Triacylglycerols affect the water content and cohesive strength of collagen fibrils

Abstract

The function of lipids in collagen fibrils is not well understood. The types and amounts of lipids present in collagen fibrils are particularly unknown. Triacylglycerols (TAGs) are a major class of lipids found in tendons and connective tissue. In this study, we measured the proportion of TAGs in single collagen fibrils obtained from chicken calcaneal tendons and investigated how TAGs affect the mechanical properties of collagen fibrils. An analytical protocol using 3D depth profiling with atomic force microscopy (AFM) was developed to measure the proportion of TAGs in single collagen fibrils. We found that collagen fibrils from chicken tendons contain an unexpectedly large amount of TAGs, on average 9% of the dry fibril volume, and that TAGs act as plasticizers, softening the fibrils and reducing their water content. Spectroscopic data from Raman and ¹H NMR revealed the chemical identity of TAGs in collagen fibrils. We propose that TAGs are irregularly intercalated within the collagen crystal lattice and display liquid-like disorder. This challenges the current understanding of the chemical composition of native collagen fibrils. Furthermore, our findings demonstrate that structurally embedded lipids, such as TAGs, can modulate cohesive forces between proteins at the molecular level.

---

Introduction

Collagens are the most abundant proteins in vertebrates and provide mechanical strength to connective tissues.¹ Collagen-like peptides have been instrumental in elucidating the 3D molecular structure of collagens.² Recent experimental and theoretical work showed that the alkyl chain of alkylated collagen-like amino acids stabilizes the triple helix of collagen-like peptides.^3,4 The proposed molecular mechanism is an alignment of alkyl chains along the hydrophobic groove of collagen-like peptides. This raises the question of whether and how natural lipids associate with collagen fibrils in vivo, and how this affects the biomechanical properties of collagen fibrils, which are the fundamental tensile load-bearing elements in the connective tissues of all vertebrates.¹ Collagen fibrils are a natural macromolecular material that combine high tensile strength and flexural flexibility.¹ In synthetic polymeric materials, the addition of low-molecular-weight additives (plasticizers) is often used to tailor the mechanical properties of a material to specific application needs.⁵ With this in mind, we asked what influence natural lipids might have on the mechanical properties of collagen fibrils. The interaction with water molecules is also important. Native collagen fibrils contain approximately 60% water by weight,^6–9 which acts as a plasticizer to make the fibrils soft and flexible.

Lipids are a diverse group of amphiphilic molecules with different physicochemical properties and various biological functions.¹⁰ They are closely associated with collagen molecules, as is known from the chemical purification of collagen from natural sources.^11–13 However, the only model describing the 3D molecular structure of type I collagen fibrils in native tendon based on X-ray diffraction data¹⁴ (PDB ID: 3HR2) contains no lipids intercalated within the collagen lattice structure. Triacylglycerols (TAGs) are the major lipid component of tendons, cartilage, and other connective tissues.¹⁵ Lipids are found in the tendon ground substance, which embeds the densely packed type I collagen fibrils,¹⁶ and in the synovial fluid, which acts as a lubricant between the tendon and the tendon sheath. In cases of obesity and other diseases, cholesterol and cholesterol esters accumulate in the interfibrillar matrix between collagen fibrils, where they can form granular domains (xanthomata).^17–19 In such cases, cholesterol and cholesterol esters are intercalated into the collagen crystal lattice, a phenomenon that is considered pathological.^20,21 However, the lipid content of single collagen fibrils has not been studied.

Here, we report an analytical protocol to measure the proportion of lipids in single collagen fibrils. The protocol uses 3D depth profiling with atomic force microscopy (AFM) to accurately measure the shape of collagen fibrils when they are partially covered by fluids.^22,23 The key lies in a washing sequence in which a non-polar solvent (hexane) first removes lipids adhering to the fibrils, followed by a polar solvent mixture (dichloromethane/methanol) that extracts lipids from the fibril interior. After each washing step, we used AFM to measure the resulting changes in cross-sectional area, water content, and indentation modulus of each collagen fibril.^22,23 The change in cross-sectional area yields the volume fraction of lipids that was removed from each collagen fibril by washing with solvents. Raman spectroscopy and ¹H NMR were used to identify the extracted lipids as triacylglycerols (TAGs).

