On-chip fabrication and in-flow 3D-printing of microgel constructs: from chip to scaffold materials in one integral process

Bioprinting has evolved into a thriving technology for the fabrication of cell-laden scaffolds. Bioinks are the most critical component for bioprinting. Recently, microgels have been introduced as a very promising bioink, enabling cell protection and the control of the cellular microenvironment. However, the fabrication of the bioinks involves the microfluidic production of the microgels, with a subsequent multistep process to obtain the bioink, which so far has limited its application potential. Here we introduce a direct coupling of microfluidics and 3D-printing for the continuous microfluidic production of microgels with direct in-flow printing into stable scaffolds. The 3D-channel design of the microfluidic chip provides access to different hydrodynamic microdroplet formation regimes to cover a broad range of droplet and microgel diameters. After exiting a microtubing the produced microgels are hydrodynamically jammed into thin microgel filaments for direct 3D-printing into two- and three-dimensional scaffolds. The methodology enables the continuous on-chip encapsulation of cells into monodisperse microdroplets with subsequent in-flow cross-linking to produce cell-laden microgels. The method is demonstrated for different cross-linking methods and cell lines. This advancement will enable a direct coupling of microfluidics and 3D-bioprinting for scaffold fabrication.


Introduction
Bioprinting is an emerging technology for the fabrication of three-dimensional cell-laden structures to mimic or replace body tissue [1].It plays an important role for tissue engineering and drug delivery, as well as for the study of diseases and the development of treatments.In bioprinting, cells and materials are formulated to a bioink [2,3] which is then directly printed into a hierarchically structured 3D scaffold.The ultimate aim of bioprinting is to produce cell-laden scaffolds that exhibit the full or partial function of the target tissue or organ, based on the rationale that the three dimensionally fabricated structural arrangement facilitates and accelerates biological maturation [4].
Bioinks are thus a crucial component of the bioprinting technology.They rely on materials, mostly hydrogels, which can encapsulate cells and bioactive molecules, and can be used in suited printing technologies such as extrusion printing to fabricate the desired three-dimensional scaffolds or constructs.The design of bioinks is challenging, because their flow and elastic properties need to be finely adjusted to be sufficiently shearthinning during extrusion, and subsequently developing the desired mechanical stability and elasticity quickly after solidification to form a stable scaffold.Furthermore, bioinks must be biocompatible, sufficiently biofunctional to promote cell viability and proper post-fabrication behavior, and be well perfusable for sufficient provision of nutrients and oxygen.
In order to meet these challenges, Highley et al [5] recently proposed the use of microgel bioinks.Due to their colloidal nature, microgel bioinks are well shearthinning and rapidly solidify under quiescent conditions, while cells loaded into the soft colloids are shear protected.Printed microgel scaffolds can be further stabilized by secondary cross-linking.Microgels further offer the advantage to tailor the cell microenvironment.Methods to prepare cell-laden microgels have already been developed, particularly in the area of microfluidics where microgels with very uniform size can be fabricated in a continuous process [6][7][8].It was demonstrated that microgels can be formed by physical cross-linking such as via alginate/Ca 2+ ionic complex formation [9,10] which can be modulated by EDTA-complexation [11,12], or by thermally induced gelation such as via cooling gelatin solutions below 20 • C [9,13].Chemical cross-linking reactions offer greater stability and better mechanical properties of the microgels.Examples are Michael type reactions as for functionalized gelatins, hyaluronates, polyethylene glycols or polyglycerols [12,[14][15][16], azide-alkyne click reactions as for polyglycerols [17] and photo-crosslinking [18] which requires a photoinitiator and crosslinking groups as has been shown for polyethylene glycol.Microfluidics-based microgel or microsphere fabrication with successive assembly steps have been recently demonstrated in [19] and [20].Cells that have been encapsulated include stem cells [9,12,14,15], Crypt and Peyer cells [10], liver cells (HepG2) and endothelial cells (HUVEC) [18], as well as NIH3T3 fibroblasts [6].The fabrication of functional scaffolds based on cell-laden microgels has so far been shown by Fan et al [21] using emulsion-based fabrication of gel-MA microgels and by Compaan et al [22] for gelatin microgel filler particles.For microfluidic generated microgels this has recently been demonstrated for the first time by Highley et al [5].
The reason for the so far limited number of studies on microgel based bioinks and their bioprinting to scaffolds is the required combination of microfluidics that produces only small microgel quantities, followed by a number of post-chip batch process steps involving cross-linking, the preparation of the bioink, and the subsequent bioprinting into the desired scaffold.This currently makes microgel biofabrication a time-consuming, low-productivity multistep process.
It would therefore be highly desirable if microfluidics for the production of the microgels and bioprinting for the fabrication of the desired scaffolds could be integrated into one continuous, automatable process.Here we show, that microfluidic chips can be designed for the on-chip encapsulation of cells into droplets, which are photo-crosslinked in-flow to produce microgels, which are then automatically jammed from a downstream microtubing to continuously form thin microgel filaments.The microtubing is integrated into the printhead of a 3Dprinter to directly in-flow print the filaments into three-dimensional scaffolds within a prefabricated permeation chamber.We demonstrate that the printed microgel scaffolds have sufficient intrinsic microporosity for continuous permeation of medium, and demonstrate the integration of additional complementary perfusable microchannel systems connected to the in-/outflow ports of the permeation chamber.

