Aqueous Calcium Phosphate Cement Inks for 3D Printing

Mimicking the native properties and architecture of natural bone is a remaining challenge within the field of regenerative medicine. Due to the chemical similarity of calcium phosphate cements (CPCs) to bone mineral, these cements are well studied as potential bone replacement material. Nevertheless, the processing and handling of CPCs into prefabricated pastes with adequate properties for 3D printing has drawbacks due to slow reaction times, limited design freedom, as well as fabrication issues such as filter pressing during ejection through thin nozzles. Herein, an aqueous cement paste containing α‐tricalcium phosphate powder is proposed, which is stabilized by sodium pyrophosphate (Na4P2O7·10H2O) as additive. Since high powder loadings within pastes can result in filter pressing during extrusion, various concentrations and molecular weights of hyaluronic acid (HyAc) are added to the cement paste, resulting in reduced filter pressing during 3D extrusion‐based printing. These cement pastes are investigated regarding their setting reaction after activation with orthophosphate solution by isothermal calorimetry and X‐ray diffraction, as well as their hardening performance using Imeter measurements, while the processability is assessed by extrusion through 1.2 and 0.8 mm cannulas. The 3D‐printed structures with appropriate HyAc molecular weight and concentration demonstrate suitable mechanical properties and resolution for clinical application.


Introduction
Due to their excellent biocompatibility and their chemical similarity to human bone mineral, hydroxyapatite-forming calcium phosphate cements (CPCs) are well-established materials for bone repair. [1]Several commercial products are currently available for clinical usage, [2] which are based on a starting powder containing one or more soluble calcium phosphate phases.Upon combination with an aqueous solution as mixing liquid, these phases dissolve, enriching the mixing liquid with calcium (Ca 2þ ) and phosphate (PO 4 3À ) ions.The solution then becomes supersaturated with respect to less soluble calcium phosphate hydrate phases, resulting in their precipitation.Depending on the hydration product, the cements can be assigned as either apatite or brushite cements. [3]A common constituent of apatite cements is α-tricalcium phosphate (α-TCP; α-Ca 3 (PO 4 ) 2 ).Hydration of α-TCP results in the formation of a calcium-deficient hydroxyapatite (CDHA) according to Equation (1). [4] As this reaction proceeds rather slowly, methods for acceleration are required.One option is the addition of Na 2 HPO 4 , acting as accelerator due to the common ion effect. [5]Reactivity of α-TCP powder can be further increased by partial amorphization by prolonged milling, resulting in the formation of highly soluble amorphous TCP (ATCP). [6]Equal to crystalline α-TCP, this ATCP reacts under formation of CDHA according to Equation (1). [7]ydration of α-TCP can be successfully suppressed for prolonged time periods by addition of pyrophosphate (P 2 O 7 4À ) ions, which are supposed to adsorb to the surfaces of the α-TCP particles, thus blocking their dissolution. [8]This effect can be utilized for the fabrication of premixed cement pastes, with the potential to simplify the mixing procedure during surgery by using standardized static mixing devices, reducing preparation-related effects, e.g., caused by user variations. [9]Such premixed pastes can then be activated in a controlled manner by addition of an aqueous orthophosphate solution to start the setting reaction.In a recent study, it was demonstrated that the setting kinetics after activation of premixed pastes with an aqueous orthophosphate solution, containing an overall concentration of 30 wt% Na 2 HPO 4 and NaH 2 PO 4 (Na 2 /Na) in a weight ratio of 4:1, were well adjustable by varying the concentration of tetrasodium pyrophosphate decahydrate (Na 4 P 2 O 7 •10H 2 O; PP), as well as the added amount of Na 2 /Na activator solution.PP concentrations of 0.05 wt% and an addition of 21 vol% activator solution were proposed as best suitable with respect to the resulting setting performance. [8]hile small defects surrounded by intact bone are easy to fill with injectable cement pastes, the treatment of larger size defects remains a challenge in bone regeneration, as pastes are not ideal DOI: 10.1002/adem.202300789Mimicking the native properties and architecture of natural bone is a remaining challenge within the field of regenerative medicine.Due to the chemical similarity of calcium phosphate cements (CPCs) to bone mineral, these cements are well studied as potential bone replacement material.Nevertheless, the processing and handling of CPCs into prefabricated pastes with adequate properties for 3D printing has drawbacks due to slow reaction times, limited design freedom, as well as fabrication issues such as filter pressing during ejection through thin nozzles.Herein, an aqueous cement paste containing α-tricalcium phosphate powder is proposed, which is stabilized by sodium pyrophosphate (Na 4 P 2 O 7 •10H 2 O) as additive.Since high powder loadings within pastes can result in filter pressing during extrusion, various concentrations and molecular weights of hyaluronic acid (HyAc) are added to the cement paste, resulting in reduced filter pressing during 3D extrusion-based printing.These cement pastes are investigated regarding their setting reaction after activation with orthophosphate solution by isothermal calorimetry and X-ray diffraction, as well as their hardening performance using Imeter measurements, while the processability is assessed by extrusion through 1.2 and 0.8 mm cannulas.The 3D-printed structures with appropriate HyAc molecular weight and concentration demonstrate suitable mechanical properties and resolution for clinical application.due to their poor mechanical shape stability before hardening.Furthermore, the irregular shape of such defects is often difficult to rebuild while providing sufficient porosity for nutrient supply.Here, 3D printing of ceramics has gained increasing attention within the last decades due to the freedom in design of the fabricated constructs.Extrusion-based 3D printing enables the production of macroporous and mechanical stable scaffolds in patient-specific shape by using computer-aided design (CAD) model, for example, based on CT scans and/or MRI.However, the ceramic materials often require high-temperature sintering steps to provide sufficient mechanical properties resulting in shrinkage of the final 3D shape.To overcome this, reactive cement pastes composed of cement powder and an oil/surfactant mixture have been developed, whereas an exchange of the oil by water post-printing initiates cement setting. [10]While this is successful in 3D printing of microporous calcium phosphate scaffolds [11] and even patient specific implants, [12] such cement pastes are only printable at very high solid contents >80 wt% due to viscosity reasons leading to a low microporosity hydroxyapatite matrix with a limited degradation ability in vivo.
In the current work, we have chosen a different approach for 3D printing of CPC scaffolds.Here, the previously described aqueous cement paste was further modified for an application in extrusionbased 3D printing by adding hyaluronic acid (HyAc) as swelling agent to increase paste viscosity.Although CPC modification with HyAc has been already described, [13] these cements have not yet been used for extrusion-based additive manufacturing approaches.Only one previous study used oxidized HyAc as an additive to a mixture of polyvinylalcohol and CPC to modulate the angiogenic properties of printed scaffolds. [14]The modification with HyAc allows viscosity adaption by the liquid phase and hence will enable a lower ceramic content of the paste.Different molecular weights and concentrations of HyAc were applied and resulting properties such as viscosity, setting behavior, and mechanical performance were systematically investigated.Finally, printing experiments demonstrated the applicability of the paste, whereas a direct printing into the hardening solution was beneficial to prevent fusion of the processed strands and to maintain the overall shape of the printed construct.

