formalin reacts most readily with the amino acids cysteine and lysine, creating covalent bonds within and across proteins.62-64 Many in vitro studies have determined the mechanism of methylene bridge formation between proteins when exposed to formalin.62,65,66 Cross-linking is an established method for retaining cellular structures and the organization of organelles.67,68 Although the structures are now cross-linked, formalin is thought to preserve native protein secondary structure components.69,70 However, previous studies using circular dichroism and gel electrophoresis studies have also observed structural changes occurring due to fixation and/or wax-embedding processes.70,71

The results presented here indicated that a formalin fixation does not entirely prevent protein degradation within pancreas tissues. In our prior work using 2D IR spectroscopy to study cataracts in lens tissues extracted from mice, we did not observe fixation-related protein structure changes (unpublished results). Lens tissues have the highest protein density of any major organ in the body (ca. ~35%), of which 90% are crystallin proteins, which are extremely stable.39 Pancreas tissues contain proportionally less protein, at ca. ~50 mg protein/gram compared to ca. ~240 mg/mL in the lens cortex (also reported at higher densities depending on location).72,73 Thus, the fixation may be more effective on lens than pancreas tissues.39,74 Another factor is that the pancreas contains many enzymes that digest proteins, and so formalin fixation is done quickly following an animal sacrifice to prevent enzymatic degradation.75 Thus, the change in the secondary structure here might be due to a slow aggregation of partially digested proteins or the incomplete deactivation of enzymes.

It is interesting to note that antibody stains continue to function after fixation. (If the tissue is dehydrated, as is the case for paraffin-embedded fixed tissues, heat- or chemical-induced antigen retrieval is usually necessary prior to staining.)71,76,77 One might expect that aggregation would prevent antibody binding. It is difficult for us to quantify the amount of protein that has aggregated, but we believe it is less than 10%, suggesting that antibody binding would still be greater than 90% effective. It would be interesting to perform a series of time-elapsed 2D IR and antibody imaging studies to determine if aggregation is anticorrelated to antibody binding. A study of that nature might help determine which specific proteins are aggregating. The samples in this study were stored in plastic bags in the dark (see Methods), similar to the common practice of storing FFPE tissues on benchtops or equivalent spaces. Future studies might test if storage under nitrogen slows and prevents degradation.

CONCLUSION

This study reports 2D IR bioimages of formalin-fixed pancreas tissues collected over nearly a one-year time span. The data were analyzed with three different quantities that are sensitive to a protein secondary structure: frequencies, anharmonicities, and cross peaks. All three quantities are consistent with protein aggregation into amorphous small aggregates with small amounts of non-native β-sheet. Thus, formalin fixation does not completely arrest the secondary structure of the proteins within the pancreatic tissue. It may be important to take our results into account when performing chemical, antibody, or spectroscopic analyses of formalin-fixed pancreas tissues. It would be interesting to perform a year-long series of antibody stains alongside 2D IR bioimaging to look for a correlation in antibody binding and secondary structure changes.

Supplementary Material

supplemental

ACKNOWLEDGMENTS

The authors thank the Histology Laboratory at the University of Wisconsin-Madison school of Veterinary Medicine and the Experimental Animal Pathology Laboratory (EAPL) at the University of Wisconsin-Madison for tissue embedding and slicing. The authors also thank K. M. Farrell for composing the mouse featured in the TOC graphic.

Funding

This work was supported by the National Institutes of Health (1R01DK101573-01, 1R01DK101573-06, 1R01DK102948-01A1 (A.D.A.), and R01DK079895 (M.T.Z.)).

Footnotes

Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jpcb.1c05554.

Supplementary figures and information on Gaussian fitting (PDF)

The authors declare the following competing financial interest(s): Martin Zanni is co-owner of PhaseTech Spectroscopy, Inc., which sells mid-IR and visible pulse shapers and 2D spectrometers like those used in this publication.

Contributor Information

Sidney S. Dicke, Department of Chemistry, University of Wisconsin–Madison, Madison, Wisconsin 53706, United States.

Ariel M. Alperstein, Department of Chemistry, University of Wisconsin–Madison, Madison, Wisconsin 53706, United States; Present Address: Department of Chemistry, University of Minnesota, Minneapolis, MN 55455, United States.