Materials and methods
We here provide a brief summary of the methods.Details of the materials, materials synthesis and methods are provided in the supporting information.
Device fabrication: the multi-layer master structures were produced using a 3-layer photo lithography process.The microfluidic devices were prepared from Sylgard 184 ® poly(dimethylsiloxane) elastomer using a standard soft lithographic approach.
Droplet production and cell encapsulation: droplets were produced by focusing an aqueous center stream containing the polymer precursors (polyoxazolines, alginates) and the cells (yeast, HEK293T, NIHT3T, BJ1-TERT morphology reporters) with an oil phase through a narrow channel section into the main downstream channel.Droplet formation was followed with a light microscope and a high-speed video camera.
For the aqueous streams we used as polymer solutions a 0.25 wt% solution of alginate, 5 wt% solutions of mixtures of ene-and thiol-functionalized poly(ethyloxazolines), and 7.5-15 wt% solutions of mixtures of ene-and thiol-functionalized poly(methyloxazolines) in combination with high and low molecular weight pentenoic acid modified gelatin (GelPa).ADA-GEL was used as 5 wt% solutions prepared at 37 • C, and then used at room temperature.Cell concentrations in the aqueous stream varied between 1.5 and 3 mio.cells ml −1 .To prevent cell sedimentation in the syringes, the cell suspensions were mixed 1:3 with an OptiPrep (16 wt%) solution.The oil phases (1-octanol, 1-decanol, 1undecanol, nonanoic acid, mineral oil) contained 1-3 wt% SPAN80 as an emulsifier to stabilize the water/oil emulsion droplets.Typical water flow rates were 100-1000 µl h −1 for the aqueous phase and 600-3000 µl h −1 for the oil phase.For photocrosslinking, 1 wt% (with respect to polymer) lithium-phenyl-2,4,6-trimethylbenzoylphosphinate (LAP) was added as a photoinitiator.Calcium-2-ethylhexanoate was added as an alginate crosslinker in oil at the saturation concentration.
Photo crosslinking and 3D-printing: photo crosslinking of the polyoxazoline precursor via a thiol-ene reaction was performed in-flow in a microtubing with a 385 nm power LED.The microgel filaments were 3D-printed with a commercial 3D-printer (Creality Ender-5) into a three-dimensional scaffold within a flow chamber.

Microfluidic device and controlled droplet production
Our aim was to design a microfluidic chip for (i) low shear stress cell encapsulation, (ii) variability for physical or chemical cross-linking, (iii) a large variation of microdroplet diameter, and (iv) capability to couple it directly to a 3D-printer.To allow performing different physical and chemical crosslinking reactions, the entrance channel design should allow sequential mixing of the solutions containing cells, polymers, cross-linking and further agents.The microdroplets should have controllable diameters from 300 µm down to 50 µm to reach celldensities of 10 6 cells ml −1 if single cell encapsulation would be desired.
We therefore used a channel design that allows three-dimensional flow-focusing with two subsequent mixing crosses, followed by a downstream narrow orifice for controlled droplet formation.The design is schematically shown in figure 1.It contains an inlet channel for the center stream solution containing the cells and the precursor polymers, and two side channels which can contain buffer solutions, culture media, bioactive substances or crosslinking agents.The side channel streams focus the inlet channel stream three-dimensionally into the center of the channel where shear-forces on the cells are minimal.Subsequently, the aqueous stream is three-dimensionally focused with the oil phase to enter a narrow orifice section that controls droplet formation.The narrow section enables to access different hydrodynamic regimes to vary the droplet size over a wide range.The downstream channel is kept sufficiently narrow such that the droplets form a stable droplet train on the channel center streamline.The 3D double focusing chips were fabricated by soft lithography using a multilayer technique, and the flows simulated as described in the supporting information (figures S11-S13).
Droplet break-up occurs when the viscous shear force F shear exerted by the outer fluid exceeds the pinning interfacial force F γ arising from the surface tension.Both forces are directly related by the dimensionless Capillary number Ca = F shear /F γ , which can be calculated in terms of the viscosity of the continuous oil phase η, the average inlet flow velocity v, and the surface tension γ as Ca = F shear Fγ = ηv γ .Depending on the Capillary number, different relevant hydrodynamic regimes for droplet generation can be distinguished: (a) squeezing regime with channel geometry-controlled break-up (Ca < 0.1), (b) dripping regime (0.1 < Ca < 1) and (c) jetting regime (Ca > 1) [23][24][25][26][27].As shown in figure 1, with the variable 3D-constriction design all three hydrodynamic regimes for droplet production are accessible, and the Capillary number is the main control parameter for droplet production.It allows to precisely control the droplet diameter in the target range between 50 to 300 µm by adjusting the volumetric flow rate, oil viscosity and interfacial tension.The respective viscosities and interfacial tensions are summarized in table SI in the supporting information.

