*2.4. Pore size and Fiber Diameter*

The collagen and collagen–elastin samples were prepared as described above. A 200 μL amount of the ice-cooled solution was placed in each well of a cooled 24-well μ-plate (μ-Plate 24 Well ibiTreat; Cat. No. 82406; ibidi, Gräfelfing, Germany). Subsequently, the 24-well plate was placed in an incubator at 37 ◦C to start the polymerization for 2 h. The polymerized hydrogels were washed three times using PBS and fluorescently stained by applying 20μg/mL of 5(6)-Carboxytetramethylrhodamine N-succinimidylester (TAMRA-SE; Cat. No. 21955; Merck KGaA, Darmstadt, Germany) overnight and subsequently washed three times using the PBS. Using an LSM microscope (TCS SP8; Leica, Wetzlar, Germany), three-dimensional image cubes of the fluorescence signal of the TAMRA-SE using a 561 nm excitation laser and HC PL APO CS2 40x/1.10 water immersion objective were recorded. The pore size and fiber diameter were determined as published previously [17,18]. To compensate for the apparent collagen–elastin clusters that would disturb a fiber diameter determination, a custom-built cluster deletion algorithm was used to solely measure the actual fiber diameter sizes.

#### *2.5. Directionality*

The above obtained images were also used to quantify the directionality and its standard deviation of the network. The imageJ plugin, "Directionality", with the "Fourier components" method was used. A total of 70 planes of each of 10 different random positions for both gels (collagen and collagen–elastin) were summarized into one image which was then analyzed. The clusters in the collagen–elastin samples were removed prior to analysis using the same, custom-built cluster deletion algorithm as described above.

#### *2.6. Elastin Influence on the Network Structure*

To investigate the influence of elastin polymerization on the final hydrogel structure, we used a Col-F collagen binding reagent (Col-F; Cat. No. 6346; ImmunoChemistry Technologies, Bloomington, MN, USA) and a collagen I antibody (Immunotag™ Collagen I Polyclonal Antibody; Cat. No. #ITT5769; G-Biosciences, St. Louis, MO, USA). The collagen–elastin hydrogels were prepared in 24-well μ-plates as described above.

For the collagen I antibody staining, the samples were incubated with 5% BSA solved in PBS for 30 min at room temperature—with aspirate goat serum, and incubate sections with primary antibody (ITT5769) in PBS overnight at 4 ◦C or 1 h at 37 ◦C; 3 × 1:1000 (600 μL/well). The samples were washed three times with PBS for 5 min each.

For the Col-F staining, the samples were incubated with 3% BSA solved in PBS for 30 min at room temperature—with aspirate goat serum, and incubate sections with primary antibody (ITT5769) in PBS overnight at 4 ◦C or 1 h at 37 ◦C; 3 × 1:200 (300 μL/well). The samples were washed three times with PBS for 5 min each.

Three-dimensional images were recorded using an LSM microscope (TCS SP8; Leica, Wetzlar, Germany) with a 63×/1.20 HC PL APO CS2 water immersion objective and a 488 nm (Col-F) and 561 nm (collagen antibody) excitation laser, respectively. The final image dimensions were 100 μm by 100 μm in x-y and a roughly 30 μm to 50μm z dimension.

#### *2.7. Live Polymerization*

The samples were prepared as described above. A 1 mL amount of the cooled solution was placed in a well of a pre-cooled 24-well μ-plate and then placed in a LSM microscope with an incubation chamber (TCS SP8; Leica, Wetzlar, Germany) at 37 ◦C and 100% relative humidity. Using a HC PL APO CS2 40×/1.10 water immersion objective and a 561 nm laser in the reflection mode, a 1 h recording of the polymerization process and hydrogel network formation was observed and recorded as live imaging videos. The videos had an image size of 1024 × 1024 px with a frame-rate of 1 fps.

## *2.8. Statistical Methods*

The employed statistical methods included the mean, median, standard deviation and a box plot as well as a Mann–Whitney-U test. The methods are named at the relevant position.

#### **3. Results**

#### *3.1. Plate Reader*

Figure 1 displays the polymerization curves of a collagen solution and two collagen elastin solutions at 37 ◦C. The heating curve of the collagen is in strong agreement with the literature data, as it highlights the onset of clouding after 30 min as well as no significant changes in the turbidity after 2 h [19]. The clouding curve is generally associated with fiber formation which increasingly contributes to light scattering. The addition of elastin then introduces several features into the polymerization process. The most striking feature is that the final absorption (>2 h) was only slightly increased, although 25% or 30%, respectively, should be expected as this is the net increase in the biomass for each sample. This is a strong indication that the alignment of elastin and collagen monomers must occur in a lateral manner, as the opposite case of a perpendicular alignment would contribute to light absorption and scattering. The small increase in absorption, at times >2 h, might be a consequence of an elevated fiber thickness as elastin monomers attach to the collagen triple helix. The lateral addition of elastin is also likely as circular dichroism experiments on elastin–collagen gels have shown that the addition of elastin leads to a reduced PPII (polyproline II) content, probably due to PPII helix distortion [13]. The addition of low amounts of elastin increases the polymerization rate by a factor of two while the polymerization rate maximum is shifted to an earlier time (37 min instead of 49 min, see Table 1). At these concentrations, the elastin may act as a nucleation center for polymerization. Additional effects which could fasten the assembly might include the burying of hydrophobic domains in collagen but especially in elastin, which has alternating hydrophilic and hydrophobic