Results and discussion

Intermittent contact (IC) mode AFM images of collagen fibrils extracted from healthy chicken calcaneal tendon, deposited on a Si substrate, and rinsed with deionized water, as detailed in the Methods Section, show dark areas adjacent to the collagen fibrils (Fig. 1(a)). This substance is liquid, as force–distance (FD) measurements with AFM and reconstructed 3D depth profiles show (Fig. 1(e) and (c), respectively). From FD data measured during tip approach, we determine the position d₀, where the attractive force sets in, and the position d_S, where a repulsive force starts to act on the AFM tip. The approximately constant attractive force between d_S and d₀ is caused by the capillary interaction with the AFM tip. During tip retraction, a capillary bridge is pulled-off the surface, and the attractive capillary force continues to act on the tip until the capillary bridge collapses at d_off. Such type of FD behavior is characteristic for a fluid on a solid substrate.^22,24

Fig. 1 (a) and (b) 3D-rendered AFM height images of collagen fibrils, with phase images serving as color texture, after washing with H₂O (a) and after washing with heptane (b). (c) and (d) 3D depth profiles of the tip–sample force constructed from AFM-based nanoindentation data collected during tip approach after washing the fibrils with water (c) and after removing the lipids with DCM/MeOH (d). Attractive forces (blue) correspond to fluid lipids, repulsive forces (red) to solid materials (collagen fibril and substrate). (e) and (f) Corresponding FD data measured next to the fibril within the lipid region after washing with water (e) and after removing the lipids (f).

The fluid substance adhering to the collagen fibrils can be washed off with non-polar solvents (hexane, heptane) without changing the morphology of the fibrils (Fig. 1(b) and (d)), suggesting that it is composed of hydrophobic lipids. For Raman and ¹H NMR spectroscopic analysis, a calcaneal tendon was cut into smaller pieces and washed with a 2 : 1 (v/v) mixture of dichloromethane and methanol (DCM/MeOH), a lipid extraction procedure. The Raman spectrum of the extracted substance is characteristic of TAGs, as the comparison with pure triolein, a TAG with three identical alkyl chains, and Raman spectra reported in the literature^25,26 reveals (Fig. 2(a) and (b)). TAGs are the most abundant form of glycerolipids, which display a large variation in length and number of alkyl chains, as well as in the number and the position of C=C double bonds within the alkyl chains. The details depend on tissue type and species.^10,15,27,28 Therefore, we measured Raman spectra of three reference compounds: triolein with three identical alkyl chains (Fig. 2(b)), 1,3-diolein with two identical alkyl chains, and monocaprin with only one, saturated alkyl chain. The Raman spectrum of 1,3-diolein is very similar to that of triolein (Fig. S1, SI), but the detailed shape of the carbonyl vibrational band ν(C=O) differs markedly from those of triolein and monocaprin (see the inset in Fig. 2(a)). Furthermore, monocaprin is crystalline at ambient temperature, resulting in sharper peaks in the Raman spectrum (Fig. S1(a), SI). Thus, Raman spectroscopy shows that TAGs are the major component of the substance extracted from tendon tissue with DCM/MeOH. ¹H NMR spectroscopy confirms this finding. The majority of the NMR signal originates from TAGs, as the comparison with the NMR spectrum of triolein shows (Fig. 2(c)). The minor signals present in the NMR signal of lipids extracted from tendon and marked with an asterisk (*) are attributed to other natural lipids according to previous work on the chemical composition of lipids in tendons, ligaments, and joints.¹⁵

Fig. 2 (a) Raman spectra of lipids from chicken calcaneal tendon (blue) and triolein (red). Vibrational bands are assigned according to the literature.^25,26 Inset: Carbonyl stretching vibration ν(C=O) for lipids from tendon (L) and the reference compounds monocaprin (1), 1,3-diolein (2), and triolein (3). (b) Molecular structure of triolein. (c) ¹H NMR spectra of lipids from tendon and triolein in CDCl₃. Peaks A₁ to J correspond to triolein^29,30 and are also found in the spectrum of lipids extracted from tendon. Peaks marked with S are due to residual solvents.³¹ See Fig. S2 (SI) for a more detailed assignment of peak positions.

Thermodynamically, TAGs and collagen fibrils constitute two coexisting phases. Inspired by the Flory–Huggins theory of polymer solutions and the thermodynamics of binary mixtures, we expect the fibrils to contain a small amount of TAGs. Based on our methodology for 3D depth profiling fluid polymer solutions²² and collagen fibrils with AFM,²³ we designed an analytical protocol to measure the amount of TAGs in individual collagen fibrils: we first dissolved the lipids surrounding the fibrils using the non-polar solvent hexane. In the second step, we used a 2 : 1 v/v mixture of the polar solvents DCM/MeOH to dissolve the lipids that are embedded within the collagen fibrils (Fig. 3(a)). The DCM/MeOH mixture resembles the classic chloroform/MeOH mixture used by Folch and coworkers^32,33 and Bligh and Dyer,³⁴ which is commonly used for lipid extraction. However, DCM is less toxic than chloroform.³⁵ The water miscibility and hydrogen-bonding ability of MeOH is required in order to split lipid–protein complexes in the sample.¹⁰ Hexane is a common solvent used to extract natural oils and fats,³⁶ which are primarily composed of TAGs. This shows that TAGs are soluble in hexane and the DCM/MeOH mixture. Moreover, the solvents are applied in excess. The solubility of hexane in water is very low (9.5 mg L⁻¹).³⁷ Thus, we suppose that hexane does not diffuse into the polar collagen fibrils.