Cell encapsulation
The cells were encapsulated into the droplets by dispersing them at a given concentration in the center stream (see figure 1).The flow-focusing into the center streams at the two mixing crosses is shown in the supporting information (figure S27).For initial tests we used yeast cells (saccharomyces cerevisiae) that were fluorescently labeled with Nile Red.To prevent sedimentation in the syringes, we used OptiPrep solution (16 wt%).We further investigated the encapsulation of Yellow fluorescent protein (YFP)-tagged HEK293T cells, NIH3T3 cells and BJ1-TERT cells.The number of encapsulated cells per droplet follows a Poisson distribution, as shown in the supplementary information (figure S21).The microgels were fluorescently labeled by the incorporation with FITC-dextrane which was added in the center stream A. Polyoxazoline, GelPa, alginate and ADA/GEL were used to as microgel components.

In-flow photo-crosslinking
To enable continuous on-chip crosslinking, the crosslinking reaction times should be in the millisecond regime such that for routine flow velocities of v = 2-10 mm s −1 full conversion is reached during the residence time of the microdroplets in the respective microfluidic channel section.It has been previously shown that UV-crosslinking of polymer precursors within microdroplets is a fast and efficient method to produce microgels of controlled size and shape [28][29][30].We employed an LED operating at an emission wavelength of λ = 385 nm suitable for the efficient use of the water-soluble photoinitiator LAP for a thiol-ene reaction between thiol and vinylfunctionalized polyoxazolines.The LED has an illumination power of P = 530 mW from a 1 mm 2 source area.The LED power can be varied to reach high conversion at minimal illumination time to minimize potential cell damage.For a standard flow velocity of v = 10 mm s −1 , the residence times over an illuminated distance of 1 mm is t = 100 ms.Thus, during this time the droplet is illuminated with an energy of up to E = P .t = 5.3 J cm −2 .This is more than 10-times higher than an energy density of E = 0.45 J cm −2 reported for complete photo-crosslinking of microgels using a power density of P = 15 mW cm −2 over a time period of 30 s [5].Thus, we expected to reach full conversion for continuous in-flow crosslinking using LAP as a photoinitiator and functionalized polyoxazoline.
In the first device versions of the present work the LED was positioned directly above the microfluidic reaction channel.We found that the onchip LED illumination caused prepolymerization of the polymer precursor in the entrance channel sections due to the strong light scattering of the chip material.Since the Rayleigh scattering intensity depends on the wavelengths as 1/λ 4 , short wavelength light shows very high scattering intensity.We therefore performed the in-flow LED photo-crosslinking step in a separate and shielded microtubing section downstream the microfluidic chip as shown in figure 2. We found that an issue with having the cross-linking step in the downstream microtubing section is the occurrence of microgel bunching.This leads to the formation of microgel clusters having aperiodic exit times at the microtube exit.This disturbs stable microgel filament formation which is necessary for subsequent 3D-printing, as discussed below.
An important factor causing droplet and microgel bunching is the difference between the microchannel and microtube width.It is particularly critical for the transition section from microchannel to microtube, which often is designed as a 90 • -bend where the horizontal microchannel enters the vertical punch hole containing the microtube.If the punch hole or the microtube have larger diameters than the microchannel, radial velocity components in the widening sections move the droplets or microgels to off-center streamlines, particularly in bent sections.As in the widened sections the center streamline, and even more the off-center streamlines, have lower flow-velocities, microgels jam into contact, thereby forming bunches or clusters.This was investigated in more detail using FE-simulations as described in the supporting information (figures S17 and S18).
We therefore transferred the droplet stream from the microfluidic exit channel (250 µm diameter) into a microtubing (PE, 280 µm inner diameter) having nearly the same inner diameter in a straight co-axial connection as shown in figure 2. We observed that upon transfer of the droplet train into the microtubing, the droplets remain well separated and localized in the center stream.With this Figure 2. Scheme of the microfluidic/3D-printing device as used for (a) physically cross-linked microgels with the example of alginate/Ca 2+ , and for (b) photo-chemically cross-linked microgels with the example of polyoxazoline thiol-ene photo-crosslinking using LAP as a photo-initiator.For each case, the components used in the inlet channels are indicated.Photos of the droplet generation, of the droplet transfer from chip to microtube, of the microgel jamming process, and of the resulting microgel fiber are provided, together with videos in the supporting information.The print head is driven by a commercial 3D FDM-printer.
setup we achieve stable in-flow photo crosslinking of separated droplets for continuous microgel production.

