Intrinsic variation of the polarization-resolved SHG from collagen: Multiscale analysis and application to parchments Open Access

Collagen is a major structural protein in mammals.¹ It is found in many organs such as skin, tendons, arteries, bones, and cornea and as such is involved in many pathologies. There are many collagen types, all of which are characterized by triple helical domains composed of three α chains at the molecular scale [Fig. 1(a)]. In collagen fibrillar types, these triple helices align to form fibrils with diameters ranging between 10 and 300 nm. Fibrils can further organize in various ways, most usually in fibers that are tight assemblies of aligned fibrils.¹ This so-called hierarchical organization of collagen is specific to each tissue and crucial for its biological function. Collagen remodeling in the course of pathologies such as fibrosis or cancer may affect this three-dimensional (3D) organization and impede the tissue function. Accurate multiscale characterization of collagen distribution in tissues is, therefore, a major biomedical issue.

Second harmonic generation (SHG) microscopy is nowadays the gold standard technique for collagen structural imaging with sub-micrometer resolution.² As SHG is a coherent multiphoton process specific for dense and non-centrosymmetric materials, it provides three dimensional (3D) imaging of fibrillar collagen without any staining and with unequaled specificity and sensitivity.³ Moreover, it can be combined with other multiphoton modalities, notably two-photon excited fluorescence (2PEF), to provide multimodal 3D images of thick collagen-rich tissues.² 

Polarization-resolved SHG microscopy (P-SHG) has been shown to provide more accurate information about collagen distribution in tissues.^4–16 P-SHG consists in recording series of SHG images for different orientations of the linear polarization of the incident laser excitation. A polarimetric diagram is thus measured in every voxel, and its main axis provides the mean direction of the collagen fibrils within this voxel. The so-called anisotropy parameter, which is noted as ρ (or alternatively γ), is also of great interest for collagen structural imaging: it is defined as the square root of the ratio of the SHG signal measured for an incident polarization parallel to the collagen fibrils to that for an incident polarization perpendicular to these fibrils. This anisotropy parameter has been shown to be directly related to the pitch of the collagen triple helix.^5,17

However, P-SHG measurement of the anisotropy parameter is a complex issue and has been shown to be affected by several experimental artifacts. First, collagen-rich tissues can exhibit birefringence and diattenuation that distort the incoming polarization and affect the polarimetric diagrams in depth.^4,6,7,18,19 Second, high focusing scrambles polarizations, which affects the polarimetric diagrams in a different way for forward vs backward SHG detection.²⁰ Third, these measurements can also be affected by the noise background.^21,22 Fortunately, all these experimental artifacts have been extensively studied and can be either corrected or at least estimated to provide reliable error bars.

In addition to these experimental measurement issues, the anisotropy parameter also exhibits intrinsic variations due to the variety of collagen hierarchical organizations in tissues. P-SHG microscopy probes this multiscale three-dimensional (3D) structure at the optical scale, within the focal volume (usually around 1 µm³) that may contain several fibrils with different directions. The anisotropy parameter has been shown to vary with the orientational disorder of these fibrils^{5,8–10,13,16} and with their out-of-plane orientation.^11–16 Both effects have been described theoretically, but tractable analytical equations describing both effects within the same formalism are still lacking. Moreover, these effects have been evidenced experimentally in different samples, which is prone to experimental artifacts.

In this work, we present a unified multiscale theoretical approach of the intrinsic variations of the anisotropy parameter measured in collagen using P-SHG microscopy. We then measure these variations in the very same collagen samples to get rid of experimental artifacts affecting the P-SHG measures. To that end, we use parchments, which are animal skins that are preserved by liming, scraping, and drying under tension.²³ Parchments thus contain almost solely collagen. Due to the last manufacturing step, the collagen fibers are stretched and well aligned parallel to the parchment surface, which is in the imaging plane.²⁴ However, a different finishing treatment can result in the presence of out-of-plane fibrils at the surface of some limited region of the same parchment. In the same way, artificial aging by heat for different durations of samples from the same parchment can induce increasing levels of disorder in the fibrillar structure. Our series of measurements in these parchments exhibit a good agreement with our theoretical approach and show the relevance of these P-SHG measurements of collagen multiscale organization in tissues.

