Multimodal microscopy for characterization of amyloid-{\unicode[Times]{x3B2}} plaques biomarkers in animal model of Alzheimer's disease
Renan Cunha, Lucas Lafeta, Emerson A. Fonseca, Alexandre Barbosa, Marco A. Romano-Silva, Rafael Vieira, Ado Jorio, Leandro M. Malard
MMultimodal microscopy for characterization of amyloid- β plaquesbiomarkers in animal model of Alzheimer’s disease † Renan Cunha ‡ a , Lucas Lafeta ‡ a , Emerson A. Fonseca ‡ a,b , Alexandre Barbosa a,c Marco A. Romano-Silva d , Rafael Vieira e , Ado Jorio a,b , and Leandro M. Malard ∗ a Given the long subclinical stage of Alzheimer’s disease (AD), the study of biomarkers is relevant bothfor early diagnosis and the fundamental understanding of the pathophysiology of AD. Biomarkersprovided by Amyloid- β (A β ) plaques have led to an increasing interest in characterizing this hall-mark of AD due to its promising potential. In this work, we characterize A β plaques by label-freemultimodal imaging: we combine two-photon excitation autofluorescence (TPEA), second harmonicgeneration (SHG), spontaneous Raman scattering (SpRS), coherent anti-Stokes Raman scattering(CARS), and stimulated Raman scattering (SRS) to describe and compare high-resolution imagesof A β plaques in brain tissues of an AD mouse model. Comparing single-laser techniques images,we discuss the origin of the SHG, which can be used to locate the plaque core reliably. We studyboth the core and the halo with vibrational microscopy and compare SpRS and SRS microscopiesfor different frequencies. We also combine SpRS spectroscopy with SRS microscopy and present twocore biomarkers unexplored with SRS microscopy: phenylalanine and amide B. We provide high-resolution SRS images with the spatial distribution of these biomarkers in the plaque and comparedthem with images of the amide I distribution. The obtained spatial correlation corroborates thefeasibility of these biomarkers in the study of A β plaques. Furthermore, since amide B enables rapidimaging, we discuss its potential as a novel fingerprint for diagnostic applications. INTRODUCTION
Since the first report by Alois Alzheimer , Alzheimer’sdisease (AD) has been described as a progressive neu-rodegenerative disorder characterized by cognitive andbehavioral impairments, ultimately leading to death .For a long time, its diagnosis relied on the analysisof changes present when cognitive impairment was al-ready installed. Nevertheless, advances in basic researchshow that AD has a long subclinical stage, with brainchanges preceding the onset of symptoms . One ofthe efforts in the study of AD is the characterization ofsuch disease-related structural changes with the poten-tial to indicate the state of the progression before clin-ical manifestations . Such effort led to the search forbiomarkers through imaging of extracellular amyloid- β plaques, a pathological hallmark of AD, which can allowa better understanding of the AD pathophysiology and a Departamento de F´ısica, ICEx, Universidade Federal de MinasGerais, Belo Horizonte, MG, 31270-901, Brazil. b Programa de P´os-Gradua¸c˜ao em Inova¸c˜ao Tecnol´ogica, Uni-versidade Federal de Minas Gerais, Belo Horizonte, MG, 31270-901, Brazil. c Departamento de Oftalmologia, Faculdade de Medicina, Uni-versidade Federal de Minas Gerais, Belo Horizonte, MG,31270-901, Brazil. d Departamento de Sa´ude Mental, Faculdade de Medicina, Uni-versidade Federal de Minas Gerais, Belo Horizonte, MG, 31270-901, Brazil. e Departamento de Bioqu´ımica e Imunologia, Universidade Fed-eral de Minas Gerais, Belo Horizonte, MG, 31270-901, Brazil. † Supplementary Information (SI) avaiable. ‡ These authors contributed equally to this work. ∗ E-mail: lmalard@fisica.ufmg.br. provide new perspectives into its early diagnosis .Amyloid- β (A β ) plaques consist of a core formedpredominantly by misfolded A β peptides, rich in β -sheet conformation, surrounded by a predominantly li-pidic halo, in the brain extracellular matrix . Dueto their structural complexity, several imaging methodshave been implemented with the aim to provide a bettercharacterization of A β plaques. Xenobiotic stain-basedapproaches have been extensively applied, using exoge-nous compounds such as methoxy-XO4, Congo red, andThioflavin S (ThioS) . Due to the specificity of the lig-ands, they provide a limited amount of information. Asa result, it is necessary to use different labels to obtaininformation on different biomolecules, as in the case ofusing the immunolabels Iba1, GFAP, and Lamp1 to studythe halo . Moreover, the labels can interfere with theinformation of interest and hamper the interpretation ofrelevant data , reinforcing the use of label-free imag-ing techniques.Nonlinear and vibrational optical imaging have beenincreasingly used in biology and medicine due to its label-free character, sensitivity, three-dimensional optical sec-tioning, and high spatial resolution . Furthermore,it has been shown that multimodal imaging, in whichseveral techniques are combined, can be used to max-imize the gain of information from complex biologicalsystems . Previous studies of A β plaques reportedlabel-free imaging by multiphoton fluorescence , sec-ond harmonic generation (SHG) , spontaneous Ramanscattering (SpRS) , coherent anti-Stokes Raman scat-tering (CARS) , and, more recently, stimulated Ra-man scattering (SRS) . In some studies of brain imag-ing obtained from mice, autofluorescence, SHG show- a r X i v : . [ phy s i c s . m e d - ph ] F e b –17 | ing the protein-rich core, and CARS images showingthe lipid-rich halo, are presented with limited spatialresolution . While the SRS images are presented withhigher spatial resolution , the