A synthetic cell density signal can drive proliferation in chick embryonic tendon cells and tendon cells from a full size rooster can produce high levels of procollagen in cell culture

Cells and cell culture

Embryonic tendon cells were isolated from 16 day chick embryos by a modification Petzold & Schwarz (2013) of the procedures described by Dehm & Prockop (1971). The basic medium used was a 50/50 mix of F12/DMEM and if serum was used it was fetal bovine serum deactivated at 56 °C for 30 m.

Adult tendon cells were from full size roosters that had been freshly killed at the local Muslim butcher. The rooster was washed in a disinfecting solution, ~1% solution of BDD (Fisher Scientific, Waltham, MA, USA) for ~5 min and then ~3 min wash in 70% ethanol. The legs were removed. Then the shank dissected from the foot to the hock. The thick, scaly, yellow skin covering the shank was slit open revealing a sheath with several tendons running through it. The tendons were dissected and removed from the sheath, cut into small pieces, and then put in dissociation medium (20 ml) containing 120 mg collagenase (type I; Sigma or Worthington, Columbus, OH, USA); 80 μl bovine serum albumin (at 1 mg/ml; Sigma, Columbus, OH, USA); 200 μl, 100x penicillin/streptomycin. The 50 ml centrifuge tube was incubated horizontally for 3 h at 37 °C on a rotator at 100 rpm and for the last 20 m, 0.5 ml of 1 mg/ml of DNAse (Cooper Biomedical, San Ramon, CA, USA) was added to the dissociation solution. Then, the tube was held upright and the large undigested tendon tissue was allowed to settle to the bottom. The supernatant was collected and 10ml of medium was added and tube inverted several times. Again the second supernatant was collect and added to the first. This was put through a nylon mesh (192 μ) to further remove large fragments. At this point the cells were spun down in a small clinical centrifuge (speed 6 out of 7), resuspended in 20 ml of fresh medium and counted on a hemocytometer slide.

The 16 day chick embryos were exempt from requiring an animal use protocol as determined by the LBNL Animal Welfare and Research Committee. Similarly, the purchase of a dead rooster from the butcher shop was also an exempt use.

Freezing embryonic tendon cells freshly isolated from the tendon

Tendon cells were spun down and resuspended in freezing media, 90% FBS and 10% DMSO, at 4 × 10⁶ cells/ml. A total of 1 ml aliqouts were put into a CoolCell (Corning, Corning, NY, USA) and this was put into a −80 °C freezer. Within a couple days the tubes were transferred to a liquid nitrogen freezer. When needed, cells were quickly thawed and 250 μl of cells were added to a flask containing 10 ml of F12/DMEM medium. The cells were placed in a 5% CO₂ incubator for 3 h and then the medium was changed to the appropriate growth medium.

Production and purification of SNZR P

To make an exact copy of the 94 AA protein (SNZR P), an Impact kit (New England BioLabs, Ipswich, MA, USA) was used with the following modifications. The vector pTYB21 was used so that an N-terminus methionine would not be added to SNZR P. Because the cDNA encoding the 94 AA protein contains Sapl restriction enzyme site, Sapl could not be used in the cloning. Instead, another kit was used, NEBuilder HiFi DNA Assembly Master Mix/NEBuilder HiFi DNA Assembly Cloning (New England BioLabs, Ipswich, MA, USA) that allows assembly of the vector to SNZR P without the need for restriction enzymes. Another change was to transform the E. coli strain DH5a first and then subcloned into T7 express (New England BioLabs, Ipswich, MA, USA). The induction of these cells used 0.4 mM IPTG at 30 °C for 5 h. The sonicator (diagenode, Denville, NJ, USA) used to break open the E. coli was set at 4 °C at high power with 20 cycles of 30 s on/30 s off. The other directions in the kits were closely followed.