Direct in-flow 3D printing
Our aim was to integrate microfluidics with 3D printing into a continuous process.We therefore skipped the bioink fabrication steps and directly inserted the microtubing outlet of the microfluidic device into the print head of a commercial 3D-printer used for fused deposition modeling (FDM).FDM is a 3D printing method that is based on the extrusion of a continuous filament of a molten thermoplastic material through a moving print head nozzle to fabricate freestanding three-dimensional constructs.Fused filament deposition is the most widespread printing methodology for additive manufacturing.It has been originally developed to print filaments from thermoplastic materials, and later been extended to print hydrogel filaments for the fabrication of biological tissue constructs.
A first precondition for the direct coupling of microfluidic flows and 3D-printing is that the flow velocity v m of the microgels is equal to the velocity v p of the print head.In microfluidics, routine droplet flow velocities are in a range of v m = 1-50 mm s −1 , which notably is in the range of typical FDM print head speeds of v p = 10-100 mm s −1 or bioprinting nozzle speeds of 1-50 mm s −1 , at pressures of 0.5-3 bar [31].Therefore, direct coupling is possible with FDM printing as well as with bioprinting commercial devices.
A second precondition is that filaments are formed after exiting the print head microtube.We observe that upon exiting the microtube the microgels jam along their center streamline to form stable pearl-necklace type filaments.A video of the jamming process is provided as a supporting information, together with a video snapshot which is shown in figure 2. Jamming occurs along the center streamline directly after exiting the microtube, where the overall fluid stream suddenly widens, and concomitantly the center stream velocity suddenly decreases, such consecutive microgels jam into contact.After jamming, microgel adhesion stabilizes the pearl-necklace type filaments.
Microgel adhesion needs to be sufficiently fast and strong to stabilize the formed filaments for 3Dprinting.The filaments are initially stabilized by contact adhesion (tack) of the microgels.Contact adhesion, e.g. as in pressure-sensitive polymer adhesives, can be mediated by chain interpenetration and entanglement formation of highly mobile chain segments as at the periphery of the microgels.In our case for the alginate microgels, the respective filaments are further stabilized by Ca 2+ -crosslinking with Ca-salt present in the liquid phase containing the printed construct.For the polyoxazoline microgels rely on weak adhesion, which however, is sufficient to stabilize initially scaffolded constructs over a short time.
The freshly formed filament is then deposited at the desired location by the print head.For this, the microtube end is positioned ca. 1 mm above the substrate or construct surface.We find that standard substrates such as polystyrene or glass provide sufficient adhesion to deposit filament surface patterns.After 3D-filament deposition the oil phase separates and floats atop the aqueous medium, where it can be removed by a pipette or syringe.We note that in the case of mineral oil and polyoxazoline microgels filament formation was difficult, since the oil/microgel phase separation is retarded.
As the two preconditions can be well fulfilled, a direct coupling of microfluidic and 3D-printing platforms is possible.We find that using a typical polymer stream volumetric flow rate of e.g.Q = 1000 µl h −1 , construct volumes of 1 cm 3 h −1 can be printed with microtube inner diameters of 280 µm, strand widths of 200-500 µm, and printing velocities of v p = 5 m s −1 .This is in the range of commercial 3D-printing or bioprinting devices [32].For typical microgel diameters of 100 µm and cell numbers of 5 cells per microgel, cell densities of 10 7 cells cm −3 can potentially be achieved, which is suitable for tissue engineering applications.The direct coupling of microfluidics and 3D-printing enables continuous one-step construct fabrication from microgels, thereby removing the role of microfluidics as a typical fabrication bottleneck.The formulation of shearthinning bioinks with appropriate elastic properties is not necessary, which also removes the issue of large bioprinting nozzle shear stresses with its negative effects on cell viability.
To assess the capabilities and current limitations of the microfluidic printing process, we first investigated the printing of standard line and meander patterns that are used for biofabrication.Figure 3(a) shows a standard parallel line pattern and figure 3(b) a meander pattern where line widths of 300 µm can be achieved.We find that for the alginate microgels the line widths are nearly equal to the filament diameters (see figure 2(a)), which are equal to the microgel diameters, such that potentially very thin filaments and scaffold lines can be fabricated in a controlled way.
We find that the current main limitation of the method is the relatively weak initial adhesion between the microgels, which limits the stability of the filaments and of the printed constructs.This can be observed in figure 3(c), where microgel filament lines have been deposited on top of each other, but due to the weak adhesion, the microgels show the tendency to slide off, leading to line broadening to 1 mm in figure 3(c), and to construct instability.
Microgel adhesion, and thus the stability of the filaments and printed constructs, can be improved by a second crosslinking step.For alginate-and ADA/GEL-microgels this is possible by printing into a medium containing Ca 2+ -ions, leading to inter microgel-crosslinking.POx-based scaffolds can further be stabilized by using polymer solutions with an excess ene-component.After production, an aqueous solution of thiol-containing POx can be added to the scaffold.A second UV irradiation after 2 min of incubation leads to scaffolds that are stable in aqueous medium for at least two weeks.(see SI figure S25) As microgel adhesion can be improved, but is not very fast, it is still currently difficult to directly print three-dimensional constructs with large aspect ratios or overhangs, such as for 3D-crossbar patterns.Yet, 3D-objects with low aspect ratio such as the donut-shaped structure in figure 3(d) can readily be printed.
Prolonged contact between microgel filaments over minutes, leads to the desired strong adhesion.Figure 3(e) shows a cube-shaped 3D-construct consisting of microgel filaments printed in cross-bars.During printing the construct is supported by an external scaffold that can be removed directly after printing to release the self-supporting construct.We generally observe that the initially weak contact adhesion between the microgel filaments after printing develops into a sufficiently strong adhesion to support high aspect ratio structures.We attribute the developing adhesion to chain interpenetration and entanglement formation at the periphery of the microgels, thereby mediating adhesion (tack) between microgels already at short contact times and low contact pressures.Additional adhesion can be provided by further physical or chemical crosslinking, as described above.
A further feature observed in figure 3(d) and f is the intrinsic microporosity of the microgel constructs.As the printed microgel filaments initially have limited contact adhesion, they show variable displacements from the line patterns.This leads to void structures, whose extension, diameter and volume depend on the printed line density.These voids provide microgel constructs with an intrinsic microporosity as seen in the donut shaped construct in figure 3(c).Typical voids are in the form of channels with widths of the order of the microgel diameters, i.e. 50-200 µm.Depending on the line density, microchannel lengths can be of the order of a few hundred microns up to the dimension of the construct.The intrinsic microporosity of microgel constructs is important for perfusion to provide cells with nutrients.Culture media can perfuse through the micropore channels at much higher velocity compared to solid homogeneous hydrogel constructs.Thus, for microgel constructs the integration of threedimensional micro vascular structures is not necessary for the provision of nutrients or drugs to encapsulated cells.
We find the currently achievable line width of microgel filament deposition to be in the range of 200-500 µm.In principle, for ideal pearl-necklace filaments that could be positioned with sufficient precision, the achievable line widths would be equal to the microgel diameters, i.e. in the range of 50-200 µm.In practice we needed to let the microgels jam into slightly thicker filaments with cross-sections containing typically 2-5 microgels.This thickness is a compromise between a printhead velocity to be as large as possible to draw thin microgel filaments, but not too large to avoid filament rupture.If the contact adhesion between microgels and microgels filaments could be further increased, the direct fabrication of high-aspect ratio structures with down to 100 µm feature size should well be possible.Yet, with the temporal stabilization of the constructs by scaffolds during the printing process, a direct fabrication of threedimensional microgel constructs is already possible as will be subsequently demonstrated.