investigation concernedthe feasibility of narrowband SRS microscopy to studyplaques, restricting the core imaging to the amide I vi-brational mode. Here, we perform a multimodal opticalcharacterization, exploring the optical properties of en-dogenous fluorescent biomolecules, molecular structure,and molecular vibrations of A β plaques to implementhigh-resolution imaging based on two-photon excitationautofluorescence (TPEA), SHG, SpRS, CARS, and SRS.We provide a two-photon excitation fluorescence (TPEF)image of the ThioS stained tissue as a gold standard pro-cedure to validate our label-free imaging. With a higherspatial resolution, we compare the TPEF ThioS staining,TPEA, and SHG images, which highlight the protein-richcore of the plaques, and also discuss the origin of the SHGsignal in A β plaques. Comparing SpRS and SRS imagesfor various frequencies of different plaques, we observe analmost identical correspondence, even in different scat-tering geometries. Among those frequencies, we com-pare the spatial distribution of Phenylalanine (Phe) andamide I in the plaque, showing that Phe, not yet exploredin SRS microscopy, is very specific for the core, charac-terizing a possible biomarker in the low-frequency regionin addition to amide I. Combining Raman spectroscopywith SRS microscopy, we look for possible biomarkers inthe high-frequency region. We present the plaque imagebased on the unsaturated lipids vibrational mode, whichexhibit a good correlation with other halo vibrations. Wepropose a core biomarker based on the amide B vibra-tional mode, also not yet explored in SRS (nor CARS andSpRS) microscopy. We also compare the spatial distribu-tion of amide B and amide I in the plaque. Since amideB enables rapid imaging, this novel fingerprint could beused for diagnostic applications. RESULTS AND DISCUSSION
Figure 1 shows the zoom at the A β plaque shown bythe blue square in Fig. 4F (Materials and Methods)acquired by using different microscopic modalities. Alltechniques enable the identification of the A β plaque,contrary to the bright field image in Fig. 1A, which ex-hibits no features. A comparison of the findings of mul-timodal imaging techniques (Figs. 1C-H) with TPEFThioS staining (Fig. 1B), established as a gold standard,allows an unambiguous plaque identification. Two-photon excitation autofluorescence microscopy
Figure 1C shows the TPEA image obtained with thelaser excitation wavelength at 810 nm and no filters at thedetection in order to collect the entire emission window.There is a small contribution of the SHG emission (see below) to the TPEA image, and therefore the SHG is sub-tracted from the TPEA image, generating Fig. 1C. Theorigin of TPEA of biological tissues can be attributed toendogenous fluorophores such as nicotinamide adeninedinucleotide, flavin adenine dinucleotide, flavoproteins,tyrosine, tryptophan, collagen, elastin, and lipopigments,to name a few . These fluorophores emit lightacross the visible region with overlapping emissions .Lipopigments such as lipofuscin, present in neurons, andassociated with age-related neurodegeneration , ex-hibit a broad spectrum, typically in the entire visiblerange . For this reason, although the A β plaqueis present in the TPEA image of Fig. 1C, the precisecontribution of autofluorescent biomolecules to the ob-served signal is still under discussion, and proper charac-terization of biomarkers is a non-trivial task. Moreover,the extraction of information from the TPEA image onlyby optical characterization can be impractical . Addi-tionally, since those autofluorescent biomolecules are notexclusive to A β plaques, other bright features are alsopresent. In our TPEA images, blood vessels are observed,highlighted by the arrows in Fig. 1C. Second harmonic generation microscopy
Another technique that can also be implemented usinga single laser wavelength is SHG microscopy. SHG canbe present in materials lacking inversion symmetry .This signal relies on the anisotropy that particular biolog-ical structures exhibit, giving rise to appreciable second-order susceptibility values . Fig. 1D shows an SHGimage acquired with the laser excitation wavelength at810 nm and a bandpass filter centered at 405 nm (10 nmwidth). In this case, the narrow bandpass filter is im-portant to block the TPEA emission. As shown in Figs.1C-D, these techniques can be reliable in identifying theA β core without any extrinsic fluorescent label. Never-theless, the molecular origin of the SHG and proper char-acterization of symmetry-based biomarkers provided byA β plaques is also still under discussion. Biological struc-tures that typically produce SHG include collagen, micro-tubules, and other non-centrosymmetric proteins .In the brain, collagen fibrils have been suggested to affectthe detection of amyloid fibrils , and also that the originof SHG in A β plaques can be attributed to collagen ,or microtubules in neurites of plaques . However, ourimage in Fig. 1D clearly shows that the primary sourceof SHG is the core, where it would not be expected toobserve interaction with collagen and where there are noneurites. Previous studies with immunohistochemistry ofneurites, astrocytes, and microglia, demonstrated thatthese cells are present in the surrounding halo, ratherthan in the core of the A β plaques . Additionally,comparing SHG, TPEA, and TPEF ThioS staining im-ages in Figs. 1B-D allows us to observe similarities inthe core images, suggesting a common origin. Further-more, ThioS is considered a specific label for the core of –17 | FIG. 1. High-resolution multimodal imaging of an A β plaque from the hippocampus of a 12-month-old mouse brain tissue.