Cell-Free Gene Expression in Bioprinted Fluidic Networks

The realization of soft robotic devices with life-like properties requires the engineering of smart, active materials that can respond to environmental cues in similar ways as living cells or organisms. Cell-free expression systems provide an approach for embedding dynamic molecular control into such materials that avoids many of the complexities associated with genuinely living systems. Here, we present a strategy to integrate cell-free protein synthesis within agarose-based hydrogels that can be spatially organized and supplied by a synthetic vasculature. We first utilize an indirect printing approach with a commercial bioprinter and Pluronic F-127 as a fugitive ink to define fluidic channel structures within the hydrogels. We then investigate the impact of the gel matrix on the expression of proteins in E. coli cell-extract, which is found to depend on the gel density and the dilution of the expression system. When supplying the vascularized hydrogels with reactants, larger components such as DNA plasmids are confined to the channels or immobilized in the gels while nanoscale reaction components can diffusively spread within the gel. Using a single supply channel, we demonstrate different spatial protein concentration profiles emerging from different cell-free gene circuits comprising production, gene activation, and negative feedback. Variation of the channel design allows the creation of specific concentration profiles such as a long-term stable gradient or the homogeneous supply of a hydrogel with proteins.

The mScarlet-I (mSc) and mTurquoise2 (mTur) proteins were produced with our housemade E. coli based cell extract as described before.GFP was synthesized in commercial myTXTL Sigma 70 Master Mix with the control plasmid P70a-deGFP.The original samples were prepared with standard concentrations.For the diluted samples, the concentrations of all components except the template plasmid were halved by adding the same volume nf water.We consider a simple kinetic model for the expression of fluorescent proteins in the cell extract that accounts for the degradation of resources over time.We will use the following abbreviations for the involved species and their concentrations.We further introduce the dimensionless variable X that accounts for the availability of resources in the extract.
Transcription, translation and protein maturation are modeled by the following equations: (1) Here, α R , α P are the RNA and protein production rates, δ R is the RNA degradation rate.The term δ 1 corresponds to the maturation rate of the fluorescent protein.Moreover, protein production (Eq.2) is assumed to depend on the presence of RNA and the availability of resources, which are measured by the (dimensionless) parameter X.As mentioned in the main text, our experiments indicate that the fluorescent proteins are stable on the time scale of the experiment, and we thus omit a degradation term.We assume simple first order degradation of the resource term: which leads to The first equation has the solution: As RNA is expected to be unstable in the cell extract (δ R > δ 1 , δ 2 ), we make the simplifying assumption that RNA is present at its steady state concentration This reduces our initial model (Eq.1-4) to: where The first equation has the solution (initial condition [P](0) = 0): Integrating of Eq. 8 then gives, with [F](0) = 0, which is the Equation (1) used in the main text.
For further analysis, we normalized the data by dividing through the end level value ξ, which is: This endpoint normalization yields an expression [F](t)/ξ, which reduces the number of fitting parameters from three to two (δ 1 and δ 2 ) to avoid over-fitting.In Figure S4-S6, we show fits of that expression to the data for different values of δ 1 and δ 2 , for the original and diluted samples for different agarose concentrations.

Remarks
The fits show overall good agreement with the data, with realistic values for the lifetime of the cell extract (on the order of 4-5 hours in the 0% samples, and 7-10 hours in the gel containing samples).With about 40-50 min, the time scale for the second fit parameter δ 1 corresponds well with that expected for the maturation of a fluorescent protein.
It is impossible to fit the sigmoidal shape of the fluorescence time courses with a single rate parameter.We also considered other processes that could generate the initially observed lag phase.In fact, there is no clear time scale separation between the RNA degradation time scale (1/δ R ), which is expected to be on the order of minutes up to tens of minutes, and the time required for protein maturation (1/δ 1 ).Omitting the protein maturation step, and instead including RNA degradation, results in the following, alternative expression for the fit function: Fitting this function appeared to result in unrealistically long RNA lifetimes on the order of hours, however.
One could include both processes -RNA degradation and protein maturation -in the fitting procedure, but many combinations of δ 1 and δ R would result in similarly good fits.
To avoid overfitting, we decided to only use two fit parameters, bearing in mind that the fitted parameter might not only represent the maturation phase of the fluorescent protein, but contain also contributions by the RNA transcription and degradation process.