Kathryn L. Schueler, Department of Biochemistry, University of Wisconsin–Madison, Madison, Wisconsin 53706, United States

Donald S. Stapleton, Department of Biochemistry, University of Wisconsin–Madison, Madison, Wisconsin 53706, United States

Shane P. Simonett, Department of Biochemistry, University of Wisconsin–Madison, Madison, Wisconsin 53706, United States

Caitlyn R. Fields, Department of Chemistry, University of Wisconsin–Madison, Madison, Wisconsin 53706, United States.

Farzaneh Chalyavi, Department of Chemistry, University of Wisconsin–Madison, Madison, Wisconsin 53706, United States.

Mark P. Keller, Department of Biochemistry, University of Wisconsin–Madison, Madison, Wisconsin 53706, United States

Alan D. Attie, Department of Biochemistry, University of Wisconsin–Madison, Madison, Wisconsin 53706, United States

Martin T. Zanni, Department of Chemistry, University of Wisconsin–Madison, Madison, Wisconsin 53706, United States.

REFERENCES

  • (1).Biancalana M; Koide S Molecular Mechanism of Thioflavin-T Binding to Amyloid Fibrils. Biochim. Biophys. Acta, Proteins Proteomics 2010, 1804 (7), 1405–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (2).Howie AJ; Brewer DB; Howell D; Jones AP Physical Basis of Colors Seen in Congo Red-Stained Amyloid in Polarized Light. Lab. Invest 2008, 88 (3), 232–242. [DOI] [PubMed] [Google Scholar]
  • (3).Yakupova EI; Bobyleva LG; Vikhlyantsev IM; Bobylev AG Congo Red and Amyloids: History and Relationship. Biosci. Rep 2019, 39(1). DOI: 10.1042/BSR20181415 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (4).Kayed R; Glabe CG Conformation-Dependent Anti-Amyloid Oligomer Antibodies. Methods Enzymol. 2006, 413, 326–44. [DOI] [PubMed] [Google Scholar]
  • (5).Franken LE; Grünewald K; Boekema EJ; Stuart MCA A Technical Introduction to Transmission Electron Microscopy for Soft-Matter: Imaging, Possibilities, Choices, and Technical Developments. Small 2020, 16 (14).1906198. [DOI] [PubMed] [Google Scholar]
  • (6).Qiang W; Yau WM; Lu JX; Collinge J; Tycko R Structural Variation in Amyloid-β Fibrils from Alzheimer’s Disease Clinical Subtypes. Nature 2017, 541 (7636), 217–221. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (7).Tycko R Solid-State NMR Studies of Amyloid Fibril Structure. Annu. Rev. Phys. Chem 2011, 62 (1), 279–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (8).Reif B; Ashbrook SE; Emsley L; Hong M Solid-State NMR Spectroscopy. Nat. Rev. Methods Prim 2021, 1 (1), 1–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (9).Molugu TR; Lee S; Brown MF Concepts and Methods of Solid-State NMR Spectroscopy Applied to Biomembranes. Chem. Rev 2017, 117 (19), 12087–12132. [DOI] [PubMed] [Google Scholar]
  • (10).Noreen R; Moenner M; Hwu Y; Petibois C FTIR Spectro-Imaging of Collagens for Characterization and Grading of Gliomas. Biotechnol. Adv 2012, 30 (6), 1432–46. [DOI] [PubMed] [Google Scholar]
  • (11).Ali MHM; Rakib F; Al-Saad K; Al-Saady R; Goormaghtigh E An Innovative Platform Merging Elemental Analysis and Ftir Imaging for Breast Tissue Analysis. Sci. Rep 2019, 9 (1), 18–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (12).Ali MHM; Rakib F; Abdelalim EM; Limbeck A; Mall R; Ullah E; Mesaeli N; McNaughton D; Ahmed T; Al-Saad K Fourier-Transform Infrared Imaging Spectroscopy and Laser Ablation-ICPMS New Vistas for Biochemical Analyses of Ischemic Stroke in Rat Brain. Front. Neurosci 2018, 12(647). DOI: 10.3389/fnins.2018.00647 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (13).Bhargava R Infrared Spectroscopic Imaging: The next Generation. Appl. Spectrosc 2012, 66 (10), 1091–1120. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (14).Baker MJ; Trevisan J; Bassan P; Bhargava R; Butler HJ; Dorling KM; Fielden PR; Fogarty SW; Fullwood NJ; Heys KA; et al. Using Fourier Transform IR Spectroscopy to Analyze Biological Materials. Nat. Protoc 2014, 9 (8), 1771–1791. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (15).Stewart S; Priore RJ; Nelson MP; Treado PJ Raman Imaging. Annu. Rev. Anal. Chem 2012, 5, 337–360. [DOI] [PubMed] [Google Scholar]