(A) Bright field zoomed image from Fig. 4F. (B) TPEF image of the tissue after ThioS staining. (C) TPEA and (D) SHGimages. (E) SpRS image at 1675 cm − . (F) SRS image at 1675 cm − and (G) 2850 cm − . (H) CARS image at 2850 cm − .All images are of the same region. All scale bars indicate 20 µ m. A β plaques, being a characteristic ligand of the β -sheetconformation . Studies of β -sheet conformation of invitro and in situ amyloid structures suggest that the ori-gin of SHG emission is due to interfacial and symmetryproperties of the protein structures , which is sup-ported by our core image. Therefore, we consider thatthe origin of the SHG is better associated with thesesymmetry aspects of protein structures in the core ratherthan due to the above-mentioned suggestions regardingcollagen, neurites, and microtubules.In addition to plaque identification based on tech-niques that highlight the core, such as TPEA and SHG,a full biochemical characterization of the core and halorequires techniques that can retrieve chemical informa-tion. Vibrational microscopy
In vibrational microscopy, the plaque core can beaccessed through the C=O stretching mode of amideI . The misfolding of A β peptides, one of theleading events for plaque formation, can be identified bythe blueshifted ( ∼
15 cm − ) amide I on SpRS, consistentto a β -sheet secondary structure conformation in relationto the α -helix secondary structure peak (at 1660 cm − ). Figure 1E shows a SpRS image set at such a vibrationalfingerprint (1675 cm − ), in which the core is clearly vis-ible, and therefore can be used as a label-free imagingtechnique. Since SpRS is an incoherent effect, its majordisadvantage is that a hyperspectral image like the oneshown in Fig. 1E took approximately 20 hours, mak-ing this a time-consuming technique for histopathologicalstudies and unfeasible for in vivo studies. On the otherhand, such a disadvantage is circumvented by coherentnonlinear vibrational microscopy approaches, which keepthe rich chemical information provided by Raman spec-troscopy without the drawback of low throughput .Figures 1F-G show the SRS images taken at 1675 cm − (amide I) and 2850 cm − , respectively. The 2850 cm − Raman peak is ascribed to the C-H stretching modespresent in lipid-rich regions, allowing Raman-based mi-croscopies to identify the halo around the core as shownin Fig. 1G. Figure 1H shows the CARS image tuned at2850 cm − , also revealing the lipid-rich halo around theplaque core. For the CARS image, we have used a band-pass filter (660/13 mn) to measure only the CARS contri-bution; however, there is a residual TPEA signal from thecore. Hence, although at the same vibrational frequencyof the SRS (Fig. 1G), the CARS image also shows asignal from the core. Additionally, the implemented ho-modyne detection scheme does not prevent non-resonant –17 | background signals . Furthermore, due to the poor sen-sitivity of our PMT in the near-infrared region, our ex-perimental setup for CARS does not allow us to measurevibrational modes below 2500 cm − .As shown in Fig. 1, label-free, nonlinear optical mi-croscopies such as TPEA and SHG can be very useful tolocate A β plaques; however, it is limited for retrievingchemical information. Vibrational microscopies circum-vent this limitation, also allowing to locate other struc-tures, such as the halo shown in Figs. 1G-H. Addition-ally, since the SRS signal is unaffected by a non-resonantbackground and proportional to the scatterers concentra-tion, SRS and SpRS share virtually identical spectra .We show in Fig. 2 a comparison of SRS and SpRS im-ages for different vibrational frequencies of the plaque inFig. 1 (Figs. 2A-B) and of other two different plaques(Figs. 2C-F). The plaques were located at the hippocam-pus (Figs. 2A-D) and at the cortex (Figs. 2E-F). We alsoshow a comparison of SRS and SpRS spectra from 2800to 3075 cm − (SI, Fig. S2).The broad Raman band from 2800 to 3000 cm − (SI,Fig. S2) is commonly found in organic samples, as it isthe composition of important chemical vibrations, suchas the CH stretching modes at ∼ − (lipids)and ∼ − (proteins/lipids), and CH stretch-ing mode at ∼ − (proteins/lipids) . Oppositelyto the case of SHG, TPEA, and TPEF ThioS staining,the halo can be imaged clearly by SRS and SpRS: at2850 cm − only the halo is visible. By changing tohigher frequencies, the halo vibrations start to overlapwith vibrational modes of proteins and, therefore, theplaque core is also present in the SRS and SpRS imagestaken at 2880 cm − and 2930 cm − , as shown in Fig. 2.These are the most intense Raman modes in the plaques ,and give rise to signals of lipids forming the halo sur-rounding the A β core, whose correlation with dystrophicneurites, microglia and astrocytes has been extensivelystudied . Previous studies have explored the rela-tion between neuroinflammation and AD , which isassociated with the presence of astrocytes, microglia, anddystrophic neurites distributed in the halo surroundingthe A β fibrillar deposits . The influence of size,location, speed, and extent of plaque growth and theirrelation to the evidence of neurotoxicity of the plaqueswas reported in a sequential analysis of the area of neu-ritic dystrophy in the halo . These findings highlight theneed for tools capable of characterizing halo biomarkersand show how the specific information provided by SRSand SpRS microscopies can be useful, especially regard-ing plaque expansion and neurodegeneration progression. Vibrational biomarkers