Direct in-flow 3D printing of permeable constructs
In figure 4 we show the direct filament printing of 3D constructs into a perfusion chamber by using stabilizing scaffolds, which can be removed after printing.Figure 4(a) shows a scheme of the perfusion chamber having in-flow and out-flow ports C in and C out connecting to an upstream and downstream chamber, which contain the scaffolds.With the design in figure 4(a) simple rectangular box scaffolds can be printed, which seals to the side walls of the perfusion chamber.Other scaffold shapes are possible to produce microgel constructs of variable shapes.
The perfusion chamber serves to provide stable media perfusion for long-term culture, and to be compatible with optical and fluorescence microscopy for live cell imaging.The chamber is fabricated from an FDM-printed PLA-housing that is glued onto a thin petri dish polystyrene base plate having the dimension as a standard microscopy slide.The chamber can be connected via Luer lock inserts to external tubings.The perfusion chamber with the integrated construct can be used as an open device to manually add solutions and reagents, or sealed with a thin cover slide to obtain a closed system.The temperature is controlled by mounting on a Peltier stage.The permeation chamber allows to provide the construct with fresh medium, and to perform construct perfusion experiments with controllable flow rates.
The oil phase separates from the aqueous phase and the microgels due to its lower density.We observed that for some viscous oils the separation process can be slow, especially for some oil/emulsifier combinations (e.g.certain mineral or plant oils) requiring additional means to accelerate phase separation.
Scaffold removal can be done after printing for reusable scaffolds, or by subsequent dissolution in the aqueous medium for sacrificial scaffolds.We observe that microgel filament adhesion is sufficiently strong to provide cohesion and thus three-dimensional mechanical stability of the printed constructs in water and in air.This is shown in figure 4(b), where the scaffolds have been removed from a 14 × 4 × 4 mm 3 (width × depth × height) construct directly after printing.Here, the construct is printed to be in contact with the perfusion chamber side walls for sealing and subsequent perfusion experiments.Examples of completely free-standing 3D-constructs are shown in the supporting information.We find that the microgel constructs well preserves the shape of the scaffold and are long time stable as shown in the supporting information (figures S33 and S34).
We observe that the intrinsic microporosity of the microgel constructs provides complete permeation already at hydrostatic pressures of a few millimeters, which can be realized by filling aqueous solutions into the upstream chamber.The confocal microscopy image in figure 4(c) shows the hydrostatic pressuredriven permeation of a solution of a fluorescent dye (ATTO 647N) through the construct via the micropores.We observe that the fluorescent solution front advances via advection through the micropore channels, together with the subsequent diffusion into the microgels, which is apparent from the increasing homogeneous fluorescence intensity in the construct near the construct edge.
As examples for cell encapsulation we chose yeast cells (S. cerevisiae) for optimization of the method, and YFP-tagged HEK293T-cells for the final device tests.The confocal microscopy image in figure 4(d) shows a homogeneous distribution of the yellow fluorescent HEK cells in the final microgel construct matrix, imaged in the perfusion chamber.
When printed into the perfusion chamber the constructs can be fabricated to make direct contact and sealing with the side walls, and direct contact with the top cover slide to allow continuous microperfusion of the whole construct.This configuration can be used for applications involving cell tests for drugs, or as bioreactors where production cells in the constructs convert precursors to molecular or macromolecular drugs that are released into the downstream chamber for continuous production.