Fig. 3 (a) Washing sequence for the measurement of the volume fraction of TAGs in a single collagen fibril. (b) Relative change in cross-sectional area of individual collagen fibrils in dry air after washing first with hexane (Hex) and then with DCM/MeOH (D/M). Data points are the mean of tendons 1–3 (shown in black, red, and blue, respectively), error bars are the 95% confidence interval of the mean.

After each washing step, we measured the resulting changes in fibril cross-section, water content, and indentation modulus of each collagen fibril with AFM.^22,23 The change in cross-sectional area yields the volume fraction of TAGs, f_TAG, that was removed from each collagen fibril by washing with hexane and DCM/MeOH (Fig. 3). Because native collagen fibrils vary in diameter and mechanical properties,^23,38,39 we examined a random sample of 40 fibrils taken from three tendons of three different chickens and measured 3D depth profiles of each fibril at the same fibril location after each washing step (Fig. S3 and S4).

In the 3D depth profiles that we reconstructed from the FD data (Fig. 1(c)), the regions of attractive forces (blue, corresponding to the liquid TAGs) and repulsive forces (red, corresponding to the solid collagen fibril and the solid substrate) can be distinguished. This allowed us to accurately measure the shape of the collagen fibril (the d_S height profile, Fig. S4, SI) beneath the liquid TAGs that adhered to the fibrils. This yielded the cross-sectional area A₀ in the dry state and the cross-sectional area A after washing with hexane and DCM/MeOH.

The volume fraction of material washed out of the fibril is the TAG volume fraction, f_TAG = 1 − A/A₀. The remaining area fraction f = A/A₀ in the dry state after washing with DCM/MeOH was assigned to the volume fraction of collagen f = 0.906 ± 0.049 (mean ± SD, n = 40). This shows that the mean volume fraction of TAGs in collagen fibrils is 9% of the dry volume. Using the density⁴⁰ and molecular mass¹⁴ of collagen (1.36 g cm⁻³ and 287 kDa) and that of triolein (0.908 g cm⁻³ and 885 Da), we found that f_TAG corresponded to 20 ± 11 (mean ± SD) TAG molecules per tropocollagen molecule. This is an unexpectedly large number, and it also serves as an estimate of the solubility limit of TAGs in collagen fibrils, since an excess of TAGs remained on the fibrils after washing with H₂O.

The volume fraction f_W of free water in the collagen fibril was obtained from the ratio of the cross-sectional area of each fibril measured in humid and dry air (Fig. 4). After washing with water, f_W decreases with increasing f_TAG (Fig. 4). After washing with DCM/MeOH, this correlation is lost, and f_W does not depend on f_TAG. In tendon 1, with a change in relative humidity (RH) from 55% to 88%, f_W = 0.23 ± 0.04 (mean ± SD, 16 fibrils, Fig. 4), and in tendon 2, with a change in RH from 5% to 95%, f_W = 0.61 ± 0.08 (18 fibrils, Fig. S6 and S7, SI). This increase in volume with increasing RH agrees very well with the estimate of the water content in collagen fibrils⁴⁰ and with drying/swelling measurements on macroscopic tendon specimens.^9,41

Fig. 4 TAGs affect the water fraction in collagen fibrils. (a) The volume fraction f_W of free water in a collagen fibril is given by f_W = (A_humid − A_dry)/A_dry. (b) f_W of collagen fibrils from tendon 1 plotted as a function of f_TAG after washing with water (black), and after washing with hexane and DCM/MeOH (D/M, red). Lines are guides for the eye.

To quantify fibril stiffness, which is a measure of the cohesive strength between tropocollagen molecules, we determined the effective indentation modulus E* from a fit of a modified Hertz contact model to FD data measured during tip approach (Fig. 5(a)) along the crest of each fibril. Examples of the regions of interest used to determine E* are shown in Fig. S5 (SI). The mean indentation modulus measured in humid air at 88% RH increases by a factor of 1.24 ± 0.019 (mean ± SD, n =16) after washing with hexane and by a factor of 1.79 ± 0.25 after washing with DCM/MeOH (Fig. 5(b)). An increase in E* is also observed at 55% RH. For fibrils from tendon 2, E* also increases after washing with DCM/MeOH (Fig. S7(b) and (c), SI). Reliable data on E* is unavailable for tendon 3 because the tip was very broad.

Fig. 5 Biomechanical function of TAGs in collagen fibrils. (a) FD data measured at the fibril crest yield the indentation modulus E*. (b) and (c) Box plots of indentation modulus E* at 55% RH (dry) and at 88% RH (humid) after sequential washing steps (tendon 1, n = 16).