SRS and SpRS microscopies also take a step furtherin the study of the plaque core compared to TPEA andSHG. The overlapping autofluorescent signals from dif-ferent biomolecules hampers the tracing of the origin of autofluorescence in the plaque only by optical methods,and the origin of the SHG signal is still uncertain. Dueto the well-established chemical specificity of SRS andSpRS, it is possible to resolve different signals producedin the core. The fourth column (from left to right) ofFig. 2 shows the amide I (1675 cm − ) frequency, throughwhich the core is identified. The amide I band is a typ-ical protein band in Raman spectra associated with thepolypeptide backbone . To date, it has been the onlyvibrational signature used to identify the core with SRSmicroscopy , which raises the question about the feasi-bility of using new signatures and the information theycan reveal about such a complex structure as A β plaques.In this regard, we image the plaque through the vibra-tional mode of the phenylalanine (Phe, 1007 cm − , lastcolumn of Figs. 2A-F), a hydrophobic, aromatic aminoacid found in two positions of the A β peptide primarystructure . Its position and contact with other aminoacids seem to have a fundamental role in the β -sheet fibrilconformation . Furthermore, a significant differenceof Phe concentration was obtained between the core andsurrounding tissue , making the image of the core clear,which suggests its use in the study of A β plaques. Wealso show a comparison between the spatial distributionof Phe and amide I in the plaque (SI, Fig. S3), showingthe specificity of Phe in the core. Therefore, we regardPhe as a core biomarker in the low-frequency region inaddition to amide I.In all our images in Fig. 2, there is a good correspon-dence between SRS and SpRS microscopies. The slightdifferences between the two are due to the scattering ge-ometry: the measurement of SRS was performed in atransmission geometry, in contrast to SpRS, measured ina reflection mode. Additionally, the SpRS images for thetwo core biomarkers (1675 and 1007 cm − ) were back-ground subtracted to enhance the contrast. Neverthe-less, the SpRS and SRS spectra from 2800 to 3075 cm − match very well (SI, Fig. S2). Therefore, all the chem-ical information delivered by spontaneous Raman spec-troscopy can be used for narrowband SRS imaging with-out loss of information. This demonstrates that we cancombine the chemical specificity and full spectral data ofSpRS spectroscopy with the speed and resolution of SRSmicroscopy reliably and straightforwardly.An important difference between the low-frequencycore and high-frequency halo biomarkers presented hereis that the latter has the highest intensity. For this rea-son, most studies with coherent Raman microscopy werecarried out in this region . To investigate vibrationalsignatures of the core in the high-frequency region, wefurther applied the combination of Raman spectroscopyand SRS microscopy. Figure 3A shows typical SpRSspectra from the core (blue) and the halo (red) of anA β plaque at the high-frequency region. This region isusually unexplored to obtain information about proteins.Nevertheless, besides the already commented Ramanbands from 2800 to 3000 cm − , there is a pronouncedRaman band between 3050 and 3090 cm − with a peak –17 | FIG. 2. Comparison of narrowband stimulated Raman microscopy and spontaneous Raman microscopy in different scatteringgeometries. Three different A β plaques imaged by SRS (A, C, E) and hyperspectral SpRS (B, D, F), taken at differentvibrational energies, as listed at the bottom of the images. All scale bars are 20 µ m. Images without a scale bar share the scalebar of the leftmost image in the same line. –17 | FIG. 3. Amide B and unsaturated lipids as core and halo biomarkers, respectively, in the high-frequency region. (A) TypicalSpRS spectrum of the halo (upper, red) and core (lower, blue) in the high-frequency region. The purple vertical bar identifiesthe frequency region of the SRS image in (C), while the orange bar identifies the frequency region of the image in (D). (B)Overview SRS image of the hippocampus with several A β plaques. The white rectangle identifies the plaque in (C-H). Scalebar is 60 µ m. (C) SRS image taken at 3070 cm − , attributed to amide B vibration. (D) SRS image taken at 3019 cm − ,attributed to unsaturated lipids and exhibiting a good correlation with lipids vibration in (F). (E) SRS image taken at 2930cm − (protein/lipids), (F) SRS image taken at 2850 cm − (lipids), and (G) the subtraction of images in (E) and (F). (H) SRSimage taken at 1675 cm − , attributed to amide I vibration. All scale bars in (C)-(H) are 20 µ m. at ∼ − in the core spectrum. This signal hasbeen assigned by infrared spectroscopy as the amide B vi-brational mode present in proteins due to N-H stretchingmodes . It is usually part of a Fermi resonance doublet(amide A and B), and it is resonant with an amide IIcombination mode in β -sheet conformation . This isof substantial significance since the amide II band is con-sidered too weak or absent in the Raman spectrum ,but it provides valuable structural information, and it issuggested for secondary structure prediction as a coun-terpart of amide I . Figure 3B shows an overview SRSimage of a hippocampus with several A β plaques basedon lipids CH stretching mode (2850 cm − ) and acquiredwith a 20X objective (numerical aperture, 0.75). Thewhite rectangle identifies the plaque shown in Figs. 3C-H. Figure 3C shows the SRS image taken at 3070 cm − ,which images the core clearly, thus demonstrating thefeasibility of using the amide B frequency to locate thecore of the A β plaques. To the best of our knowledge, nocore signature in this high-frequency region has been re-ported. Such a specific image in this region could only bepresented by subtracting the images based on frequenciessuch as 2930 cm − (Fig. 3E, proteins/lipids CH stretch-ing mode) and 2850 cm − (Fig. 3F, lipids CH stretchingmode). Comparing this subtraction in Fig. 3G with theimage based on amide I ( ∼ − ) in Fig. 3H, a good correlation is observed. Such a correlation is alsopresent when comparing with amide B image (Fig. 3C),demonstrating the applicability of this vibrational mode.We compare additional plaques images (SI, Fig. S4),showing the robustness of the method. A comparison ofthe spatial distribution of amide B and amide I in theplaque is also provided (SI, Fig. S5). Therefore, we re-gard amide B as a core biomarker in the high-frequencyregion.The importance of these results for microscopy can alsobe appreciated in two ways: the analysis of different sig-natures for the core, such as amide I, Phe, and amideB vibrational modes, can reveal complementary informa-tion about the plaque, which can be valuable for its basicunderstanding and its precise role in AD-related neurode-generation. Also, while amide B is described as a weaklyabsorbing component in infrared spectroscopy , it al-lowed the acquisition of core images with high contrast,high intensity, and few accumulations with SRS. In ourexperiment, we obtained SRS images at the amide B fre-quency with approximately 7 times higher mean contrastthan those obtained at the amide I frequency. This resultsuggests amide B as a novel fingerprint to be incorporatedin the study of A β plaques for diagnostic applications,where the acquisition time is of great relevance.Correspondingly, the halo spectrum shows a band be- –17 | tween 3010 cm − and 3025 cm − (peak at 3019 cm − ),which is absent in the core spectrum shown in Fig. 3A.This band is associated with unsaturated =CH bindingsand is attributed to unsaturated lipids . Figure 3Gshows the SRS image at 3019 cm − , which correlateswith the characteristic SRS image of the halo at 2850cm − in Fig. 3F. Therefore, it is also possible to use thisRaman band to locate the halo by SRS imaging in thehigh-frequency region. We also provide images based onthe same vibrations shown in Figs. 3C-H for two otherplaques in the hippocampus (SI, Fig. S6). CONCLUSION
In summary, we performed label-free multimodal imag-ing by nonlinear and vibrational microscopies to studyand present high-resolution images of A β plaques in thehippocampus and cortex of brain tissues of bitransgenicmice AD model. We validated the presence of A β plaquesby TPEF ThioS staining. While imaging with exoge-nous labels is undesirable, autofluorescence may be non-ideal in some applications due to the different overlap-ping signals. We presented SHG imaging of the coreand compared with TPEF ThioS staining and TPEA im-ages, which, together with an analysis of results in theliterature, suggest a protein-related origin of the SHGsignal. Also, using vibrational microscopies based onSpRS, CARS, and SRS, we studied both the core andthe halo of the plaques. While the CARS imaging isaffected by spurious backgrounds, we showed that SpRSand SRS microscopies produce virtually identical images,even in different experimental setups. This allows for astraightforward combination of Raman spectroscopy in-formation with the advantages of SRS microscopy. Wealso presented a halo biomarker based on unsaturatedlipids and two core biomarkers not yet reported in SRSmicroscopy studies of A β plaques: phenylalanine (Phe),in the low-frequency region, and amide B, in the high-frequency region. We compared the spatial distributionof the three core biomarkers. While the Phe image ap-pears to be more specific to the central part of the corethan amide I, amide B allowed us to obtain images withhigher acquisition rates than amide I and Phe. This re-sult suggests amide B as a novel fingerprint in the studyof A β plaques for diagnostic applications. MATERIALS AND METHODSAnimal model and tissue preparation
APP:PS1 double transgenic male mice (Tg) were pur-chased from the Jackson Laboratories. These mice havethe human amyloid precursor protein gene containingthe Swedish mutation K594N/M595L and delta E9 mu-tant human presenilin 1 gene, causing accelerated A β deposition in the brain after 5 months of age . For this study, we used mice of 6 and 12 months old. Themice were anesthetized and sacrificed, then had the brainremoved and kept overnight in 4% paraformaldehyde(PFA), followed by cryosectioning of the frontal cortexand hippocampus (40 µ m thick slices). These sliceswere washed three times with phosphate-buffered saline(PBS) and covered with a coverslip to spectroscopic anal-yses. After the multimodal imaging, tissue sections werestained using 1% ThioS solution (w/v) in 50% ethanol(v/v) (Sigma-Aldrich). The samples were immersed for8 minutes in ThioS solution, immediately immersed threetimes in 50% ethanol, washed with PBS, and seededdirectly onto coverslips for TPEF imaging. Figure 4Fshows a TPEF ThioS staining image of a sample obtainedfrom the hippocampus containing several plaques. Allimages in Fig. 1 refer to the one highlighted by the bluesquare in Fig. 4F. Our investigation is in agreement withthe Guide for the Care and Use of Laboratory Animalsand was approved by the Ethics Committee for AnimalUtilization in Research (CEUA) of the Federal Universityof Minas Gerais (protocol 225/2014), under the criteriaof the National Animal Experimentation Control Council(CONCEA). Optical effects and multimodal plataform