Integration of complementary perfusion channels
Because of the yet limited filament contact adhesion the direct printing of larger perfusion channels is still difficult.To fabricate 3D-constructs with integrated perfusable microchannels we follow the same printing method as described above, employing a scaffold that can be removed after printing.The integration of the microchannel proceeds via a layer-by-layer procedure as schematically shown in figure 5(a).After the first layer is printed, a hollow polyvinylalcohol (PVA) microtube is inserted on top of the layer.The microtubes are fabricated by 3D-FDM printing from commercial PVA filaments with 0.4 mm inner diameter and 0.9 mm outer diameter.PVA dissolves well in water, is an FDA-approved polymer, and has excellent FDM-printing properties.We found that a wall thickness of 1 mm provides sufficient integrity of the PVA microtube during perfusion over time periods of several hours, before it dissolves.To provide sufficient microchannel integrity after PVA-dissolution, we coated the PVA microtubes with a photo-crosslinked polyoxazoline layer using the same photochemistry as for the polyoxazoline microgel fabrication.The microtube inlet and outlet were equipped with short PDMS microtubing connections.These serve to directly as connections of the microtube to the inlet and outlet perfusion needles of the perfusion chamber directly after insertion of the microchannel, as shown in figure 5(c).
After integration of the microchannel, microgel filaments are printed around the microtube to finish the second layer.Subsequently, a third layer is printed on top of the second layer to finish the construct.The microchannel is already functional after its integration and can be perfused during the subsequent construct printing process.This is shown in figures 5(c) and (d), where a fluorescein solution that is perfused through the microchannel can be well observed.Typical flow rates in the microchannel are 1-2 ml h −1 .At an inner diameter of 500 µm this corresponds to flow velocities of 1-2 mm s −1 which is in a typical range for blood capillaries in view of later applications.
The integration of the complementary perfusion channel system allows to microperfuse the construct as a whole via a flow c in from the upstream chamber through the microporous construct, and then exiting via the downstream chamber as c out .In addition, the construct can be separately perfused via a flow v in from the inlet port through the microchannel system, exiting downstream via the outlet port v out .The separate channel systems allows infusion with endothelial cells (e.g.HUVECs) for vascularization, and a spatially defined provision of nutrients or factors to the cells in the construct.Besides the simple straight channel shown in figure 5, also multiple channels with bifurcation points can, in principle, be integrated into the construct and connected to the perfusion chamber.

In-flow printing of alginate microcapsule constructs
So far, the method had been established for polyoxazoline microgels, which can be well fabricated in microfluidic devices by fast photo-crosslinking.To extend the method from synthetic polymers with chemical cross-linking to biopolymers with physical cross-linking, we considered the fabrication of alginate microgel constructs.Therefore, an alginate solution was focused with an oil stream (1-undecanol) into the nozzle to generate alginate solution droplets.For rapid on-chip crosslinking calcium ethylhexanoate was dissolved in the oil-phase.We find that a continuous production of alginate microcapsules at high flow rates is possible.When exiting the microtubing, alginate capsules form very stable pearlnecklace filaments as shown in figure 2. These can be printed into alginate microcapsule constructs similarly as demonstrated for the polyoxazoline microgel constructs.
We note that very recently a similar two-step method to encapsulate cells into gel beads and subsequently use 3D-deposition to print the gel beads into scaffolds has been published [33].In this case mesenchymal stem cells (MSCs) were encapsulated into 450-650 µm beads consisting of gelatin, Matrigel, HAMA or GelMa, which were then deposited by air-jet assisted printing into stable 3Dscaffolds.The authors found that using this method leads to scaffolds with increased MSC-proliferation with significantly improved therapeutic performance, e.g. for skeletal muscle and hair follicle regeneration.

Cell viability
To assess the cell viability after encapsulation, photo crosslinking and extrusion through the printer, we performed live/dead cell viability assays for NIH3T3 cells using Calcein and Ethidium homodimer-1 (EthD-1) after different production steps.The images were captured with a fluorescence microscope and live and dead cells were counted with the software Fiji.
In a first step, cell viability was investigated after encapsulation, but without printing the produced microgels with the FDM printer.Therefore, the microgels were collected directly after leaving the microfluidic chip at the microtube exit.Images were taken 1 and 7 d after cell encapsulation.We investigated microgels with low (7.5 wt%) and high (15 wt%) polyoxazoline concentration.
The cells in the microgels reveal a good viability.We observe (see figure 6) that the viability of the cells in the low (7.5 wt%) polymer concentration microgels was 62.8 ± 5.3% (n = 4) at day 1 and 84.08 ± 5.6% (n = 4) at day 7.The viability of the cells for the higher polymer concentrations (15 wt%) polymer microgels at day 1 amounts to 73.5 ± 5.8% (n = 4) and at day 7 72.2 ± 6.02% (n = 4) after microgel production.The viability in the 7.5 wt% microgels increases within 7 d, indicating a successful recovery of the cells after the microfluidic process.The viability in the 15 wt% microgels remains the same within 7 d.Due to the bio-inert nature of the used polymers, the cells are round and do not stretch in the microgels [34,35].
In a second step, cell viability was investigated after the microgels have been extruded through the printing nozzle of the FDM printer (see figure 7).Here, microgels were collected after extrusion in aqueous medium.Viability was investigated at day 1, 3, and 7 after production.We investigated microgels with 10 wt% of a mixture of thiolmodified poly(methyloxazoline) and low molecular weight GelPa (2 kDa).At day 1, the viability was 75.7 ± 11.4% (n = 4) and it did not change significantly at day 7 (76.5 ± 9.8% (n = 4)).We observe cell viabilities that are in the same range as directly after encapsulation, indicating that the extrusion process does not lead to increased cell death.
To gain further information about the behavior of cells after the extrusion of microgels, we encapsulated BJ morphology reporter cells that contain the cDNA sequence for a farnesylated tdTomato protein targeted to the plasma membrane [36].Since cells remained round in the previously used microgel compositions due to their bio-inert nature [34,35], we now used added low and high molecular weight GelPa to the POx-precursor solution for the encapsulation experiments.In comparison to prior experiments, viability improved further when using high molecular weight GelPa (figure 8(a)).Here, the reporter cells could only be stained with Calcein to detect living cells in green since the signal of EthD-1 overlaps with the tdTomato signal of the reporter cells.However, the high amount of green cells indicates improved viability.Comparing microgels containing low molecular weight GelPa (figure 8(b)) and high molecular weight GelPa (figures 8(a) and (c)-(f)) reveals differences in cell behavior.For the latter one, we observed that cells start occurring on the outside of microgels from day 1 onwards.These findings indicate that cells do not only survive the encapsulation and extrusion process but also give first evidence of cell proliferation and migration (figures 8 (c)-(f)).