Here, $E̲$ is the complex amplitude of the incident electric field, $E̲(t)=E̲e−iωt+cc$⁠, and the dependence on the position $r̲$ has been dropped for simplicity.

The susceptibility tensor χ⁽²⁾ is thus obtained as the summation of all hyperpolarizability tensors within a unit volume. Equally, it means that the SHG signal measured in a multiphoton microscope is obtained as the coherent summation of all SH electric fields radiated by the elementary nonlinear dipoles of each peptide bond within the focal volume.

Importantly, collagen exhibits a hierarchical organization from the α chain to the triple helix, the fibril, and the assembly of fibrils, as depicted in Fig. 1(a). Due to this multiscale structure, the summation of the hyperpolarizability tensors must be considered on two increasing scales.

• On the sub-micrometer scale that corresponds to the α chain, the triple helix or the fibril. On this scale, all the peptide bonds have the same polarity, so that the coherent summation of their contributions results in constructive interferences.

• On the micrometer scale, that is, the optical scale of the focal volume. On this scale, there are multiple fibrils with various 3D orientations, including opposite polarities, which can lead to partially destructive interferences. The resulting polarization thus depends on the precise spatial distribution of the collagen fibrils within the focal volume, which is specific to each tissue. Furthermore, this assembly of fibrils may have different orientations with respect to the light propagation axis Z of the microscope.

In many tissues such as skin and tendon, fibrils assemble to form fibers that can be considered as tight assemblies of aligned fibrils.¹ Because parchments are made from animal skin composed of collagen fibers, SHG microscopy probes these fibers at the micrometer scale. Whatever the scale under consideration, the collagen susceptibility tensor χ⁽²⁾ is calculated from the peptide bond hyperpolarizability β in two steps.

• Changing the frame of reference from the peptide bond frame to the α chain/triple helix/fibril frame or from the fibril frame to the laboratory frame

• Summation over all the peptide bonds within the α chain/triple helix/fibril or over all fibrils within the focal volume

This anisotropy parameter ρ_fib is thus a constant characterizing the collagen triple helical structure.

Note that this calculation can be generalized to the strong focusing regime, beyond the plane wave approximation.^19,20 In this regime, numerical calculations based on Green’s function formalism have to take into account the tensor components with a Z coordinate, in which the calculation is similar to Eq. (9).

We now consider specific fibril assemblies that correspond to common geometries in tissues. The aim is to simplify Eq. (9) to gain insight into the effect of multiscale collagen organization on the P-SHG signal.

P-SHG imaging of collagen samples exhibiting well-aligned fibrils in the imaging plane thus provides a measure of ρ_fib. Such measurements have been reported in stretched tendons and corneal lamellae and give values of 1.36 and 1.45, respectively.^8,20 This corresponds to a pitch angle of around 50°, close to the pitch angle of 45° measured by x-ray diffraction.²⁶ The difference is attributed to mainly three effects: (i) the nonlinear dipole may be not perfectly aligned along the peptide bond, as shown by quantum-chemistry simulations;^27–29 (ii) the pitch angle may vary with the hydration state of collagen;³⁰ and (iii) the triple helical structure leads to a dispersion of the peptide bond angles to the main axis. Moreover, fibrils can exhibit a super-coiled structure that also leads to a dispersion of the peptide bond angles.^31,32 These effects can be taken into account by a more complex calculation and provides a better agreement with the collagen structure.^5,13,33

It increases from ρ_fib to 3 as the out-of-plane angle ψ increases from 0 to $π2$⁠.^4,11,12

This calculation also applies to a disordered assembly of collagen triple helices within a single fibril within the focal volume, that is a partial loss of the fibril structure.