  • (16).Opilik L; Schmid T; Zenobi R Modern Raman Imaging: Vibrational Spectroscopy on the Micrometer and Nanometer Scales. Annu. Rev. Anal. Chem 2013, 6, 379–398. [DOI] [PubMed] [Google Scholar]
  • (17).Zhang Y; Hong H; Cai W Imaging with Raman Spectroscopy. Curr. Pharm. Biotechnol 2010, 11 (6), 654–661. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (18).Jones RR; Hooper DC; Zhang L; Wolverson D; Valev VK Raman Techniques: Fundamentals and Frontiers. Nanoscale Res. Lett 2019, 14 (1), 1–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (19).Wei L; Hu F; Shen Y; Chen Z; Yu Y; Lin CC; Wang MC; Min W Live-Cell Imaging of Alkyne-Tagged Small Biomolecules by Stimulated Raman Scattering. Nat. Methods 2014, 11 (4), 410–412. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (20).Hu F; Zeng C; Long R; Miao Y; Wei L; Xu Q; Min W Supermultiplexed Optical Imaging and Barcoding with Engineered Polyynes. Nat. Methods 2018, 15 (3), 194–200. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (21).Chen Z; Wei L; Zhu X; Min W Extending the Fundamental Imaging-Depth Limit of Multi-Photon Microscopy by Imaging with Photo-Activatable Fluorophores. Opt. Express 2012, 20 (17), 18525. [DOI] [PubMed] [Google Scholar]
  • (22).Fernandez DC; Bhargava R; Hewitt SM; Levin IW Infrared Spectroscopic Imaging for Histopathologic Recognition. Nat. Biotechnol 2005, 23 (4), 469–474. [DOI] [PubMed] [Google Scholar]
  • (23).Zohdi V; Whelan DR; Wood BR; Pearson JT; Bambery KR; Black MJ Importance of Tissue Preparation Methods in FTIR Micro-Spectroscopical Analysis of Biological Tissues: “Traps for New Users. PLoS One 2015, 10 (2), e0116491. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (24).Choo LP in.; Wetzel DL; Halliday WC; Jackson M; LeVine SM; Mantsch HH In Situ Characterization of β-Amyloid in Alzheimer’s Diseased Tissue by Synchrotron Fourier Transform Infrared Microspectroscopy. Biophys. J 1996, 71 (4), 1672–1679. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (25).Gazi E; Dwyer J; Gardner P; Ghanbari-Siahkali A; Wade AP; Miyan J; Lockyer NP; Vickerman JC; Clarke NW; Shanks JH; et al. Applications of Fourier Transform Infrared Microspectroscopy in Studies of Benign Prostate and Prostate Cancer. A Pilot Study. J. Pathol 2003, 201 (1), 99–108. [DOI] [PubMed] [Google Scholar]
  • (26).Kole MR; Reddy RK; Schulmerich MV; Gelber MK; Bhargava R Discrete Frequency Infrared Microspectroscopy and Imaging with a Tunable Quantum Cascade Laser. Anal. Chem 2012, 84 (23), 10366–10372. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (27).Petibois C; Desbat B Clinical Application of FTIR Imaging: New Reasons for Hope. Trends Biotechnol. 2010, 28, 495–500. [DOI] [PubMed] [Google Scholar]
  • (28).Palombo F; Cremers SG; Weinberg PD; Kazarian SG Application of Fourier Transform Infrared Spectroscopic Imaging to the Study of Effects of Age and Dietary L-Arginine on Aortic Lesion Composition in Cholesterol-Fed Rabbits. J. R. Soc., Interface 2009, 6 (37), 669–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (29).Ling Y; Rios HF; Myers ER; Lu Y; Feng JQ; Boskey AL DMP1 Depletion Decreases Bone Mineralization in Vivo: An FTIR Imaging Analysis. J. Bone Miner. Res 2005, 20 (12), 2169–2177. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (30).Tiwari S; Kajdacsy-Balla A; Whiteley J; Cheng G; Hewitt SM; Bhargava R INFORM: INFrared-Based ORganizational Measurements of Tumor and Its Microenvironment to Predict Patient Survival. Sci. Adv 2021, 7 (6), eabb8292. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (31).Reddy RK; Bhargava R Accurate Histopathology from Low Signal-to-Noise Ratio Spectroscopic Imaging Data. Analyst 2010, 135 (11), 2818–2825. [DOI] [PubMed] [Google Scholar]