In summary, our washing experiment shows that TAGs occupy 9% of the dry volume of collagen fibrils (Fig. 3(b)). After dissolving TAGs from collagen fibrils with DCM/MeOH, the collagen fibrils take up more water in humid air (Fig. 4(b)), and their indentation modulus E* increases (Fig. 5(b)). This seemingly contradictory effect reveals the biomechanical function of TAGs: they act as plasticizers and limit the water content of native collagen fibrils. The proposed atomistic mechanism is that TAGs reduce the total number of hydrogen bonds and water-mediated hydrogen bonds between adjacent tropocollagen molecules^42,43 and replace them with weaker van der Waals interactions between TAG alkyl chains and tropocollagen molecules. This reduces the cohesion between the tropocollagen molecules and the indentation modulus of the collagen fibrils.

Our finding prompts the question of how and where TAGs are incorporated in the collagen fibrils. The latter form a quasi-hexagonal lattice of tropocollagen triple helices surrounded by water molecules^14,40 (Fig. 6), whereby water molecules and collagen side chains undergo liquid-like reorientations.^44,45 TAGs from animal tissues are also liquid and disordered at room temperature. This may explain why the presence of TAGs in collagen fibrils has been overlooked, since the diffuse background present in the X-ray scattering data is subtracted prior to detailed analysis of the crystalline order.^14,46 Long-chain alkanols were found to intercalate between tropocollagens, causing the type I collagen crystal lattice to expand.⁴⁷ This intercalation was found to be reversible, suggesting physisorption of alkanols onto collagen molecules.⁴⁷ We propose that TAGs intercalate into collagen fibrils in a similar manner. Egli et al.^3,4 found that alkyl ligands of lipidized amino acids attach to the surface of collagen-like peptides in a hydrophobic groove. This increases the stability of the collagen-like peptide triple helix. Based on this, the non-polar alkyl chains of TAGs may align in the hydrophobic groove via hydrophobic interactions, and the carbonyl groups of TAGs may form hydrogen bonds with hydrogen bond donors (NH, NH₂, and COH groups) located at the surface of tropocollagen molecules. The hydrophilic glycerol group of TAGs may also form hydrogen bonds with water molecules.

Fig. 6 Proposed intercalation of TAGs (orange) into collagen fibrils (blue-gray). (a) Crystal structure of a 67 nm long segment of type I collagen fibril. (b) and (c) Cropped cross-sections in the gap and overlap regions with TAGs manually placed between tropocollagens. The rendering is based on X-ray diffraction data from Orgel et al.¹⁴ (PDB: 3HR2), and was created using ColBuilder⁴⁸ and ChimeraX.⁴⁹

The crystalline order of the collagen fibrils provides additional opportunities for TAG incorporation. The gap regions are less densely packed, with only four tropocollagens per unit cell, compared to five tropocollagens per unit cell in the overlap regions. The free lattice sites in the gap regions may provide space for TAGs. Indeed, transmission electron microscopy showed that liquid paraffin, refined wood oil, and other non-polar liquids can accumulate in the gap regions of collagen fibrils.⁵⁰ In addition, the crystalline order in the plane perpendicular to the fibril axis is polycrystalline,⁴⁶ and we suggest that TAGs may accumulate at grain boundaries.

Conclusions

The discovery of TAGs and their biomechanical function in native collagen fibrils challenges the current understanding of the composition of native collagen fibrils. More importantly, our results show that structurally embedded lipids, in the case of collagens TAGs, can modulate the cohesive forces between proteins at the molecular level. This also suggests that TAGs, and probably other lipids, may be involved in the mechanochemistry of collagens^51,52 during tissue remodeling, aging, and disease.

Methods

Sample preparation and washing procedure

Collagen fibrils were extracted from chicken calcaneal (Achilles) tendons and deposited on a Si substrate, as described in Magerle et al.⁵³ After drying, the specimen was washed three times with deionized water, air dried, and rehydrated with humid air in the AFM. Hexane (95%, CAS RN: 110-54-3, Grüssing GmbH, Filsum, Germany) and a 2 : 1 v/v mixture of dichloromethane (DCM, ≥99.9%, CAS RN: 75-09-2, Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany) and methanol (MeOH, ≥99.8%, CAS RN: 67-56-1, Thermo Fisher Scientific, Geel, Belgium) were used as solvents for lipids in native collagen fibrils. After washing the specimen with water and measuring the first set of FD data, the specimen was removed from the AFM and washed with hexane. This was done by placing a drop of hexane on the specimen and then removing the solvent by touching the edge of the drop with a paper tissue (Precision Wipes, Kimberly-Clark GmbH, Koblenz, Germany). This procedure was repeated 3–5 times until no lipid droplets were visible under an optical microscope with bright-field illumination. After measuring the second set of FD data, the same procedure was used to wash with DCM/MeOH (2 : 1 v/v) and a third set of FD data was measured. Triolein (98%, CAS RN: 122-32-7, Thermo Fisher Scientific, Geel, Belgium), 1,3-diolein (≥99%, CAS RN: 2465-32-9, Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany), and monocaprin (>98%, CAS RN: 26402-22-2, TCI Europe N.V., Zwijndrecht, The Netherlands) were used as reference materials for Raman and ¹H NMR spectroscopy.