SDS gel electrophoresis

The standard gel format was Novex™ 4–20% Tris-Glycine Mini Gels, WedgeWell™ format, 10-well, 1 mm thick (Invitrogen, Waltham, MA, USA) run for 90 m at 120 v. Samples were diluted with equal amounts of 2x sample buffer (Invitrogen, Waltham, MA, USA), β-mercaptoethanol was added (5%),and then heated at 80 °C for 10 m. In the case where we were tracking the production of SNZR P there was sufficient amounts of protein that the gels were stained with Simply Blue (Invitrogen, Waltham, MA, USA) and photographed with a standard camera and light box. When we were tracking the loss of the SNZR P band when it would bind to SNZR L, the levels were in the ng range so the staining shifted to Sypro ruby (Invitrogen, Waltham, MA, USA) and bromphenol blue was removed from the sample buffer in order to reduce background fluorescence at the front of the gel. Because of limited access to the lab during COVID-19, the Sigma protocol was used where the gel was run and fixed on 1 day and stained on another day. The gels were photographed using excitation at 472 nm and an emission at 684 nm (Azure Biosystems 600; Azure Biosystems, Dublin, CA, USA). The Azure Biosystems software was used to quantitate the bands. The recommended rolling ball method was used to define background. A small value for the ball (#4) allowed the software to draw an accurate boundary between the fluorescent signal and the background. When using RNase A (Worthington, Columbus, OH, USA) to make a standard curve between 3 and 21 ng (Fig. S2), the software generated a linear fit going through the origin. Similarly, the software quantitatively calculated the ratio of the procollagen α1 band to the procollagen α2 band as 1.8 ± 0.3, n = 14 (Table 1).

Chemical synthesis of SNZR L, the unique lipid of the chicken tendon cell density signal

In a previous article (Petzold & Schwarz, 2013) a working model of SNZR L was developed and is now shown as the “original” lipid structure (Fig. 1). This model was consistent with mass spectrometry (ms) data and enzymatic data on specific lipid cleavage sites. Other analytical methods were not available because they required a higher concentration of the lipid than we could produce. Our goal was to synthesize a biological active lipid knowing that a small error in the structure usually results in little or no activity. Our only solution for confirming the lipid structure was to synthesize the structure that best fit all the available data and then test its biological activity on tendon cells.

This came into focus when Avanti Polar Lipids was contracted to make the original lipid. They realized that joining a fatty acid to a phosphate, forms a high energy, acid anhydride bond. This could be problematic because the intermediate form might not be stable enough to allow a second fatty to be added. The synthesis might require that both fatty acids be added at the same time resulting in a mixture of three products. And even then, these molecules might not be stable. A group of lipid chemists and ms specialists at Avanti Polar Lipids met and discussed this problem. They came up with an “alternative” structure (Fig. 1) where the fatty acids are reduced yielding an alcohol with a ketone, a common metabolite. This would change the bond to a stable ester while not altering the MW of the molecule or the ms/ms fragmentation pattern.

The better candidate between the “original” form and the “alternative” forms relied on our characterization of the active molecule, SNZR PL. Previous experiments with SNZR PL (in conditioned media) showed that it was stable to heat, 90 °C for 10 min (Zayas & Schwarz, 1992), to storage at 4 °C in conditioned medium for 1 year, and to purification through four columns with changes in salt concentration, pH, and solvents (Petzold & Schwarz, 2013). Stability strongly favored the “alternative” structure. Avanti Polar Lipids synthesized 0.4 g from which a concentrated solution was made by adding 0.1 mg to 50 ml of ethanol (95%) and this solution has been stable for over a year at 4 °C.

Molecular synthesis of the 94 amino acid (AA) chicken protein, SNZR P

SNZR P is the opposite of SNZR L being highly conserved between species and tissues. Even the larger gene, 424 AA from which SNZR P is cleaved at the carboxy end, is highly conserved between chicken and human (Petzold & Schwarz, 2013). The chicken form can be an active participant in human cells even with the addition of a methionine at the amino terminus and tags at the carboxy terminus. Nevertheless, in making a synthetic version for the first time it was important to make the protein identical to the native form to reduce the chance for potential artifacts. Because SNZR P was shown to be unmodified when expressed in human and chicken cells (Petzold & Schwarz, 2013), this protein can be made in E. coli.