Figures 4A-D shows the exchange of energy diagramsfor the optical effects studied in this work. Fig. 4A showsthe TPEF, where a high-energy electronic state becomesaccessible through the simultaneous absorption of twolow-energy photons at frequency ω p . In this third-ordernonlinear effect, part of the absorbed energy is lost innon-radiative processes, while the other part is convertedinto a spontaneous fluorescent emission at frequency ω E .Due to the non-radiative processes, such fluorescent emis-sion occurs at a frequency lower than the sum of theabsorbed frequencies . Figure 4B shows the SHG, asecond-order nonlinear optical effect where the two low-energy photons at ω p can be up-converted into one pho-ton with exactly twice the incident frequency. This spon-taneous parametric up-conversion can only occur in ma-terials lacking inversion symmetry in the electric dipoleapproximation . Figure 4C shows the energy diagramfor two different, but related effects. When one laserbeam of frequency ω p interacts with a material, it can un-dergo a spontaneous inelastic scattering at the frequency ω S . In the case where ω s < ω p , the material is left inan excited vibrational state at the frequency Ω ν , whichcharacterizes the Stokes process of SpRS . When twolaser beams at frequencies ω p and ω S interact coherentlywith the material, additional third-order nonlinear opti-cal effects arise. If the beating frequency ( ω p − ω S ) isresonant with the vibrational transition at Ω ν , the beamat ω S stimulates the conversion of the beam at ω p into ω S , also at the cost of leaving the material in an excitedvibrational state. Such a process is the SRS . Si-multaneously with the SRS, the second beam at ω S also –17 | FIG. 4. Optical effects, experimental implementation, and Thioflavin validation. Energy level description for (A) TPEF, (B)SHG, (c) SpRS and SRS, and (D) CARS. (E) Experimental setup. (F) Left side: bright field image of the mouse brain slice(scale bar is 500 µ m). Right side: A β plaque highlighted in the TPEF ThioS staining image of the hippocampus region shownby the gray square in the bright field image (scale bar is 100 µ m). stimulates a four-wave mixing process that generates the ω aS frequency. In this case, the nonlinear generation of ω aS is orders of magnitude more likely to occur than thelinear generation of the anti-Stokes Raman scattering,which characterizes the CARS in the diagram of Fig.4D . All Raman-mediated effects probe vibrationalstates, therefore providing specific chemical informationabout the material.We obtained SpRS images and spectra using theWITec alpha300 SAR confocal system operating with a532 nm CW laser excitation, focused by a 60X oil objec-tive (numerical aperture, 1.4), and scanning with stepsof ∼
260 nm. We generated both images and spectra withthe Project 5 WITec software. All nonlinear optical ef-fects were studied through the setup shown in Fig. 4E.For TPEF and SHG, we used a pump beam of 180 fspulse-width and 80 MHz repetition rate, tuned at 810nm. The beam is transmitted through the dichroic mir-rors DM1 and DM2 (following the path of the purplebeam) and is focused by a 60X apochromatic oil objectiveO (numerical aperture, 1.4). The sample image is donethrough a scanning laser microscope (LaVision Biotec).The dichroic mirrors DM3 and DM4 direct the reflectedsignal generated by the material to the photomultipliers PMT1 and PMT2. Bandpass filters F1 and F2 can beinserted before the PMT1 and PMT2, respectively, allow-ing an unambiguous measurement of the effect. For theimplementation of SRS and CARS, the optical paramet-ric oscillator OPO (APE picoEMERALD) provides twobeams: a 5-6 ps pump beam of tunable frequency ω P anda 7 ps Stokes beam at a fixed frequency ω S , in blue andred, respectively, as in Fig. 4E. Frequency ω P is tuned sothat the beat frequency is resonant with the frequency ofthe vibrational mode. The dichroic mirror DM1 splits thebeams in a Mach-Zehnder configuration where the filtersFP and FS ensure no contamination of the pump beam(Stokes) in the arm of the Stokes beam (pump), respec-tively. The electro-optical modulator EOM connectedto the function generator FG produces a high-frequency(10 MHz) polarization-modulation in the Stokes beam.The dichroic mirror DM2 recombines both beams ina collinear configuration. Temporal synchronization isachieved through a delay line. The polarizer POL af-ter the recombination of the beams transforms the EOMpolarization-modulation into amplitude-modulation andensures that both the unmodulated (pump) and modu-lated (Stokes) beams have the same polarization state.The scanning mirrors SM allow the beams to map the –17 | sample when focused by the 60X apochromatic oil ob-jective O (numerical aperture, 1.4) in the scanning lasermicroscope. The backscattered CARS signal is then di-rected to the PMT1 photomultiplier by the dichroic mir-rors DM3 and DM4. A bandpass filter eliminates signalcontamination by TPEF. The forward SRS signal is col-lected by the collimator C and the filter FSRL eliminatesthe Stokes component of the beam. The telescope T re-duces the beam movement on the detector area of thephotodiode PD, where the SRS signal is extracted withthe help of the lock-in amplifier LIA coupled to the PDand connected to the EOM via the synchronization portof the FG. This setup allows the SRS signal to reach ahigh signal/noise ratio, as shown by previous works .After the measurements, we compare our findings witha TPEF image of the tissue stained with ThioS to val-idate the label-free techniques. This fluorescent label iswidely recognized as a dye in neurodegenerative diseasesstudies, particularly for visualizing amyloid plaques inAD models. . Figure 4F shows the bright field im-age of a section of the brain and the TPEF image of anenlarged section in the hippocampus. The blue squareidentifies the plaque studied in our multimodal analysisof Fig. 1. A saturated image is also provided (SI, Fig.S1), favoring the identification of the tissue and the otherA β plaques. AUTHOR CONTRIBUTIONS