segments [20,21]. Additionally, a potential entropy gain by a helix distortion, as mirrored in the reduced PPII helix content in collagen after an elastin addition, supports the thesis of a conformation-dependent collagen–elastin interaction [13]. Such an entropy gain by a helix distortion was described for an alpha helix [22]. Taken together, collagen's and especially elastin's propensity to bury their hydrophobic domains, as well as a general increase in the monomer concentration, might contribute to an increase in the polymerization rate. In terms of the type of fiber alignment, we argue that hydrophobic burying implies a parallel alignment, as in the otherwise perpendicular type no significant burying can take place.

**Figure 1.** (**a**) Mean polymerization curves at 37 ◦C for a 2 mg/mL collagen solution as well as two collagen–elastin solutions containing 20 w% (0.6 mg/mL elastin) and 33.3 w% (1.2 mg/mL elastin), respectively. (**b**) Derivative of the mean of the curves in (**a**). Six wells were recorded per sample and the color-coded curves in figure (**a**) denote one standard deviation. Supplementary Figure S3 shows the extended curves. The random spikes in the beginning of the collagen curve result from water condensation and evaporation under the well plate cover.


**Table 1.** Characteristic values for polymerization. *t* <sup>1</sup> 2 stands for the time where the derivative of the turbidity curves has the greatest value, i.e., the increase in turbidity is the greatest, while Abs. at t(1/2) (a.u.) stands for the absorption value (turbidity) at the time of *t* <sup>1</sup> 2 .

A further addition of elastin, however, decreased the polymerization rate. This was unexpected as the polymerization rate is always proportional to the monomer concentration. Thereby, at relevant concentrations, elastin can be viewed as a perturbation towards polymerization as it may interfere with the proper alignment of collagen triple helical monomers. A fingerprint of this feature was the additional absorption shoulder at 20 min which probably signified a second polymerization process introduced by the elastin. This shoulder is believed to originate from the formation of elastin–collagen clusters which form at elevated elastin concentrations. This is a likely process, as elastin to collagen ratios of more than 0.22 will exceed elastin–collagen equimolarity. In fact, based on the molar masses of collagen (300,000 Da) and elastin monomers (ca. 67,000 Da), 0.5 mg/mL of elastin is sufficient to accommodate 2 mg/mL of collagen in an equimolar manner [23,24]. The plate reader experiments fell well within this consideration, as they showed that 0.6 mg/mL of elastin did not lead to an additional clustering peak at 20 min, while the 1.2 mg/mL sample did so; therefore, the upper limit for an elastin addition seems to lie between these values. As the absorption value in the 33.3 w% curve of the 20 min peak was much smaller than the maximum absorption, it can be argued that most of the biomass was polymerized into the gel. Another interpretation may be that the shoulder at 20 min signified clouding by elastin coacervation, a well-known effect which describes the heat-induced elastin aggregation by an association of hydrophobic elastin segments. However, elastin coacervation is quite fast and usually complete after several minutes; therefore, we can exclude this effect here [25]. It is important to note that a similar experiment was performed by Vazquez-Portalatin et al. by also using collagen type I and bovine neck ligament elastin. They similarly recorded the clouding of elastin–collagen solutions for several elastin–collagen ratios [26]. Opposed to our experiments, they observed an overall increase in the polymerization rate and a shift in the polymerization start to earlier times with an increasing elastin proportion; however, the maximum polymerization rate was around 21 min, which was twice as fast as our observation of around 40 min. Additionally, the turbidity-dependence on the elastin concentration was much lower than in our experiments. This might not only have to do with the fact that they used only rat tail collagen (R collagen), another wavelength (313 nm) and PBS (phosphate-buffered saline) instead of a phosphate buffer. In fact, they used comparable elastin–collagen ratios but with a 1:10 dilution. This gives credit to our above claim of a saturative process during polymerization. Obviously, in our experiments, the addition of elastin at elevated concentrations seemed to induce a second polymerization process apart from the "classical" polymerization which we introduce as a cluster formation, probably because the monomers met more often which also increased the chance that the monomers met without being optimally aligned in the gel. These clusters grew on their own without participating in the classical polymerization. In fact, Paderi et. al. discuss how a perpendicular chain alignment can inhibit collagen polymerization which may, in our case, have been the nucleation center for the cluster formation [27].