The IMPACT kit from New England BioLabs was developed in part for this purpose. The E. coli expression system makes a 56 kD protein that can be extended with our 94 AA protein (10.8 kD) yielding a composite protein of 66.8 kD. The 56 kD portion has two important functions. It has a chitin binding site that allows it to bind tightly to a chitin column. This allows the separation the 66.8 kD protein from the vast majority of E. coli proteins. The second property is inclusion of an intein, an enzyme that cleaves at its carboxy terminus when induced with a reducing agent. This releases the 10.8 kD protein in almost pure form (Fig. 2). A liter culture of E. coli yields approximately 1 mg of SNZR P.

Reconstituting SNZR PL from its component parts

SNZR P and SNZR L are bound tightly together in the active signaling molecule. Since the protein can have regions where it is hydrophobic and other regions where it is hydrophilic, these could interact independently to bind to similar regions in the phospholipid. There are also divalent ion interactions because chelators are known to inhibit activity (Zayas & Schwarz, 1992).

The binding will change its apparent molecular weight on SDS gels from ~11 to ~16 kD and activate its biological activity. The protein by itself has no biological activity (Petzold & Schwarz, 2013) and the lipid biological activity by itself is not known. The first two attempts to recombine SNZR PL from its components focused on biological activity. The first was to mix equal molar amounts SNZR P with SNZR L in PBS at room temperature at 1,000x, where x = 100 pg/ml of SNZR P. The second attempt was mixing equal molar amounts with 0.5 mM Ca⁺⁺ and 0.5 mM Mg⁺⁺ at 37 °C for 17 h at 10,000x. In these two cases, testing heat, concentration, and divalent ions for their ability to drive the two components together and form an active molecule. Initially, when added to primary cultures of embryonic chick tendon cells, all the concentration allowed the cells to attach to the flask and seemed to be growing (1x, 5x, 10x, 25x, and a control with 0.2% FBS (fetal bovine serum)) but by the second medium change on the 4th day, cells given high levels stopped growing and all of them were not confluent by the 7th day in sharp contrast to the control cells growing in low levels of FBS. Simply stated, the data from this experiment was unexpected. Furthermore, using a biological assay when the role of SNZR L was not better understood only added to the complexity of the problem.

To gain clarity, our focus shifted to gel analysis of the assembly of SNZR PL from its two components. Preliminary experiments showed that Sypro ruby, a red fluorescent protein dye, had sufficient sensitivity to detect ng levels of SNZR P. These experiments also showed that heating for 10 min at 50 °C or 60 °C in the presence of divalent ions would reduce the SNZR P signal while 37 °C was not sufficient. These experiments also showed that we could not detect the SNZR PL band. This is expected because adding a lipid component to the protein could alter the binding of the protein dye. A similar problem had been seen before with Western blotting when different monoclonal antibodies to SNZR P, or to tags added to SNZR P, would not bind well to SNZR PL (Petzold & Schwarz, 2013). During a one hour time course experiment at both 50 °C and 60 °C, the SNZR P band is reduced by about two third by 20 min and ~90% by 60 min using either temperature (Fig. 3). The ~10% of SNZR P that does not find its SNZR L partner, implying there is a small excess, is the preferred result because SNZR P alone does not have biological activity. Because the transition between working at 50 °C and not working at 37 °C has not been defined, 60 °C would offer a safer margin for our standard conditions for recombining SNZR PL.

Biological activity of reconstituted SNZR PL

To confirm that the reconstituted SNZR PL were correctly aligned and biologically active, it was tested on embryonic tendon cells at 1x, 5x, and 10x and a control with 0.2% FBS. In contrast to the earlier experiment, cells continued to grow over the week but did not become confluent except for the control cells. This data triggered a standard response when cells do not grow fast enough—add more factor. So we tested 0x, 5x,10x, 20x, 40x, 80x, 0.2% FBS on primary avian tendon cells. In basal medium, cells seeded poorly and grew slowly. Initially, the higher levels (≥10x) would grow better than the 5x cultures but growth would slow by mid week. This is the expected behavior in our two factor model where feedback control is important part of the model. In the next experiment, the focus was on 5x cultures and growing them for longer times. Nevertheless, after 9 days they had grown well but were not confluent. At that point, we drove them to confluency by raising the level of SNZR PL to 10x for 2 days (Fig. 4, day 11).