R.C., L.L., and E.A.F. performed the experiments andanalyzed the data. R.C., L.L., E.A.F., A.B, A.J., andL.M.M. designed the experiments. M.R.S. provided theearly conceptual idea and contributed to the establish-ment of the colony of mice. R.C. and L.M.M wrote themanuscript. All the authors discussed the results andreviewed the manuscript.
CONFLICTS OF INTEREST
The authors declare no competing interest.
ACKNOWLEDGEMENTS
We acknowledge financial support from FINEP(01.13.0330.00), CAPES, Fapemig (TEC - RED-00282-16, APQ-03052-15), CNPq and CNPq project302775/2018-8.
REFERENCES
Psychiatry (Edgmont) ,2005, , 30–33.2 K. A. Jellinger, Journal of Neural Transmission , 2006, , 1603–1623. 3 A. Alzheimer’s,
Alzheimer’s & Dementia , 2018, , 367–429.4 M. Citron, Nature Neuroscience , 2002, , 1055–1057.5 E. A. Fonseca, L. Lafet´a, R. Cunha, H. Miranda, J. Cam-pos, H. G. Medeiros, M. A. Romano-Silva, R. A. Silva,A. S. Barbosa, R. P. Vieira, L. M. Malard and A. Jorio, Analyst , 2019, , 7049–7056.6 W. E. Klunk, B. J. Bacskai, C. A. Mathis, S. T. Kajdasz,M. E. McLellan, M. P. Frosch, M. L. Debnath, D. P. Holt,Y. Wang and B. T. Hyman,
Journal of Neuropathology &Experimental Neurology , 2002, , 797–805.7 D. Wilcock, M. Gordon and D. Morgan, Nature Protocols ,2006, , 1591–1595.8 M. E. McLellan, S. T. Kajdasz, B. T. Hyman and B. J.Bacskai, Journal of Neuroscience , 2003, , 2212–2217.9 C. Condello, P. Yuan, A. Schain and J. Grutzendler, Na-ture Communication , 2015, , 6176.10 P. Yuan, C. Condello, C. D. Keene, Y. Wang, T. D.Bird, S. M. Paul, W. Luo, M. Colonna, D. Baddeley andJ. Grutzendler, Neuron , 2016, , 724 – 739.11 C. Condello, A. Schain and J. Grutzendler, Scientific Re-ports , 2011, , 19.12 S. Liebscher and M. Meyer-Luehmann, Frontiers in Psy-chiatry , 2012, , 26.13 A. D. Cohen, M. D. Ikonomovic, E. E. Abrahamson, W. R.Paljug, S. T. DeKosky, I. M. Lefterov, R. P. Koldamova,L. Shao, M. L. Debnath, C. A. Mason, N. S. Mathis andW. E. Klunk, Letters in Drug Design & Discovery , 2009, , 437–444.14 W. Zipfel, R. Williams and W. Webb, Nature Biotechnol-ogy , 2003, , 1369–1377.15 J.-X. Cheng and X. S. Xie, Science , 2015, , aaa8870.16 T. Meyer, N. Bergner, C. Krafft, D. Akimov, B. Dietzek,J. Popp, C. Bielecki, B. F. M. Romeike, R. Reichart andR. Kalff,
Journal of Biomedical Optics , 2011, , 1–10.17 C. P. Pfeffer, B. R. Olsen, F. Ganikhanov and F. L´egar´e, Journal of Structural Biology , 2008, , 140 – 145.18 S. Heuke, N. Vogler, T. Meyer, D. Akimov, F. Kluschke,H.-J. R¨owert-Huber, J. Lademann, B. Dietzek andJ. Popp,
British Journal of Dermatology , 2013, , 794–803.19 N. Mazumder, N. K. Balla, G.-Y. Zhuo, Y. V. Kistenev,R. Kumar, F.-J. Kao, S. Brasselet, V. V. Nikolaev andN. A. Krivova,
Frontiers in Physics , 2019, , 170.20 H. Chen, H. Wang, M. N. Slipchenko, Y. Jung, Y. Shi,J. Zhu, K. K. Buhman and J.-X. Cheng, Optics Express ,2009, , 1282–1290.21 A. C. Kwan, K. Duff, G. K. Gouras and W. W. Webb, Optics Express , 2009, , 3679–3689.22 S. Wang, B. Lin, G. Lin, C. Sun, R. Lin, J. Huang, J. Tao,X. Wang, Y. Wu, L. Chen and J. Chen, Neurophotonics ,2019, , 1–11.23 S. Chakraborty, S.-T. Chen, Y.-T. Hsiao, M.-J. Chiu andC.-K. Sun, Biomedical Optics Express , 2020, , 571–585.24 J. Kiskis, H. Fink, L. Nyberg, J. Thyr, J.-Y. Li and A. Ene-jder, Scientific Reports , 2015, , 13489.25 J. H. Lee, D. H. Kim, W. K. Song, M.-K. Oh and D.-K.Ko, Journal of Biomedical Optics , 2015, , 1–7.26 M. Ji, M. Arbel, L. Zhang, C. W. Freudiger, S. S. Hou,D. Lin, X. Yang, B. J. Bacskai and X. S. Xie, ScienceAdvances , 2018, , eaat7715. –17 |
27 L. Bachmann, D. M. Zezell, A. da Costa Ribeiro, L. Gomesand A. S. Ito,
Applied Spectroscopy Reviews , 2006, ,575–590.28 M. Monici, Biotechnology Annual Review , 2005, , 227 –256.29 R. Richards-Kortum and E. Sevick-Muraca, Annual Re-view of Physical Chemistry , 1996, , 555–606.30 A. Moreno-Garc´ıa, A. Kun, O. Calero, M. Medina andM. Calero, Frontiers in Neuroscience , 2018, , 464.31 R. Boyd, Nonlinear Optics , Academic Press, London,2008.32 Y.-R. Shen,
The Principles of Nonlinear Optics , Wiley-Interscience, New York, 2003.33 P. Campagnola, A. Millard, M. Terasaki, P. Hoppe,C. Malone and W. Mohler,
Biophysical Journal , 2002, ,493–508.34 W. R. Zipfel, R. M. Williams, R. Christie, A. Y. Nikitin,B. T. Hyman and W. W. Webb, Proceedings of the Na-tional Academy of Sciences , 2003, , 7075–7080.35 N. H. Green, R. M. Delaine-Smith, H. J. Askew, R. Byers,G. C. Reilly and S. J. Matcher,
Scientific Reports , 2017, , 13331.36 M. Rivard, M. Lalibert´e, A. Bertrand-Grenier, C. Har-nagea, C. P. Pfeffer, M. Valli`eres, Y. St-Pierre, A. Pigno-let, M. Khakani and F. L´egar´e, Biomedical Optics Express ,2011, , 26–36.37 P. Hanczyc, M. Samoc and B. Norden, Nature Photonics ,2013, , 969–972.38 J. Tsai, J. Grutzendler, K. J. Duff and W.-B. Gan, NatureNeuronscience , 2004, , 1181–1183.39 H. B. Rajamohamedsait and E. M. Sigurdsson, Histolog-ical Staining of Amyloid and Pre-amyloid Peptides andProteins in Mouse Tissue. In: Sigurdsson E., Calero M.,Gasset M. (eds) Amyloid Proteins , Humana Press, 2012,vol. 849, pp. 411–424.40 L. Fu, Z. Wang, B. T. Psciuk, D. Xiao, V. S. Batista andE. C. Y. Yan,
The Journal of Physical Chemistry Letters ,2015, , 1310–1315.41 S. J. Roeters, C. N. van Dijk, A. Torres-Knoop, E. H. G.Backus, R. K. Campen, M. Bonn and S. Woutersen, TheJournal of Physical Chemistry A , 2013, , 6311–6322.42 J. Tan, J. Zhang, Y. Luo and S. Ye,
Journal of the Amer-ican Chemical Society , 2019, , 1941–1948.43 D. Kurouski, R. P. Van Duyne and I. K. Lednev,
Analyst ,2015, , 4967–4980.44 A. Barth and C. Zscherp,