Direct coupling between microfluidics and 3D-printing
The aim of this study was the integration of microfluidics and 3D-printing to fabricate microgel constructs in one continuous process.The first prerequisite to directly connect both methodologies is the compatibility of the microgel flow velocity in the microfluidic system with the print head speed of the 3D printer.The second prerequisite is the formation of the microgel filaments for 3D-printing.We find the first device-related condition to be well fulfilled for standard operation conditions for both methodologies.The fulfillment of the second condition is a surprising finding of this study.The observation of microgel filament formation can be rationalized by the reduction of the oil/water surface area leading to jamming, together with adhesion when the microgels are jammed into contact.Microgels are characterized by moderate cross-linking densities and high polymer chain mobility at their periphery [37], such that chain interpenetration and entanglement formation can mediate adhesion of microgels already at short contact times and low contact pressures.High polymer segment mobility, e.g.well above the glasstransition temperature, together with solvent swelling to facilitate entanglement formation are known properties promoting fast contact adhesion, known as tack, e.g. in polymer adhesives [38].
Thus, the direct coupling of both methodologies for the fabrication of microgel constructs is possible, and a separate formulation-step of a bioink is not needed.An issue to be further addressed is an improved filament contact adhesion for the direct fabrication of free-standing high aspect ratio constructs, which could be approached externally by temporary scaffold structures or internally by endowing the microgels with more rapidly acting physical or chemical adhesion groups.
The special microfluidic channel design allows 3D-focusing to localize the cells in the center stream to minimize shear stress.Double focusing allows cellprefocusing and on-chip addition of cross-linking agents.The 3D orifice design enables to access all relevant hydrodynamic regimes for a controlled formation of a wide range of droplet diameters, where the droplet size can be precisely controlled by the Capillary number Ca.The post-chip fast in-flow photo-crosslinking step continuously generates monodisperse, cell-laden microgels.
Besides microfluidics and 3D-printing, a further aspect was the direct integration of the printed construct into a perfusion chamber suitable for long-term live cell imaging, which can be used as a bioreactor, for cell testing and for tissue regeneration.Therefore, we realized a one-step continuous production process based on microfluidics and 3Dprinting to directly fabricate permeable constructs in functional perfusion chambers for immediate use in cell studies.Photographs of the complete setup are shown in the supporting information (figures S23 and S24).The process can in principle be automated to print series of construct/perfusion chamber designs.The direct coupling of the methods removes the role of microfluidics as a bottleneck in biofabrication processes.

Structural hierarchies
The described method allows to control the structural hierarchy of fabricated constructs on three hierarchy levels.On the first level within individual microgels, i.e. on length scales below ∼100 µm, the local cell environment (pericellular environment) can be adjusted via the control of • the microgel diameter via the droplet size and capillary number as shown in figure 3, • the number and types of encapsulated cells via the cell concentrations in the initial solutions and the flow rate ratios in the central and first focusing streams, • the local viscoelastic properties of the polymer matrix via the degree of functionalization, molecular weight of the precursor polymer, the concentration of the photoinitiator, and the light intensity, • the local (bio)chemical functionality via specific functionality of the polymer, • the internal structure from homogeneous structures as for the POx-microgels, to core/shellstructures as for the alginate capsules.
The control of the cell number is possible within the limits of the Poisson distribution, whereas all other parameters can be controlled with good accuracy.The microgel diameter controls the diffusion times of agents to the cells.The microgels thus represent micro compartments where the cells are located in a bio-chemically and mechanically tailored microenvironment.
The second hierarchy level concerns the level of the filaments and micropores within the construct, i.e. on length scales 100-500 µm.The average inter filament distance can be controlled by the line distance during 3D-printing, which sets the diameter, volume and degree of percolation of the micropore channel system.The micropore system allows the sustainable permeation with buffer solutions or culture medium potentially also for larger constructs.
Furthermore, similar as for homogeneous hydrogel constructs, a third hierarchy level on length scales from >500 µm to the size of the total construct can be realized, where multilayers, integrated microchannels, and specific construct shapes can fabricated and designed.The direct printing and thereby integration into a specially designed perfusion chamber allows whole construct permeation via the intrinsic microchannels, as well as complementary localized perfusion through the microchannel system with separate in-and outflow connections.In principle, by adjusting the flow rates of different cell, nutrient and growth factor streams in the microfluidic entrance channels A, B 1 , and B 2 during 3D printing, construct regions with specific cell types can be realized to mimic native tissue architecture.Therefore, on the level of the construct within the permeation chamber, flows of nutrients and drugs can be controlled for live cell studies.