The simple conclusion is that one can reduce the protein level in 0.2% serum a thousand fold by using SNZR PL and stimulate primary avian tendon cells to grow from low cell density to a confluent culture. This is a good confirmation that the reconstituted SNZR PL are biologically active. But this result helps shed light on several other issues. For instance, why does growing embryonic tendon cells with 0.2% FBS allow the cells to grow faster and reach confluent densities in 6 days instead of the 11 days using SNZR PL? This very low level of serum probably does not influence the cell as a growth factor when serum usually is used at 25 to 50 fold higher levels. Low levels of serum are more likely to stimulate the metabolism of the cell (Valmassoi & Schwarz, 1988), and as a consequence, the cell produces more SNZR PL. This would raise the basal level but at a local level and at lower amounts. In comparison to standard cell culture approach where we add ng/ml of SNZR PL to the medium, flooding the environment. This can quickly overload the response triggering feedback control. So having a good delivery system for SNZR PL is important in producing a desired response from the cell.

These experiments also give us insight into SNZR L when not bound to its protein partner. In the early experiments to combine SNZR L with SNZR P where the temperature was 37 °C, no binding occurred and the biological activity was due to SNZR L. Free SNZR L outside the cell, even at low levels, acts like high levels of SNZR PL. Free SNZR L would still interact with the membrane of the cell but would no longer be diffusible, effectively increasing its biological activity. Increasing its levels with each media change, it would quickly transition from a stimulator of growth to an inhibitor of growth. Nevertheless, free SNZR L is not normally found outside the cell and its role inside the cell should remain as predicted in our model. Only now SNZR L binds to the inside of the plasma membrane and this can influence SNZR PL that binds to the outside of the membrane. When SNZR L on the inside membrane is close enough to interact with SNZR PL, this “complex” changes the biological signaling to an inhibitor of proliferation and a stimulator of procollagen production. While SNZR L continues to be a good candidate for the second factor, this remains to be determined.

Proliferation and differentiation in adult chick tendon cells

Embryonic chick tendons are formed rapidly so that the hatched chick can walk and find food and water. The length is controlled by a growth plate that is postulated to interact with muscle connective tissue to insure that the tendon is always the correct length as the organism grows (Petzold & Schwarz, 2013). The growth plate “moves” by having cells at the front proliferate, cells in the middle differentiate, and cells at the back apoptose. A small subset of cells at the periphery of the growth plate never reaches high cell density and do not apoptose. These cells end up between collagen fibrils where they make the collagen that expands the diameter of the fibril so that the tendon is strong enough to support the weight of the growing chicken (Fig. S1). When full size is reached, tendon cells need to reduce proliferation and collagen production to maintenance levels. How is this transition achieved and is it reversible, at least to some extent, by changes in the levels of SNZR PL?

Releasing cells from the tendons of full size roosters is a challenge. The tissue is made up of highly dense, cross-linked collagen with relatively few cells residing between the fibrils (Fig. S1). To free the cells one has to incubate the tissue for several hours in collagenase. This lowers the viability of the released cells. In this experiment, the cells were seeded at 36 × 10³ cells/cm² but the number of viable cells was estimated at only 25%. This lower cell density slows initial growth and pushes the cells to grow in patches because cell distribution is not uniform, cells in moderate density areas grow better than cells in low density areas. This low viability also restricts our ability to analyze growth curves. So our focus was on how adult tendon cells respond to cell density and ascorbate in their regulation of procollagen production.