We have introduced the stage frame (X₀, Y₀, Z), where X₀ is the reference axis for the incident linear polarization in the microscope and corresponds to the X₀ axis of the microscope stage and of the laser scanning direction [Fig. 2(b)]. It can be different from the X axis that is attached to the collagen sample and corresponds to the direction of the fibril assembly in the imaging plane (Fig. 1). We thus introduce the angle $α0=X0X̂$ [Fig. 2(b)].

In the specific case of well-aligned fibrils with out-of-plane orientation ψ, Eq. (12) shows that $χXXX(2)$ and $χXY Y(2)$ are both proportional to cos ψ. As a consequence, I_SHG is proportional to cos² ψ: the SHG signal decreases as the out-of-plane orientation increases.^11,12

Importantly, Eq. (18) does not hold for more complex tissues, such as disordered assemblies of collagen fibrils oriented out of the imaging plane. In the latter case, there are additional terms and the parameter ρ has no direct physical interpretation.

Multiphoton imaging was performed using a custom-built laser scanning upright microscope, as previously described³⁴ [Fig. 2(a)]. Briefly, the excitation source is a femtosecond Ti-Sa laser tuned to 860 nm and scanned in the X₀Y₀ directions using galvanometric mirrors. It is focused by a 10× air objective (TL10X-2P, LWD Super Apochromat, Thorlabs) with a high working distance: WD = 7.77 mm to guarantee non-contact imaging even for not perfectly plane parchments, and a medium numerical aperture: NA = 0.5 to enable analyses in the plane wave approximation. The resolution (FWHM) was measured as 0.7 μm in the lateral direction and 5.8 μm in the axial direction,³⁵ and the full field of view was 750 × 750 μm². The excitation power at the objective focus was adjusted between 10 and 30 mW. We checked that no degradation of the parchments was induced under these conditions by verifying the reproducibility of the SHG measurements in a set of images recorded sequentially in the same area. SHG and 2PEF signals were epi-detected by photon-counting photomultiplier tubes (P25PC, Sentech) using dichroic mirrors (695DCXRU, Chroma and FF458-Fi01, Semrock) and appropriate spectral filters (FF01-680/SP and FF01-427/10, Semrock, for the SHG signal).

First, Z-stacks of en face SHG images were recorded with circularly polarized excitation, using 200 kHz pixel rate, 0.78 × 0.78 μm² pixel size, and 1 μm Z-step over 20–70 μm depth. Transverse reslices were then numerically processed [Fig. 2(b)]. In the following, these SHG images are displayed using a green look-up table and correspond to maximum intensity projections of the Z-stack for en face images or along a transverse direction for transverse reslices.

Second, P-SHG imaging was performed by controlling the orientation of the linear polarization of the incident laser beam by two motorized achromatic waveplates at the back pupil of the objective [Fig. 2(a)]. First, a quarter waveplate (MRAC2/40070707M, Fichou, France) was used to compensate for any ellipticity introduced by the microscope optical components and obtain a well-defined linear polarization: the resulting ellipticity, defined as the ratio of the minimal to maximal electric fields, was less than 3% in a central reduced field of view (FOV) of 351 × 351 μm². Second, a half waveplate (MRAC2/20070707M, Fichou, France) was used to rotate this linear polarization from 0° to 170° using 10° steps. A P-SHG image stack was recorded in the reduced FOV at every depth Z using these 18 incident polarizations. The pixel rate and voxel size were the same as for SHG imaging with circularly polarized excitation.

The precision of these measurements is related to the total number of detected photons.²¹ We first apply a 2 × 2 binning filter to all images to improve the signal to noise ratio, so that the precision is around 0.05 for ρ under our experimental conditions.

The parameter R² is used to determine the pixels considered as valid for subsequent quantitative analyses: pixels with R² < 0.7 are filtered out and depicted as black pixels in the images. We thus obtain R²-filtered maps of $SHG$⁠, α₀, and ρ at every depth Z.