Figure S2 :
Figure S1: Fluorescein-filled channel structures in agarose hydrogel.We printed sacrificial structures consisting of Pluronic F-127 and cast preheated agarose on top.The Pluronic ink could be removed after cooling to 4 • , leaving vascularized hydrogels that were filled with fluorescein solution for visualization.Various designs of different complexity are shown here, ranging from a simple curved serpentine (a), a branched structure with parallel channels (b), and a grid network (c).All scale bars represent 3 mm.
[R] denotes the concentration of RNA, [D] the constant concentration of plasmid DNA provided (2 nM), [P] is the concentration of the nascent fluorescent protein P, while [F] is the concentration of the fluorescent protein F after the maturation step.The recorded signals are proportional to [F].

Figure S4 :Figure S5 :Figure S6 :
FigureS4: Comparing different δ 2 values in the normalized kinetic fit of 1% samples.The normalized protein production was fitted to the kinetic data of mScarlet-I (mSc) gene expression with fixed guess-values for the parameter δ 2 .A range of 0.2 -0.3 1/h was compared for the original 0% sample set (upper plots), and 0.2 -0.3 1/h for the diluted 0% sample set.

Figure S7 :
Figure S7: Single-channel chamber.(a) Illustration of the container design with dimensions.(b) Photograph of single sample.The container was glued to a circular glass slip with diameter 20 mm and filled with agarose.The chamber surface was sealed with oxygen permeable foil and the inlets with Picodent.(c) Photograph of four samples placed into a 12-well plate.A cylindrical PLA cover was applied to each relevant well for avoiding sample-crosstalk.

Figure S8 :
Figure S8: Additional datasets to Figure 3 of cell-free protein synthesis and diffusion in singlechannel hydrogels.The samples were constructed as shown in the schematics (a,d) featuring a central channel within an agarose hydrogel.These channels were filled with original mixtures containing 75 % cell-free extract and buffer plus the indicated plasmids.The hydrogels were composed of the diluted mixture with 2 % agarose, 37.5 % extract and buffer, and for (d) additionally of mTurquoise2 (mTur) plasmid.(b-c) Profile plots over time of mScarlet-I (mSc) fluorescence intensity, obtained by averaging the scan area over the vertical direction.(e-f) Temporal profile plots showing mTurfluorescence intensity averaged over the vertical direction.Gene expression of the reporter was activated by the σ 28 produced in the channel only.

Figure S9 :
Figure S9: Additional datasets to Figure 3 and control of feedback gene circuit.(a-b) Plasmids encoding the two regulatoray proteins sigma factor (σ 28 ) and repressor TetR, and two reporter proteins mTurquoise2 (mTur) and mVenus (mVen) were supplied into central channels of agarose hydrogel supplemented with cell-free reaction mixture.The fluorescence intensity within the channel area was monitored and the average value plotted over time.(c) Two control datasets without the plasmid encoding the repressor TetR.The other three plasmids of the circuit were injected within cell-free reaction mixture into agarose hydrogel channels.

Figure S10 :
Figure S10: Additional datasets of protein synthesis in printed channel structures.The channels were filled with original mixtures comprising 75 % cell-free extract and buffer plus mScarlet-I (mSc) template plasmid.The hydrogels consisted of the diluted mixture with 2 % agarose and 37.5 % extract and buffer.(a-c) Single-channel design with temporal fluorescence intensity profile plots, generated by averaging the scan area over the vertical direction.(d-f) Dual-channel design and corresponding fluorescence intensity profile plots over time, created by vertical averaging.

Figure S11 :
Figure S11: Cell extract quality control.(a) We measured the protein content of our 1:10 and 1:20 diluted homemade cell extract in a BCA-assay using a bovine serum albumin (BSA) standard.(b) The extract activity was evaluated by measuring the increase in fluorescence intensity by mScarlet-I (mSc) protein expression within the 0 % original sample.