  • (32).Su WH; Sun DW Fourier Transform Infrared and Raman and Hyperspectral Imaging Techniques for Quality Determinations of Powdery Foods: A Review. Compr. Rev. Food Sci. Food Saf 2018, 17 (1), 104–122. [DOI] [PubMed] [Google Scholar]
  • (33).Glassford SE; Byrne B; Kazarian SG Recent Applications of ATR FTIR Spectroscopy and Imaging to Proteins. Biochim. Biophys. Acta, Proteins Proteomics 2013, 1834 (12), 2849–2858. [DOI] [PubMed] [Google Scholar]
  • (34).Wrobel TP; Mukherjee P; Bhargava R Rapid Visualization of Macromolecular Orientation by Discrete Frequency Mid-Infrared Spectroscopic Imaging. Analyst 2017, 142, 75. [DOI] [PubMed] [Google Scholar]
  • (35).Lewis EN; Levin IW; Treado PJ; Reeder RC; Story GM; Dowrey AE; Marcott C Fourier Transform Spectroscopic Imaging Using an Infrared Focal-Plane Array Detector. Anal. Chem 1995, 67 (19), 3377–3381. [DOI] [PubMed] [Google Scholar]
  • (36).Wrobel TP; Kole MR; Bhargava R Emerging Trends and Opportunities in Discrete-Frequency Infrared and Raman Spectroscopic Imaging. Spectroscopy 2016, 31 (6), 28–45. [Google Scholar]
  • (37).Baiz CR; Schach D; Tokmakoff A Ultrafast 2D IR Microscopy. Opt. Express 2014, 22 (15), 18724. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (38).Ostrander JS; Serrano AL; Ghosh A; Zanni MT Spatially Resolved Two-Dimensional Infrared Spectroscopy via Wide-Field Microscopy. ACS Photonics 2016, 3 (7), 1315–1323. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (39).Alperstein AM; Ostrander JS; Zhang TO; Zanni MT Amyloid Found in Human Cataracts with Two-Dimensional Infrared Spectroscopy. Proc. Natl. Acad. Sci U. S. A 2019, 116 (14), 6602–6607. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (40).Fournier F; Guo R; Gardner EM; Donaldson PM; Loeffeld C; Gould IR; Willison KR; Klug DR Biological and Biomedical Applications of Two-Dimensional Vibrational Spectroscopy: Proteomics, Imaging, and Structural Analysis. Acc. Chem. Res 2009, 42 (9), 1322–1331. [DOI] [PubMed] [Google Scholar]
  • (41).Ghosh A; Ostrander JS; Zanni MT Watching Proteins Wiggle: Mapping Structures with Two-Dimensional Infrared Spectroscopy. Chem. Rev 2017, 117 (16), 10726–10759. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (42).Petti MK; Lomont JP; Maj M; Zanni MT Two-Dimensional Spectroscopy Is Being Used to Address Core Scientific Questions in Biology and Materials Science. J. Phys. Chem. B 2018, 122 (6), 1771–1780. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (43).Middleton CT; Woys AM; Mukherjee SS; Zanni MT Residue-Specific Structural Kinetics of Proteins through the Union of Isotope Labeling, Mid-IR Pulse Shaping, and Coherent 2D IR Spectroscopy. Methods 2010, 52 (1), 12–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (44).Hamm P; Zanni M Concepts and Methods of 2D Infrared Spectroscopy; Cambridge University Press, 2011. [Google Scholar]
  • (45).Baiz CR; Błasiak B; Bredenbeck J; Cho M; Choi JH; Corcelli SA; Dijkstra AG; Feng CJ; Garrett-Roe S; Ge NH; et al. Vibrational Spectroscopic Map, Vibrational Spectroscopy, and Intermolecular Interaction. Chem. Rev 2020, 120 (15), 7152–7218. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (46).Lasch P; Noda I Two-Dimensional Correlation Spectroscopy (2D-COS) for Analysis of Spatially Resolved Vibrational Spectra. Appl. Spectrosc 2019, 73 (4), 359–379. [DOI] [PubMed] [Google Scholar]