Atomic force microscopy

Atomic force microscopy measurements were performed at room temperature in humid air using a NanoWizard II AFM (JPK Instruments AG, Berlin, Germany) as described by Magerle and coworkers.^23,53 Here, we used Si cantilevers (type PPP-NCSTR, NANOSENSORS, Neuchâtel, Switzerland) with a resonance frequency f_R ≈ 167.5 kHz and a half opening angle α = 15°. The spring constant of the cantilever, measured with Sader's method,⁵⁴ was typically 6.3 N m⁻¹.

Measurement procedure

The AFM measurements were performed as follows. After placing the sample in the swelling chamber, the sample was allowed to equilibrate for approximately 30 to 40 minutes. Then, a suitable area with individual fibrils was selected. We measured a 50 × 50 μm image in IC mode with 800 × 800 pixels. With Gwyddion, we manually determined the x and y positions and the directions of individual collagen fibrils, which were stored as a list. Using a self-written MATLAB script, we converted this list into a Python script for automated AFM measurements using the Experimental Planner module within the AFM control software. The script allowed automatic measurements of force–distance data at the previously defined positions. We used a force setpoint of 20 nN with a tip velocity of 3 μm s⁻¹ and a z-length of 200 nm to fully retract the tip from the surface. On each fibril, we measured an array of 100 × 25 FD data within an area of 1 × 0.25 μm². After measuring the FD data in the dry state, we turned on an airflow of about 0.5 L min⁻¹ through a washing bottle filled with deionized water. The specimen was allowed to equilibrate in humid air at 88% relative humidity (RH) for 30 to 40 minutes. We used the same AFM tip for all AFM measurements performed on all collagen fibrils extracted from one tendon. In a second and third set of measurements on collagen fibrils extracted from different tendons (tendons 2 and 3), a new AFM tip was used for each set of measurements.

All measurement steps were repeated with the specimen in humid air. The z-length was increased to 300 nm to account for a longer range of attractive capillary force at high relative humidity. After measuring the FD data in the humid state, a higher resolution IC mode image of 2048 × 2048 pixels was measured over an area of 50 × 50 μm. The specimen was then removed from the AFM and washed three times with hexane. After the solvent evaporated, the sample was placed back into the AFM and the exact same position was searched. We used an optical microscope with top-view optics to compare images before and after washing. Large bundles of collagen fibrils, located next to the individual fibrils, were used as characteristic landmarks to find the same spot and reposition the AFM cantilever on the region of interest. Then, we repeated the measuring procedure described above. The same force measurement scheme was used in dry and humid conditions. In a final step, the sample was washed with a mixture of DCM/MeOH, and all measurement steps were repeated.

Force–distance measurements

The FD measurements were performed with the same tip as for the IC mode AFM measurements, with a tip approach speed of 3 μm s⁻¹ until a force of 20 nN was reached, followed by tip retraction at the same speed. The tip–sample distance d was determined from the piezoelectric actuator position z, taking into account the cantilever bending.⁵⁵ The effective indentation modulus E* was determined from FD data measured using a modified Hertz model⁵⁶ that includes net adhesion, similar to the Derjaguin–Müller–Toporov (DMT) theory of contact mechanics,⁵⁷ yieldingfor d < d_S. Here, F_C is the adhesive force at d = d_S. We assumed a tip radius R = 8 nm as specified by the manufacturer and set the Poisson ratio v = 0.5 as in the work of Grant and coworkers.^7,8 In fitting eqn (1) to the FD data, we varied E* and d_S. The fitting range was limited to the indentation range from d–d_S = 0 to 15 nm.

We note that collagen fibrils display an elastoplastic deformation behavior.⁵³ Therefore, elastic contact models, like the Hertz model and related models, such as eqn (1), are not strictly applicable. Despite this limitation, the functional form of eqn (1) is commonly used for a phenomenological description of measured FD data.