We first perform a global quantitative analysis of the ρ parameter. The percentage of valid pixels (i.e., pixels with R² > 0.7) is calculated at every depth Z and a threshold is set at 10% and used to select valid Z-planes. The ρ parameter is then analyzed as follows: (i) the values of ρ are averaged over the valid pixels at every valid depth Z; (ii) these averaged values are averaged over all valid Z-planes to obtain a mean value and a standard deviation of ρ in the Region Of Interest (ROI) under study. The standard deviation is calculated in the same way. Several ROIs can be studied in the same parchment. This quantitative analysis is performed using homemade routines written in MATLAB or ImageJ.

The ρ parameter is also analyzed locally in selected fibers of interest 3D renderings. Instance segmentation of fibers is automatically performed using the following Python pipeline: application of a median filter (σ_filter = 3), Otsu thresholding, sphere binary erosion, masked Voronoi labeling, and removing of small labels using CLIJ via the pyclesperanto-prototype Python library⁴⁰ (supplementary Fig. S3). The mean value and standard deviation of ρ are calculated on the mask of selected fibers of interest using regionprops from skimage Python library.⁴¹ The out-of-imaging-plane angle ψ of every fiber is then calculated by using coordinate points placed by hand on the Z upper and lowest pixels in the fiber using Napari point layer.⁴² The measurement error is estimated to be roughly 10°.

Two parchments are used in this study and imaged on the flesh side, the grain side, or both. The first one is a goat parchment, which was crafted in 2020 by one of the authors (L. Robinet). Two types of finishing treatments were applied in two different areas on the flesh side (supplementary Fig. S2a). The area that was just scrapped while the skin was wet is called Scrapped, and the area that received chalk as the parchment was wet and sanded once dry is called Chalk and sanded. For each finishing treatment, three samples of 1 × 2 cm² were collected. Each sample was imaged in four different ROIs on the flesh side. In total, 4 × 3 = 12 Z-stacks of P-SHG images were thus recorded for each finishing treatment (supplementary Table T1).

The second parchment under study is an ancient sheep parchment dating from 1816 (supplementary Fig. S2b). Three samples of 1 × 2 cm² were collected: one was kept as a reference (reference, D0), while the two others were artificially aged by heat in an oven at 100 °C for 14 days (D14) and 32 days (D32). For each side of the three samples, four Z-stacks of P-SHG images were recorded in different ROIs. In total, 4 × 2 = 8 Z-stacks of P-SHG images were thus recorded for each of the three heat exposure durations (supplementary Table T2). Each sample was analyzed by differential scanning calorimetry (DSC) to assess the degradation state of the collagen by the measurement of the shrinkage temperature (Ts).⁴³ Parchment pieces of 1 mm² were soaked in water for 1 h, sealed in a 30 μl aluminum capsule, and heated from 5 to 120 °C at 10 °C/min on a Perkin Elmer DSC 8000. The measurements were repeated three times for each sample and averaged.

The parchment samples were placed under the objective of the microscope and maintained flat using lead weights (supplementary Fig. S2c). The multiphoton analyses were non-contact, non-destructive, and carried out without any labeling.

To study experimentally the variation of the anisotropy parameter ρ in an assembly of out-of-plane fibrils vs an assembly of in-plane fibrils, we choose to use the modern goat parchment because the two different finishing treatments on the flesh side result in different organizations of the collagen fibers at the parchment surface on a macroscopic scale. The Scrapped area surface is smooth and flat compared to the Chalk and sanded area, where collagen fibers are oriented out of the sample plane with various angles, as clearly observed in white-light side images of the sample sections [Figs. 3(a) and 3(b)].

Figures 3(c)–3(f) display SHG images of the two types of parchment that correspond to XZ and XY projections of en face Z-stacks. All SHG images show long and straight structures attributed to collagen fibers, as previously reported.^24,34 En face SHG image of the Scrapped area [Fig. 3(e)] shows disordered collagen fibers with homogeneous distribution, and the transverse reslice shows that all the fibers lie within the imaging plane [Fig. 3(c)]. In contrast, en face SHG image of the Chalk and sanded area [Fig. 3(f)] shows a less homogeneous collagen distribution, and the transverse reslice shows that some fibers are spread out of the imaging plane with random orientations [Fig. 3(d)].