  • (47).Murayama K; Czarnik-Matusewicz B; Wu Y; Tsenkova R; Ozaki Y Comparison between Conventional Spectral Analysis Methods, Chemometrics, and Two-Dimensional Correlation Spectroscopy in the Analysis of near-Infrared Spectra of Protein. Appl. Spectrosc 2000, 54 (7), 978–985. [Google Scholar]
  • (48).Morita SI; Miura YF; Sugi M; Ozaki Y New Correlation Indices Invariant to Band Shifts in Generalized Two-Dimensional Correlation Infrared Spectroscopy. Chem. Phys. Lett 2005, 402 (1–3), 251–257. [Google Scholar]
  • (49).Fayer MD Ultrafast Infrared Vibrational Spectroscopy, 1st ed.; CRC Press, 2013. [Google Scholar]
  • (50).Cho M Two-Dimensional Optical Spectroscopy, 1st ed.; CRC Press, 2009. [Google Scholar]
  • (51).Tracy KM; Guchhait B; Tibbetts C; Luther BM; Krummel A Visualizing Chemical Dynamics in an Ionic Liquid Microdroplet Using Ultrafast 2D IR Microscopy. ChemRxiv (2019). DOI: 10.26434/chemrxiv.9936464.v1 [DOI] [Google Scholar]
  • (52).Zhang TO; Alperstein AM; Zanni MT Amyloid β-Sheet Secondary Structure Identified in UV-Induced Cataracts of Porcine Lenses Using 2D IR Spectroscopy. J. Mol. Biol 2017, 429 (11), 1705–1721. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (53).Li DS; Yuan YH; Tu HJ; Liang Q. Le; Dai LJ A Protocol for Islet Isolation from Mouse Pancreas. Nat. Protoc 2009, 4 (11), 1649–1652. [DOI] [PubMed] [Google Scholar]
  • (54).Grechko M; Zanni MT Quantification of Transition Dipole Strengths Using 1D and 2D Spectroscopy for the Identification of Molecular Structures via Exciton Delocalization: Application to α-Helices. J. Chem. Phys 2012, 137 (18), 184202. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (55).Fields CR; Dicke SS; Petti MK; Zanni MT; Lomont JP A Different Hiapp Polymorph Is Observed in Human Serum than in Aqueous Buffer: Demonstration of a New Method for Studying Amyloid Fibril Structure Using Infrared Spectroscopy. J. Phys. Chem. Lett 2020, 11 (15), 6382–6388. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (56).Hahn S; Kim S-S; Lee C; Cho M Characteristic Two-Dimensional IR Spectroscopic Features of Antiparallel and Parallel-Sheet Polypeptides: Simulation Studies. J. Chem. Phys 2005, 123, 84905. [DOI] [PubMed] [Google Scholar]
  • (57).Lomont JP; Ostrander JS; Ho JJ; Petti MK; Zanni MT Not All β-Sheets Are the Same: Amyloid Infrared Spectra, Transition Dipole Strengths, and Couplings Investigated by 2D IR Spectroscopy. J. Phys. Chem. B 2017, 121 (38), 8935–8945. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (58).Strasfeld DB; Ling YL; Gupta R; Raleigh DP; Zanni MT Strategies for Extracting Structural Information from 2D IR Spectroscopy of Amyloid: Application to Islet Amyloid Polypeptide. J. Phys. Chem. B 2009, 113 (47), 15679–15691. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (59).Zanni MT; Gnanakaran S; Stenger J; Hochstrasser RM Heterodyned Two-Dimensional Infrared Spectroscopy of Solvent-Dependent Conformations of Acetylproline-NH 2. J. Phys. Chem. B 2001, 105 (28), 6520–6535. [Google Scholar]
  • (60).Hamm P; Lim M; Hochstrasser RM Structure of the Amide I Band of Peptides Measured by Femtosecond Nonlinear-Infrared Spectroscopy. J. Phys. Chem. B 1998, 102 (31), 6123–6138. [Google Scholar]
  • (61).Dunkelberger EB; Grechko M; Zanni MT Transition Dipoles from 1D and 2D Infrared Spectroscopy Help Reveal the Secondary Structure of Proteins: Application to Amyloids. J. Phys. Chem. B 2015, 119 (44), 14065–14075. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (62).Thavarajah R; Mudimbaimannar VK; Elizabeth J; Rao UK; Ranganathan K Chemical and Physical Basics of Routine Formaldehyde Fixation. J. Oral. Maxillofac. Pathol 2012, 16 (3), 400–405. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (63).Kamps JJAG; Hopkinson RJ; Schofield CJ; Claridge TDW How Formaldehyde Reacts with Amino Acids. Commun. Chem 2019, 2 (1), 1–14. [Google Scholar]