AFM data analysis and visualization software

We used in-house developed MATLAB scripts (MathWorks, Natick, MA, USA) for analyzing FD data and fitting as described by Magerle and coworkers.^23,53 Gwyddion⁵⁸ was used for visualization of 2D data sets and for quantitative 2D image analysis. To measure the cross-sectional area of a fibril, we measured five adjacent d_S profiles on an overlap region of each fibril and determined the mean cross-sectional area of the fibril with respect to the Si substrate using Gwyddion.⁵⁸ To determine the mean and SD values for E*, we manually marked a 5 pixel wide line corresponding to approximately 100 individual points (see Fig. S5, green areas) on the crest of the collagen fibrils. FD depth profiles and 3D surface views were combined and rendered as 3D scenes using Blender 4.0.⁵⁹

Statistical analysis

For the statistical analysis, we used Origin Pro 2024 (OriginLab Corporation, Northampton, MA, USA) and the Python package SciPy.⁶⁰ Some fibrils showed a clearly visible residue next to the fibrils after washing with water and solvents. This residue prevented an accurate determination of the cross-sectional area of these fibrils. Hence, these data were excluded from the statistical analysis and marked with a dash in the data table. The residue does not affect the measurement of the indentation modulus at the crest of the fibril. The Shapiro-Wilks test was used to test the normal distribution of measured data. We performed a Tukey test with the following significance levels: p > 0.05 (N.S.), p < 0.05 (*), p < 0.01 (**), p < 0.001 (***). Numerical data is available in a separate Excel XLSX file in the SI.

Raman spectroscopy

We extracted the lipids from the tendon as follows. We cut a 2 cm long piece of chicken calcaneal tendon, removed it from the tendon sheet, and dabbed it dry with a paper tissue. After drying the tendon in air, we removed the adhering fat by dabbing with a fresh paper tissue. Then, the cleaned tendon was cut into small pieces of about 1 mm using a razor blade. In the next step, we dissolved the lipids from the tendon fragments with 5 mL of a dichloromethane-methanol mixture (DCM/MeOH, 2 : 1, v/v). To completely dissolve all lipids, we rinsed two to three times with 1 mL of the DCM/MeOH mixture. We concentrated the resulting solvent mixture and used it for further investigations with Raman and ¹H NMR spectroscopy. With a glass pipette, we deposited a droplet of the solution (approximately 0.5–1 cm in diameter) on an Au-coated, polished Si wafer and allowed the solvent to evaporate. Triolein, diolein, and monocaprin droplets were prepared in the same way. Raman measurements were performed in backscattering geometry using a Horiba Xplora Plus setup equipped with a spectrometer comprising a 1200 lines per mm grating and an electron multiplying charge-coupled device (EMCCD) A DPSS 532 nm continuous wave (CW) laser was focused on the samples by an 80×, 0.75 NA objective. The excitation laser power was approximately 10 mW, measured under the objective. Raman peaks were assigned according to Motoyama²⁵ and Czamara et al.²⁶

NMR spectroscopy

Extracted and air-dried lipids from chicken tendon and the reference materials were dissolved in CDCl₃ and ¹H NMR spectra were recorded with an Avance III 600 NMR spectrometer (Bruker Corporation, Billerica, MA, USA). Chemical shifts (δ) were reported in parts per million (ppm) and are referenced to the residual solvent resonance as the internal standard. ¹H NMR peaks were assigned according to the Spectral Database for Organic Compounds SDBS,²⁹ Mannina et al.,³⁰ and Babij et al.³¹

Author contributions

Martin Dehnert: conceptualization (equal); data curation (lead); formal analysis (lead); investigation (lead); methodology (lead); software (lead); visualization (lead); writing original draft (equal); writing – review & editing (equal). Tiberius Klose: conceptualization (equal); data curation (supporting); formal analysis (supporting); investigation (lead); methodology (lead); visualization (supporting); writing original draft (equal); writing – review & editing (equal). Yang Pan: data curation (supporting); formal analysis (supporting); investigation (supporting); methodology (supporting); writing – original draft (supporting); writing – review & editing (supporting). Dietrich R. T. Zahn: methodology (supporting); resources (supporting); supervision (supporting); writing – review & editing (supporting). Maximilian Voigtländer: data curation (supporting); formal analysis (supporting); investigation (supporting); methodology (supporting); writing – original draft (supporting); writing – review & editing (supporting). Johannes F. Teichert: methodology (supporting); resources (supporting); supervision (supporting); writing – review & editing (supporting). Robert Magerle: conceptualization (equal); formal analysis (supporting); methodology (lead); resources (lead); supervision (lead); visualization (supporting); writing original draft (equal); writing – review & editing (equal).

Conflicts of interest

The authors declare no competing interests.

Data availability

The data supporting this article have been included as part of the SI. Supplementary information: Additional figures, numerical data, and statistical analysis of cross sections and indentation modulus of individual collagen fibrils. See DOI: https://doi.org/10.1039/d5sm00696a.

Acknowledgements

We acknowledge funding from the Volkswagen Foundation (grant I/77476) and the Deutsche Forschungsgemeinschaft (INST 270/152-1 FUGG). We thank D. Stegerer and A. Süsselbeck for technical assistance and M. Besanko-Hoppen for proofreading.