Cells were grown in medium with 0x, 3x, 5x, or 10x SNZR PL. Cells grew slowly and on the 12^th day the medium was changed to include 0.2% serum to all conditions. This increased cell growth and on the 16^th day, half the flasks were given ascorbate (50 μg/ml) and the medium changed daily. The conditioned medium (24 h) was collected and saved over the next 4 days to assay for collagen production. Embryonic chick tendon cells “synthesize and secrete prodigious amounts of procollagen (about 0.6 mg/10⁹ cells/h) for the first 8 h or so in culture” (Prockop, Sieron & Li, 1998). One can calculate that in embryonic cells induced with ascorbate—10 ml/flask, ~10⁶ cells, incubated for 24 h—that taking 35 μl and loading it on an SDS gel would yield ~30 ng of a procollagen α1 band and ~15 ng of a procollagen α2. This would be easily detected using Sypro ruby staining. However, it would be just faintly detectable in uninduced cells (minus ascorbate) making 6-fold less procollagen. In earlier experiments, adult tendon tissue made ~1% (Schwarz, Farson & Bissell, 1979) and this would be undetectable. Using this assay, the conditioned media from adult tendon cells on days 17, 18, 19, and 20 were assayed for procollagen production. The Sypro ruby stained gel with samples from day 20 is shown in Fig. 5. As expected, only serum proteins are prominent in all lanes (compare lane 10, medium only). Since we are only applying 1/286 of the total secreted proteins to the gel, this is too little to detect except for procollagen bands with a molecular weight around 150 kD, especially in the ascorbate (+C) induced cells (lanes 6–9) and faint bands in the uninduced cells (lanes 2–5). The upper band is present at twice the level of the lower band (1.8 ± 0.3, n = 14; Table 1). The molecular weight of the upper band was 152.7 kD (±2.2, n = 11) and the lower band 139.8 kD (±2.9, n = 11). The curvature of the bands on the gel from the outside lanes to the middle was corrected by calculating the molecular weight of the bright serum band (169.1 kD) running in a lane closest to the MW markers. In other lanes, changes in apparent molecular weight of that band because of curvature would be calculated and a correction applied to other bands in that lane. This only altered the average molecular weight of the procollagen band by a small amount but it reduced the standard deviation by 80% (Table S1). These three parameters were all consistent with the upper band being procollagen α1 and the lower band being procollagen α2.

Quantifying the bands intensity was done on Azure Biosystems 600 imager. Sypro ruby is known to give a linear increase in fluorescence with protein concentration. RNAse A was used to generate a standard curve (Fig. S2) and this was used to translate band intensity to ng/band. A correction was also made for loading and/or staining variations. For instance, in Fig. 5 the intensity of the serum band running at 169 kD is higher on the left side (lanes 2–6) than on the right side (lanes 7–10). We took the average intensity of this serum band in the +C bands and used that to create a correction factor for each lane (Table 1). We should point out a limitation of this assay is a lack of internal control. When one measures percent procollagen production, the non-collagen protein synthesis compensates for changes that could affect all proteins. For instance, if the cells run out of nutrients during the 24 h incubation, this would reduce all protein synthesis. Similarly, an under estimation of cell number would over estimate procollagen production. However, in this case, the induction of procollagen synthesis by ascorbate and cell density can be as much as 60-fold and this becomes the overwhelming change that we are observing. Another advantage is that this procollagen assay has minimal manipulation.

A plot of days after ascorbate induction vs ng procollagen synthesized per 35 μl sample was created (Fig. 6). On day 17 all samples were clustered around 25 ng procollagen and this is probably due to a restriction that ascorbate induction requires two days to reach maximum levels. The samples taken from the 10x SNZR PL+0.2% FBS+C made lower levels and only on days 17 and 20 were the levels sufficient to be analyzed. The other conditions 0x SNZR PL+0.2% FBS+C, 3x SNZR PL+0.2% FBS+C, 5x SNZR PL+0.2% FBS+C were all making high levels of procollagen on days 18, 19, and 20. If the three conditions and nine independent assays are combined, the average value 42 ± 12 ng of procollagen strongly supports that adult tendon are capable of making prodigious levels of procollagen if they are allowed to grow to high cell density (Table 1). Their response to ascorbate is also very similar to the embryonic tendon cells. Their proliferation rates seem to be lower than embryonic cells (Fig. 7) but the initial isolation procedure would have to be improved to confirm this. The adult cells reached a lower confluent density even in the presence of 0.2% FBS and various levels of SNZR PL and the morphology of the cells was more diverse including cells that were flatter and larger than confluent embryonic cells (Figs. 7C and 7F to Fig. 4 day 11). But most important, assays of collagen production in the intact tissue were ~1% that is more representative of an average protein compared (Schwarz, Farson & Bissell, 1979) to the high cell density cultures making large amounts. This is one the rare cases where cells in culture are more differentiated than in vivo.