Figures 3(g) and 3(h) display maps of the anisotropy parameter in a reduced FOV [dotted yellow square in Figs. 3(e) and 3(f)] and at a specific Z depth [dotted yellow line in Figs. 3(c) and 3(d)]. The Z plane chosen for the Scrapped sample is the one with the highest SHG signal. For the Chalk and sanded sample, the Z plane is selected to view a maximum number of out-of-plane fibers, which corresponds to a higher Z position than for the Scrapped sample. We observe that ρ is larger in the Chalk and sanded map [in Fig. 3(h)] than in the Scrapped one [Fig. 3(g)].

This observation is quantified by calculating the average values of ρ in every Z-stack of P-SHG data. The results for the two types of sample are quite reproducible for all ROIs under study (12 ROIs per type of sample), as displayed in Fig. 3(i) and listed in supplementary Table T1. We obtain ρ_Scrapped = 1.33 ± 0.02 and ρ_{Chalk and sanded} = 1.61 ± 0.04. It confirms that the presence of the out-of-plane fibers in the Chalk and sanded sample results in a larger anisotropy parameter ρ than in the Scrapped sample that contains only in-plane fibers, in agreement with our theoretical calculation [Eq. (13)].

As the Chalk and sanded sample contains both in-plane and out-of-plane fibers [Fig. 3(d)], a more accurate approach consists in a local analysis of the anisotropy parameter ρ and the out-of-plane orientation ψ of these two types of fibers. For that purpose, we use a transverse reslice of a P-SHG image with a large Z range in order to visualize out-of-plane fibers over their full length [Fig. 4(a)]. In-plane fibers are also easily identified as they are grouped at the bottom of the SHG image. When selecting an out-of-plane fiber within a transverse plane, the associated anisotropy parameter map clearly shows that ρ is higher in this fiber than in the fibers parallel to the imaging plane [Fig. 4(b)]. Connected component analysis in Python then allows the processing of local averages of ρ in fibers or a group of fibers, which can be plotted as a function of the out-of-plane angle ψ measured by using the coordinates of the voxels at the two fiber ends. Figure 4(c) displays these eight measurements, where the ψ angle varies from 10° to 25° with an estimated measurement error of 10°. They are superimposed on the theoretical behavior of $ρoptout(ψ)$ given by Eq. (13), where ρ_fib is set to 1.36^8,44 as in previous reports or to 1.48 that is the average measurement in the in-plane fibers of this sample. A good agreement is observed between the experimental measurements and the theoretical curves.

Let us now address the variation of ρ as a function of disorder. For this study, a series of samples were artificially aged by heat exposure to mimic collagen degradation through oxidation.²³ If the collagen is in an advanced stage of degradation, it eventually leads to the denaturation of collagen into gelatin that is the unfolding of the collagen triple-helical structure at the molecular scale. In this extreme case, no more SHG signals are detected due to the loss of the hierarchical organization.³⁴ However, early steps of degradation only correspond to a disorganization of the collagen structure that preserves the fiber distribution.^23,34

The state of degradation of collagen in parchments is first characterized by means of DSC measurements in every sample. As expected, the three samples under study exhibit a decreasing shrinkage temperature T_s for increasing heat aging duration: 51.5 ± 2.8 °C for the reference sample D0, 38.1 ± 0.2 °C for the D14 sample, and 35.3 ± 3.1 °C for the D32 sample (supplementary Table T2). It is attributed to a progressive disorganization of collagen when the duration of the heat exposure increases, as already observed in previous studies.^23,34 This disorganization is expected to include an orientation disorder in the fibril assembly within fibers as depicted in Fig. 1(f) and a disorder in the molecular packing within every fibril.⁴⁵ 