Quarterly Reviews of Biophysics ,2002, , 369–430.45 J. W. Kinney, S. M. Bemiller, A. S. Murtishaw, A. M. Leis-gang, A. M. Salazar and B. T. Lamb, Alzheimer’s & De-mentia: Translational Research & Clinical Interventions ,2018, , 575 – 590.46 M. Heneka, M. O’Banion, D. Terwel and M. P. Kummer, Journal of Neural Transmission , 2010, , 919 – 947.47 E. Martin, C. Boucher, B. Fontaine and C. Delarasse,
Ag-ing Cell , 2017, , 27–38.48 R. G. Nagele, J. Wegiel, V. Venkataraman, H. Imaki, K.-C. Wang and J. Wegiel, Neurobiology of Aging , 2004, ,663 – 674.49 M. Ahmed, J. Davis, D. Aucoin, T. Sato, S. Ahuja,S. Aimoto, J. I. Elliott, W. E. Van Nostrand and S. O.Smith, Nature Structural & Molecular Biology , 2010, , 561–567.50 R. Cukalevski, B. Boland, B. Frohm, E. Thulin, D. Walshand S. Linse, ACS Chemical Neuroscience , 2012, , 1008–1016.51 A. Barth, Biochimica et Biophysica Acta (BBA) - Bioen-ergetics , 2007, , 1073 – 1101.52 K. A. Oberg, J.-M. Ruysschaert and E. Goormaghtigh,
European Journal of Biochemistry , 2004, , 2937–2948.53 C. Heinrich, A. Hofer, A. Ritsch, C. Ciardi, S. Bernet andM. Ritsch-Marte,
Optics Express , 2008, , 2699–2708.54 D. R. Borchelt, T. Ratovitski, J. van Lare, M. K. Lee,V. Gonzales, N. A. Jenkins, N. G. Copeland, D. L. Priceand S. S. Sisodia, Neuron , 1997, , 939 – 945.55 J. L. Jankowsky, H. H. Slunt, T. Ratovitski, N. A. Jenkins,N. G. Copeland and D. R. Borchelt, Biomolecular Engi-neering , 2001, , 157 – 165.56 M. Garcia-Alloza, E. M. Robbins, S. X. Zhang-Nunes,S. M. Purcell, R. A. Betensky, S. Raju, C. Prada, S. M.Greenberg, B. J. Bacskai and M. P. Frosch, Neurobiologyof Disease , 2006, , 516 – 524. –17 | Supplementary information for
Multimodal microscopy for characterization of amyloid- β plaquesbiomarkers in animal model of Alzheimer’s disease Renan Cunha, Lucas Lafeta, Emerson A. Fonseca, Alexandre Barbosa, Marco A. Romano-Silva, Rafael Vieira, AdoJorio, and Leandro M. Malard ∗ ∗ E-mail: lmalard@fisica.ufmg.br. –17 | Fig. S1. Tissue identification in the ThioS stained TPEF image. Same thioS stained TPEF image present in Fig. 1C of themain article but with saturated A β plaques, favoring the identification of the hippocampus and showing other plaques in a12-month-old animal. Left side: bright field image of the mouse brain slice (scale bar is 500 µ m). Right side: ThioS stainedTPEF image of the region shown by the grey square in the bright field image with several A β plaques (scale bar is 100 µ m).The blue square also identifies the plaque studied in Figure 1 of the manuscript. –17 | Fig. S2. Stimulated Raman microscopy versus spontaneous Raman spectra. Comparison between stimulated Raman scattering(SRS, blue) and spontaneous Raman scattering (SpRS, red) spectra obtained from mouse brain tissue images. The imageswere acquired from different scattering geometries (transmission for the SRS and reflection for the SpRS), but demonstrate ahigh spectral correspondence. –17 | Fig. S3. Comparison between high-resolution SRS images of amide I and phenylalanine and spatial distribution of both corebiomarkers. (A,B) merged image of the halo and core, demonstrating the spatial distribution of the response based on thefrequency of amide I (A) and phenylalanine (B). There is a spatial correlation in the distribution of both biomarkers in thecore, although they present appreciable differences. (C) the merged image of amide I (A) and phenylalanine (B), as well asthe halo, also demonstrate this distribution. In all figures, the halo image is based on the frequency attributed to lipids, 2850cm − . Images without a scale bar share the scale bar of the leftmost image in the same line, which is 20 µ m. –17 | Fig. S4. Subtraction of SRS images in the high-frequency region and comparison with SRS images based on the vibrationalmode of amide I. (A-D) different A β plaques distributed in three vibrations: 2930 (proteins/lipids), revealing both the core andthe halo; 2850 (lipids), with only the halo in evidence; and 1675 (amide I), with only the core in evidence. This last columnshould be compared to the subtraction column obtained by subtracting images at frequencies 2930 and 2850 cm − . The imagesbased on amide I and those obtained by subtraction show a significant spatial correlation. –17 | Fig. S5. Comparison between high-resolution SRS images of amide I and amide B and spatial distribution of both corebiomarkers. (A,B) merged image of the halo and core, demonstrating the spatial distribution of the response based on thefrequency of amide I (A) and amide B (B). There is a spatial correlation in the distribution of both biomarkers in the core;however, no appreciable difference can be ensured. (C) the merged image of amide I (A) and amide B (B), as well as the halo,also demonstrate this distribution. In all figures, the halo image is based on the frequency attributed to lipids, 2850 cm − .Images without a scale bar share the scale bar of the leftmost image in the same line, which is 20 µ m. –17 | Fig. S6. High-resolution SRS image with core and halo biomarkers in the high-frequency spectral region for two differentA β plaques.(A,F) and (B,G) are images based on the known 2930 cm − (proteins/lipids) and 2850 cm − (lipids) frequencies,respectively. (C,H) is the subtraction of these images, showing the core in the high-frequency region. (D,I) is the image basedon the 3070 cm − (amide B) frequency, which shows the core in the high-frequency region, correlating with the image obtainedby subtraction (C,H). (E,J) is the halo image based on the 3019 cm − frequency attributed to unsaturated lipids. Imageswithout a scale bar share the scale bar of the leftmost image in the same line, which is 20 µµ