References

1. Collagen: Structure and Mechanics, ed. P. Fratzl, Springer, US, 2008.
2. J. Bella, Biochem. J., 2016, 473, 1001–1025.
3. J. Egli, C. Siebler, M. Köhler, R. Zenobi and H. Wennemers, J. Am. Chem. Soc., 2019, 141, 5607–5611.
4. J. Egli, C. Esposito, M. Müri, S. Riniker and H. Wennemers, J. Am. Chem. Soc., 2021, 143, 5937–5942.
5. P. Walters, D. F. Cadogan and C. J. Howick, Ullmann's Encyclopedia of Industrial Chemistry, 2020, pp. 1–27.
6. M. H. Pineri, M. Escoubes and G. Roche, Biopolymers, 1978, 17, 2799–2815.
7. C. A. Grant, D. J. Brockwell, S. E. Radford and N. H. Thomson, Appl. Phys. Lett., 2008, 92, 233902.
8. C. A. Grant, D. J. Brockwell, S. E. Radford and N. H. Thomson, Biophys. J., 2009, 97, 2985–2992.
9. A. Masic, L. Bertinetti, R. Schuetz, S.-W. Chang, T. H. Metzger, M. J. Buehler and P. Fratzl, Nat. Commun., 2015, 6, 5942.
10. M. I. Gurr, J. L. Harwood, K. N. Frayn and R. H. Michell, Lipids: Biochemistry, Biotechnology and Health, Wiley, United Kingdom, 6th edn, 2016.
11. T. Nikkari, E. Heikkinen, P. Ekwall and O. Smidsrød, Acta Chem. Scand., 1968, 22, 3047–3049.
12. J. L. Rabinowitz and I. M. Shapiro, Arch. Oral Biol., 1973, 18, 295–296.
13. M. Le Lous, D. Boudin, S. Salmon and J. Polonovski, Biochim. Biophys. Acta BBA, 1982, 708, 26–32.
14. J. P. R. O. Orgel, T. C. Irving, A. Miller and T. J. Wess, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 9001–9005.
15. J. L. Rabinowitz, J. R. Gregg, J. E. Nixon and H. R. Schumacher, Clin. Orthop. Relat. Res., 1979, 143, 260–265.
16. P. Kannus, Scand. J. Med. Sci. Sports, 2000, 10, 312–320.
17. A. Scott, J. Zwerver, N. Grewal, A. de Sa, T. Alktebi, D. J. Granville and D. A. Hart, Br. J. Sports Med., 2015, 49, 984–988.
18. A. Steplewski, J. Fertala, R. Tomlinson, K. Hoxha, L. Han, O. Thakar, J. Klein, J. Abboud and A. Fertala, J. Orthop. Surg., 2019, 14, 172.
19. K. Squier, A. Scott, M. A. Hunt, L. R. Brunham, D. R. Wilson, H. Screen and C. M. Waugh, PLoS One, 2021, 16, e0257269.
20. A. R. Tall, D. M. Small and R. S. Lees, J. Clin. Invest., 1978, 62, 836–846.
21. Th Nemetschek, H. Nemetschek-Gansler, M. Ratzenhofer and R. Bowitz, Virchows Arch. A, 1976, 370, 251–254.
22. M. Dehnert and R. Magerle, Nanoscale, 2018, 10, 5695–5707.
23. R. Magerle, M. Dehnert, D. Voigt and A. Bernstein, Anal. Chem., 2020, 92, 8741–8749.
24. B. Cappella, Micron, 2017, 93, 20–28.
25. M. Motoyama, Bull. NARO Inst. Livest. Grassl. Sci., 2012, 12, 19–68.
26. K. Czamara, K. Majzner, M. Z. Pacia, K. Kochan, A. Kaczor and M. Baranska, J. Raman Spectrosc., 2015, 46, 4–20.
27. W. W. Christie, J. H. Moore, A. R. Lorimer and T. D. V. Lawrie, Lipids, 1971, 6, 854–856.
28. W. W. Christie and J. H. Moore, J. Sci. Food Agric., 1972, 23, 73–77.
29. National Institute of Advanced Industrial Science and Technology, Spectral Database for Organic Compounds, SDBS No.: 10525, https://sdbs.db.aist.go.jp/CompoundLanding.aspx?sdbsno=10525, (accessed September 11, 2025).
30. L. Mannina and A. Segre, Grasas Aceites, 2002, 53, 22–33.
31. N. R. Babij, E. O. McCusker, G. T. Whiteker, B. Canturk, N. Choy, L. C. Creemer, C. V. D. Amicis, N. M. Hewlett, P. L. Johnson, J. A. Knobelsdorf, F. Li, B. A. Lorsbach, B. M. Nugent, S. J. Ryan, M. R. Smith and Q. Yang, Org. Process Res. Dev., 2016, 20, 661–667.