P-SHG stacks recorded on the grain and flesh sides of each sample are displayed in Fig. 5. SHG images of the D0 and D32 samples show similar distribution of the collagen fibers at the millimeter scale [Figs. 5(a) and 5(c), projections of the Z-stacks]. However, note that the SHG signal decreases from D0 to D32, as already reported,³⁴ which is not visible here because the look-up table is defined arbitrarily for every sample. This decrease is consistent with a loss of collagen order at the submicrometer scale. The associated ρ maps at a specific Z plane and in the same ROIs show larger values for the D32 sample than for the D0 sample [Figs. 5(b) and 5(d)]. The same behavior is observed in all P-SHG stacks on both the flesh and grain sides (4 ROIs per condition), as displayed in Fig. 5(e) and listed in supplementary Table T2. It is consistent with the theoretical variation of ρ as a function of disorder that is expressed in Eq. (15) and plotted in Fig. 5(f). ρ increases from the ρ_fib value to 2.50 for a von Mises distribution or 2.48 for a Gaussian one, when the variance σ of the fibril orientation distribution increases. As this theoretical calculation also applies to an orientation disorder of collagen molecules within a single fibril, we attribute the increase in ρ in the degraded parchments to an increase in disorder at the sub-micrometer scale.

In this paper, we have developed a comprehensive study of the variation of the anisotropy parameter ρ in the P-SHG microscopy of collagen. This parameter quantifies the contrast of the polarimetric SHG signal in collagen tissues: it is equal to the square root of the ratio of the SHG signal for incident linear polarization parallel to the collagen fiber to that for incident polarization perpendicular to the fiber.

First, we have performed a theoretical multiscale analysis of the variation of ρ as a function of the distribution of the collagen fibrils within the focal volume. We have obtained analytical expressions valid for any collagen distribution and then considered three model distributions leading to simplified equations: (i) assembly of aligned fibrils within the imaging plane, (ii) assembly of aligned fibrils oriented out of the imaging plane, and (iii) assembly of disordered fibrils within the imaging plane (Fig. 1). Our calculations show that the anisotropy parameter ρ increases when the out-of-plane angle or the disorder increases, starting from its value of around 1.4 for a unique fibril lying within the imaging plane up to 3 for out-of-plane fibrils or around 2.5 for disordered fibrils. Our analytical approach shows two main limitations: (i) it is performed in the plane wave approximation, while moderate to strong focusing is used in SHG microscopy, and (ii) it only considers model collagen distributions that do not reflect the complexity of collagen organization in tissues. To address these two limitations, more advanced numerical approaches can be developed, for instance, using Green function formalism and modeling complex 3D geometries of collagen.^19,46 It would enable to consider assemblies of fibrils with different sizes that cross at any angle and exhibit a super-helical structure.^32,38,46 However, the advantage of our analytical approach with model distributions is to provide simple expressions that can be used in a straightforward way to analyze experimental data and get insight into the collagen structure. Moreover, as discussed in the following, we obtain a good agreement with experimental P-SHG imaging of parchment with a 0.5 NA air objective, which shows that the plane wave approximation provides reliable results for moderate focusing.

Second, we have performed P-SHG microscopy experiments in parchments in order to verify the predictions of our theoretical calculations. Parchments are relevant collagen materials for this study because they are obtained from animal skins that have been dried under tension after the removal of mainly hair and fat. Parchments are, therefore, mainly composed of stretched collagen fibers parallel to the parchment surface, which is parallel to the imaging plane. These fibers are thus appropriate models for tight assembly of aligned in-plane fibrils. Moreover, specific finishing treatment with chalk and pumice, performed to remove the remaining fat and to have a white and opaque surface suitable for writing, can result in the presence of out-of-plane oriented fibers at the surface of the parchment. These type of parchments are then appropriate models for the tight assembly of aligned out-of-plane fibrils. Finally, early steps of degradation by heat have been shown to induce disorder in the fiber structure, so that heat-degraded parchments are appropriate models for the assembly of disordered in-plane fibrils.³⁴ Advantageously, tight assemblies of aligned in-plane and out-of-plane fibrils are obtained from the same parchment, which is separated in two regions subjected to two different finishing treatments. Similarly, several samples are collected in the same parchment and exposed to heat for different durations. Each of these 2 series of P-SHG measurements is thus performed in a unique parchment, which ensures reliable comparison with the reference measurement of in-plane aligned fibrils. Indeed, ρ measurements may result in different values from one parchment to the other because of the variability of the animal skins from which they are made and of the differences in the parchment manufacturing.