#### *3.2. Videos of Polymerization*

Videos S1–S3 (supplement) show the fiber formation of a collagen solution and two elastin–collagen solutions, while Video S4 (supplement) shows a comparison video. Video S1 is characterized by early and quick flashes of fibers and nodes which resulted from their diffusion through the focal plane. Small fibers and nodes could be observed as early as five minutes after combining both solutions (the collagen stock and buffer). This was contrasted to the plate reader experiments where no significant changes in the absorbance were observed before 25 min. This was because the plate reader measures absorbance which is quite small for small particles, so that only sufficiently large particles or fibers can contribute to the absorbance. The videos emphasize that the fibers assembled rather quickly while they were still subjected to convection, i.e., liquid flow. The onset of polymerization was characterized by a fiber flow velocity reduction which came to a complete stop as soon as sufficiently large enough fibers had come into contact. The fiber growth occurred in the early stages end to end and was then followed by a fiber thickening, which is in line with the literature claims that axial growth is much faster than lateral growth [28]. The sequence of the axial followed by the lateral fiber growth was retained when the elastin was added, implying that the elastin did not significantly interfere with the fiber assembly process in terms of the network structure. Moreover, when the polymerization sequence of the collagen–elastin solution was identical to the one of the pure collagen polymerization sequence, then the elastin must have been homogeneously incorporated into the collagen system, i.e., laterally. It was further obvious, that the elastincontaining networks polymerized earlier, which was in good agreement with the plate reader experiments. Consequently, elastin seemed to facilitate polymerization as described above, although this effect was concentration-dependent. The maturing collagen network was still drifting through the focal plane as seen in the appearance and disappearance of fibers and nodes. This implies that the network was subjected to density fluctuations during the polymerization. This was contrasted to the elastin containing networks, which did not drift through the focal plane. This might relate to our observations, namely, that the elastin-containing gels appeared to stick to the walls of the petri dish. This effect might limit the z-drift. A final observation was that the elastin-containing networks contained some clusters which were more prominent in the high-elastin concentration sample. Video S4 shows quite nicely how these clusters disappeared after around 30 min. We believe that the clusters sunk either to the bottom of the gel or were bound randomly to the existing fibers, although we could not observe such diffusion to the fibers. We further argue that the presence of these clusters coincided with the presence of the elevated absorbance around 20 min in the elastin-containing turbidity curves (Figure 1); however, further experiments are required to understand the interaction between the elastin and collagen R and G. This question bears some importance, as G collagen is more closely related to the formation of nodes than R collagen [19]. In fact, further applications might demand answering the question of whether elastin is also present in the nodes as a local matrix stiffness can guide the cell migration [19].

#### *3.3. Zeta Potential*

The zeta potential measurements of the individual collagen and elastin solutions in the phosphate buffer at pH = 7.5 and 4 ◦C revealed that all solutions exhibited a zeta potential around −4 mV (Figure 2). Values in this range are optimal for aggregation as values smaller than ± 30 mV are considered to induce aggregation [29]. Although biopolymers such as collagen and elastin have plenty of ionizable groups, zeta potential values around zero indicate a low degree of ionization. This low potential, as described above, favors monomer aggregation in any way, including laterally, as the resulting hydration shell will be small at these values so that the repulsion will effectively play no role. In fact, both collagen assembly and elastin assembly (coacervation) are endothermic and entropy driven at 37 ◦C, while the loss of an ordered hydration shell is the largest contribution to entropy gain [25,30]. However, a decrease in Gibbs energy is roughly twice as much in

collagen than it is in elastin, implying that collagen can more easily lose its hydration shell. Moreover, although the thermodynamics for elastin relate to the effect of coacervation, we did not see this effect in the plate reader experiments, where no early clouding could be detected. Collagen and elastin monomers could, therefore, align in a parallel manner before a temperature increase shifts the Gibbs energy change from positive to negative such that, after a loss of the respective hydration shell, the collagen–elastin association is more favorable than an elastin–elastin association (coacervation). Elsewhere, the lateral merging of a hydration shell of peptides has been described which opens up the possibility of a multistep mechanism of an early elastin–collagen interaction [31]. It is also noted that the employed collagen was already in its triple helical state so that the elastin should not have interfered with the triple helix formation; however, the data of Wilharm et. al. show that the circular dichroism of collagen–elastin is not an ideal superposition for each component and that it lacks some PPII content [13]. This shows how the presence of elastin might impact the collagen helix anyway, probably due to the destabilization of the intricate H-bond equilibrium in collagen, probably in a lateral manner.

**Figure 2.** Zeta size measurements of a collagen (2 mg/mL) and an elastin solution (2 mg/mL) in a phosphate buffer at pH 7.5 and 4 ◦C. Incremented mean (thick lines) and one standard deviation (thin lines) are shown. As data points along the curves have slightly varying total run times from 7 to 10 min (due to variations in temperature regulation by the device), the data points were averaged accordingly and plotted over the increment. Original data can be found in supporting Figure S4.