32. J. Folch, I. Ascoli, M. Lees, J. A. Meath and F. N. LeBaron, J. Biol. Chem., 1951, 191, 833–841.
33. J. Folch, M. Lees and G. H. S. Stanley, J. Biol. Chem., 1957, 226, 497–509.
34. E. G. Bligh and W. J. Dyer, Can. J. Biochem. Physiol., 1959, 37, 911–917.
35. W.-T. Tsai, Toxics, 2017, 5, 23.
36. C. Cravotto, A.-S. Fabiano-Tixier, O. Claux, M. Abert-Vian, S. Tabasso, G. Cravotto and F. Chemat, Foods, 2022, 11, 3412.
37. C. McAuliffe, J. Phys. Chem., 1966, 70, 1267–1275.
38. J. M. Wallace, Q. Chen, M. Fang, B. Erickson, B. G. Orr and M. M. Banaszak Holl, Langmuir, 2010, 26, 7349–7354.
39. S. J. Baldwin, A. S. Quigley, C. Clegg and L. Kreplak, Biophys. J., 2014, 107, 1794–1801.
40. D. J. S. Hulmes and A. Miller, Nature, 1979, 282, 878–880.
41. C. Morin, C. Hellmich and P. Henits, J. Theor. Biol., 2013, 317, 384–393.
42. I. Streeter and N. H. de Leeuw, J. Phys. Chem. B, 2010, 114, 13263–13270.
43. I. Streeter and N. H. de Leeuw, Soft Matter, 2011, 7, 3373–3382.
44. L. W. Jelinski and D. A. Torchia, J. Mol. Biol., 1980, 138, 255–272.
45. S. K. Sarkar, Y. Hiyama, C. H. Niu, P. E. Young, J. T. Gerig and D. A. Torchia, Biochemistry, 1987, 26, 6793–6800.
46. D. J. Hulmes, T. J. Wess, D. J. Prockop and P. Fratzl, Biophys. J., 1995, 68, 1661–1670.
47. Th Nemetschek, Z. Naturforsch., 1968, 23b, 507–511.
48. A. Obarska-Kosinska, B. Rennekamp, A. Ünal and F. Gräter, Biophys. J., 2021, 120, 3544–3549.
49. E. F. Pettersen, T. D. Goddard, C. C. Huang, E. C. Meng, G. S. Couch, T. I. Croll, J. H. Morris and T. E. Ferrin, Protein Sci., 2021, 30, 70–82.
50. X. W. Hu, D. P. Knight and J. A. Chapman, Biochim. Biophys. Acta, BBA, 1997, 1334, 327–337.
51. S. M. Siadat and J. W. Ruberti, Mech. Cells Fibers, 2023, 163, 50–62.
52. C. Zapp, A. Obarska-Kosinska, B. Rennekamp, M. Kurth, D. M. Hudson, D. Mercadante, U. Barayeu, T. P. Dick, V. Denysenkov, T. Prisner, M. Bennati, C. Daday, R. Kappl and F. Gräter, Nat. Commun., 2020, 11, 2315.
53. R. Magerle, P. Zech, M. Dehnert, A. Bendixen and A. Otto, Soft Matter, 2024, 20, 2831–2839.
54. J. E. Sader, J. W. M. Chon and P. Mulvaney, Rev. Sci. Instrum., 1999, 70, 3967–3969.
55. H. J. Butt, B. Cappella and M. Kappl, Surf. Sci. Rep., 2005, 59, 1–152.
56. H. Hertz, J. Reine Angew. Math., 1882, 92, 156–171.
57. B. V. Derjaguin, V. M. Muller and Yu. P. Toporov, J. Colloid Interface Sci., 1975, 53, 314–326.
58. D. Nečas and P. Klapetek, Cent. Eur. J. Phys., 2012, 10, 181–188.
59. Blender Online Community, Blender – a 3D modelling and rendering package. 2024; https://www.blender.org.
60. P. Virtanen, R. Gommers, T. E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau, E. Burovski, P. Peterson, W. Weckesser, J. Bright, S. J. van der Walt, M. Brett, J. Wilson, K. J. Millman, N. Mayorov, A. R. J. Nelson, E. Jones, R. Kern, E. Larson, C. J. Carey, İ. Polat, Y. Feng, E. W. Moore, J. VanderPlas, D. Laxalde, J. Perktold, R. Cimrman, I. Henriksen, E. A. Quintero, C. R. Harris, A. M. Archibald, A. H. Ribeiro, F. Pedregosa, P. van Mulbregt and SciPy 1.0 Contributors, Nat. Methods, 2020, 17, 261–272.

---