  • (64).Michiels TJM; Schöneich C; Hamzink MRJ; Meiring HD; Kersten GFA; Jiskoot W; Metz B Novel Formaldehyde-Induced Modifications of Lysine Residue Pairs in Peptides and Proteins: Identification and Relevance to Vaccine Development. Mol. Pharmaceutics 2020, 17 (11), 4375–4385. [DOI] [PubMed] [Google Scholar]
  • (65).Lu K; Ye W; Zhou L; Collins LB; Chen X; Gold A; Ball LM; Swenberg JA Structural Characterization of Formaldehyde-Induced Cross-Links between Amino Acids and Deoxynucleosides and Their Oligomers. J. Am. Chem. Soc 2010, 132 (10), 3388–3399. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (66).Metz B; Kersten GFA; Hoogerhout P; Brugghe HF; Timmermans HAM; De Jong A; Meiring H; Ten Hove J; Hennink WE; Crommelin DJA; et al. Identification of Formaldehyde-Induced Modifications in Proteins: Reactions with Model Peptides. J. Biol. Chem 2004, 279 (8), 6235–6243. [DOI] [PubMed] [Google Scholar]
  • (67).Graham L; Orenstein JM Processing Tissue and Cells for Transmission Electron Microscopy in Diagnostic Pathology and Research. Nat. Protoc 2007, 2 (10), 2439–2450. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (68).Wang NS; Minassian H The Formaldehyde-Fixed and Paraffin-Embedded Tissues for Diagnostic Transmission Electron Microscopy: A Retrospective and Prospective Study. Hum. Pathol 1987, 18 (7), 715–727. [DOI] [PubMed] [Google Scholar]
  • (69).Mason JT; O’Leary TJ Effects of Formaldehyde Fixation on Protein Secondary Structure: A Calorimetric and Infrared Spectroscopic Investigation. J. Histochem. Cytochem 1991, 39 (2), 225–229. [DOI] [PubMed] [Google Scholar]
  • (70).Fowler CB; O’Leary TJ; Mason JT Modeling Formalin Fixation and Histological Processing with Ribonuclease A: Effects of Ethanol Dehydration on Reversal of Formaldehyde Cross-Links. Lab. Invest 2008, 88 (7), 785–791. [DOI] [PubMed] [Google Scholar]
  • (71).Fowler CB; Evers DL; O’Leary TJ; Mason JT Antigen Retrieval Causes Protein Unfolding: Evidence for a Linear Epitope Model of Recovered Immunoreactivity. J. Histochem. Cytochem 2011, 59 (4), 366–381. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (72).Gu C; Xu H Effect of Oxidative Damage Due to Excessive Protein Ingestion on Pancreas Function in Mice. Int. J. Mol. Sci 2010, 11 (11), 4591–4600. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (73).Zhao H; Magone MT; Schuck P The Role of Macromolecular Crowding in the Evolution of Lens Crystallins with High Molecular Refractive Index. Phys. Biol 2011, 8 (4), No. 046004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (74).Truscott RJW Eye Lens Proteins and Cataracts. Protein Misfolding Aggregation, Conform. Dis 2007, 435–447. [Google Scholar]
  • (75).Hudkins KL; Pichaiwong W; Wietecha T; Kowalewska J; Banas MC; Spencer MW; Mühlfeld A; Koelling M; Pippin JW; Shankland SJ; et al. BTBR Ob/Ob Mutant Mice Model Progressive Diabetic Nephropathy. J. Am. Soc. Nephrol 2010, 21 (9), 1533–1542. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (76).Shi SR; Shi Y; Taylor CR Antigen Retrieval Immunohistochemistry: Review and Future Prospects in Research and Diagnosis over Two Decades. J. Histochem. Cytochem 2011, 59 (1), 13–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (77).Bogen S; Vani K; Sompuram S Molecular Mechanisms of Antigen Retrieval: Antigen Retrieval Reverses Steric Interference Caused by Formalin-Induced Cross-Links. Biotech. Histochem 2009, 84 (5), 207–215. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

supplemental
NIHMS1745273-supplement-supplemental.pdf (370.5KB, pdf)

RESOURCES