Conclusions
The aim of the present work was the development of a concept to integrate microfluidic droplet generation, cell encapsulation, in-flow physical or chemical crosslinking, and 3D-printing in one continuous process to directly print cell-laden microgels into three-dimensional constructs in flow chambers.We showed that the two preconditions, i.e. microgel flow rate being equal to the print head speed, and the formation of printable microgel filaments, can be fulfilled with standard microfluidic and 3D-printing devices using conventional microgel materials.These are conditions to establish an automatable process, where cell-laden microgels can be continuously fabricated and printed into perfusable scaffolds.As these scaffolds are integrated into flow chambers for longterm live cell imaging, they could be used as a bioreactor, for cell testing and tissue regeneration.A further advantage of microgel scaffolds is their intrinsic microporosity, which provides a perfusion/diffusionpathway to support a sustained provision of nutrients, factors and agents to the cells in thick constructs.The integration of and connection to hollow perfusion channel inserts was demonstrated in view of the realization of vascular channel systems.The perfusion chamber allows for continuous perfusion of the whole construct via its microporosity, and for separate perfusion through perfusion channel inserts.For microfluidic channel system a double 3D-focusing design was demonstrated to minimize cell shear stress during encapsulation, to access a large range of microgel diameters, and to allow premixing with cells, crosslinking reagents, medium and factors.The method is demonstrated for photo-chemically cross-linked polyoxazolines and GelPa, as well as physically crosslinked alginate and ADA/GEL-microgels.We showed the capability to encapsulate and print microgels yeast cells, mouse embryonic fibroblasts, human fibroblasts, and human embryonic kidney cells.This opens the way for an integral and continuous automatable fabrication process from microgel production to bioprinting that can be integrated into live cell perfusion chambers.

Figure 1 .
Figure 1.Three-dimensional schematic view of the multilayer double 3D-focusing microfluidic channel system, (b) control of droplet diameter via the capillary number Ca, and accessible hydrodynamic regimes for droplet production: squeezing (c), dripping (d) and jetting (e).1-undecanol was used as oil.The scale bars are 200 µm.

Figure 3 .
Figure 3. (a) Photograph during filament printing of alginate microgels into an array of parallel lines.(b) Printed meander-shaped lines with a thickness of 300 µm.(c) Photograph of a cross-bar pattern obtained by on-top deposition of several polyoxazoline microgel filaments.The average linewidth is 1 mm.(d) Photograph of a donut-shaped polyoxazoline microgel construct.The microgels have been fluorescently labeled by FITC-dextran to demonstrate the intrinsic microporosity corresponding to the black non-fluorescent regions, (e) photograph during printing of a 3D-construct (within dashed lines) consisting of filaments arranged in a cross-bar pattern, (f) light microscopy image of a construct edge showing that fused adhesive microgels form a continuous, three-dimensional self-supporting scaffold with intrinsic micropores.1-undecanol was used as oil.

Figure 4 .
Figure 4. (a) Scheme of the perfusion chamber consisting of an upstream and downstream chamber, perfusion ports, and removable scaffolds to stabilize the microgel construct during 3D-printing, (b) photograph of a microgel construct (14 × 4 × 4 mm) in the perfusion chamber directly after printing and removal of the scaffolds, (c) confocal microscopy images of the permeation front of a fluorescent dye, where the high dye concentration in the micropores can be clearly seen.The microgels deform upon adhesion, forming a connected continuous phase, as also shown in figure S27.(d) Confocal microscopy image of YFP-labeled HEK-cells within a microgel construct.

Figure 5 .
Figure 5. (a) Layer-by-layer printing of microgel construct with integrated perfusion channel.After printing of the first layer, a hollow perfusion channel is inserted.Subsequently, the second and third layers are printed.(b) Scheme of the perfusion chamber together with the perfused construct.The perfusion chamber provides whole construct permeation via flows c in and cout, as well as independent flow through the perfusion channel via flows v in and vout.(c) Photograph of a perfusion chamber containing the construct directly after printing.The flow of the fluorescein solution through the integrated PVA hollow channel is clearly visible, and magnified in a confocal microscopy image in (d).

Figure 7 .
Figure 7. Live/dead staining of in vitro cultured NIH3T3 cells encapsulated in 10 wt% polymer microgels after extrusion through the FDM printing nozzle at day 1 (a) and 7 (b).Mineral oil was used as oil phase.Viable cells were labeled green with Calcein-AM, dead cells were labeled red with EthD-1.Scalebar: 100 µm.

Figure 8 .
Figure 8. Behavior of BJ morphology reporter cells after being encapsulated in microgels which were extruded through the printing nozzle.Viable cells are labeled green with Calcein, a labeling of dead cells with EthD-1 was not possible due to the overlap of the signal with the tdTomato signal of the reporter cells (not shown).As precursor solutions, a 9 wt% mixture of eneand thiol-functionalized poly(methyloxazoline) was mixed with 1 wt% high (a), (c)-(d) or low-molecular weight pentenoic acid modified gelatin (b).Signs of proliferation and migration are highlighted in (c)-(d) with arrows.For better visualization of the cells, bright field images (c), (e) are shown besides the corresponding stained pictures (d), (f).Scalebars: (a), (b): 100 µm; (c)-(f): 50 µm.