Our theoretical calculations and experimental P-SHG data exhibit a good agreement. It demonstrates that an increase in the anisotropy parameter ρ can be attributed to an increase in the disorder or in the out-of-plane angle and that ρ is affected with the same order of magnitude in both cases. It may thus be difficult to discriminate between disorder and out-of-plane orientation, both in thin sections and in dense tissues, where transverse reslices do not enable the visualization of the fiber out-of-plane orientation. However, in most tissues, this out-of-plane orientation can be characterized in transverse reslice or 3D rendering, so that an increase in ρ can be attributed to disorder or not in a reliable way. It is notably the case in parchments that exhibit only in-plane fibers.³⁴ 

This study aimed at implementing a comprehensive study of the variation of the anisotropy parameter ρ, which quantifies the polarimetric contrast in the P-SHG microscopy of collagen. We developed an analytical multiscale approach of collagen P-SHG and demonstrated that the anisotropy parameter ρ increases when the disorder or the out-of-plane angle increases. These theoretical calculations were confirmed by P-SHG experiments in a series of parchments samples exhibiting different geometries of collagen fibers. Our study, therefore, provides an efficient way to attribute a variation of ρ to a variation of the collagen distribution and to get insight into the collagen structure in different tissues. It may be generalized to other protein assemblies that exhibit SHG signal, such as myosin and tubulin, although it would require to take into account the specific structure of these proteins.¹⁶ 

The supplementary material encompasses Supplementary Fig. S1: P-SHG imaging and processing; Supplementary Fig. S2: Parchments under study; Supplementary Fig. S3: Processing pipeline of out-of-plane fiber images; Supplementary Table 1: P-SHG data for the handmade goat parchment with different finishing treatments; and Supplementary Table 2: P-SHG data for the ancient sheep parchment. Artificial aging was induced by heat exposure (100 °C) during 0 (reference), 14, and 32 days.

G.G., G.L., C.C., and M.-C.S.-K. thank Xavier Solinas and Jean-Marc Sintès for technical support in the SHG setup and Nicolas Olivier and the “Advanced microscopy and tissue physiology” group at LOB for fruitful discussions. L.R. thanks Ji$ř$í Vnouek for the training on the parchment manufacturing and Olivier Deparis and the University of Namur for the organization of the parchment manufacturing workshop in September 2020.

G.G., G.L., C.C., and M.-C.S.-K. acknowledge fundings from the French “Agence Nationale de la Recherche” (Nos. ANR-10-INBS-04, ANR-11-EQPX-0029, and ANR-21-ESRE-0050). GG is funded by the MITI-CNRS.

The authors have no conflicts to disclose.

Giulia Galante: Formal analysis (equal); Investigation (equal); Software (equal); Writing – original draft (equal). Laurianne Robinet: Conceptualization (equal); Funding acquisition (equal); Methodology (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). Sylvie Heu-Thao: Investigation (supporting); Methodology (equal); Writing – review & editing (equal). Clément Caporal: Methodology (equal); Software (equal); Writing – review & editing (equal). Gaël Latour: Conceptualization (equal); Funding acquisition (equal); Methodology (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). Marie-Claire Schanne-Klein: Conceptualization (lead); Funding acquisition (equal); Methodology (equal); Supervision (equal); Validation (equal); Writing – original draft (lead).

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the corresponding author upon reasonable request.