Characterization of α-TCP Starting Powder
Using the laser-scattering particle size distribution analyzer, the α-TCP powder was investigated regarding the particle sizes and distribution indicating a bimodal particle distribution with maxima at 0.9 and 13.2 μm and an average particle size of 10.8 AE 3.4 μm.The samples were analyzed using X-ray diffraction (XRD) with Rietveld refinement and G-factor quantification regarding the phase composition of the material, resulting in 87 AE 2 wt% of crystalline α-TCP and 13 AE 2 wt% of ATCP.

Rheological Investigation of the Cement Pastes Containing HyAc
The viscosity and its dependence on the shear rate are decisive for the behavior of the cement systems before and during extrusion.Therefore, the viscosity as a function of shear rate was measured for different cement systems containing 1 wt% HyAc within the liquid phase, as shown in Figure 1a while investigating different molecular weights of HyAc.Furthermore, in Figure 1b, the influence of HyAc on the viscosity was studied from 1 to 5 wt% HyAc with a molecular weight of 2-2.5 MDa and in the following denominated as HyAc_1, HyAc_3, HyAc_4, and HyAc_5, respectively.
Figure 1a shows a clear increase in initial viscosity due to the increasing molecular weight of the HyAc used.However, the cement system containing HyAc with 1-2 MDa showed a different behavior and resulted in the highest viscosity, even greater than the cement system with 2-2.5 MDa HyAc.This might be due to variations during sample preparation, as the highly viscous pastes were difficult to mix homogenously.In the viscous cement pastes, small agglomerates can easily remain, even after mixing, which influence the measurement.Furthermore, the error bars of the samples with lower viscosity are higher compared to those with high viscosity (mean deviation, n = 3), which might be due to a decrease in stability of the lower (molecular weight) cement systems.This could be caused by the lower viscosity resulting in less stable cement pastes, which are more likely to show phase separation, leading to a cement system containing liquid parts and solid parts influencing the measurements.In general, initial viscosities of 3200 and 2450 Pa s, respectively, can be obtained for the 1 wt% HyAc formulations with molecular weights of 1-2/2-2.5 MDa. Figure 1b shows a multiplication of viscosity with an increase in the HyAc content in the liquid.With a fivefold increase (1 wt% ! 5 wt%) of the HyAc content in the fluid, the viscosity increases from about 2500 to 48 000 Pa s, corresponding to an increase of 19.2-times the initial viscosity.Due to the strong shear thinning behavior of the cement paste containing HyAc_5, the material can be used for 3D-printing application even though the viscosity increases dramatically.

Isothermal Calorimetry
All three samples investigated exhibited a pronounced initial heat flow maximum.The height of this maximum was clearly dependent on the HyAc content: it was the highest for HyAc_0 with a maximum value of 97 AE 8 mW g TCP À1 , while the maxima of the HyAc containing samples were significantly lower with 33 AE 3 mW g TCP À1 for HyAc_3, and 23 AE 2 mW g TCP À1 for HyAc_5 (Figure 2a).However, it is also evident that the heat flows of HyAc_3 and HyAc_5 started to increase earlier than those of HyAc_0.The samples also differ in the heat flow following the initial maximum: HyAc_0 and HyAc_5 showed a nearly linear decrease afterward, with the level of heat flow being significantly higher for HyAc_0 (Figure 2c).In contrast, a small, second maximum was visible after about 2 h in HyAc_3.In the later course of the reaction, the heat flows of HyAc_0 and HyAc_3 continuously decreased toward zero, while a rather broad shoulder appeared in HyAc_5 (Figure 2b).
In accordance with this, the increase in heat of hydration (HoH) was most rapid for HyAc_0 (Figure 2c).While the HoHs of both HyAc containing samples developed quite similar during the first 2 h, the HoH of HyAc_3 slightly exceeded that of HyAc_5 in the following hours.However, after around 15 h, this trend was reversed, obviously caused by the broad shoulder in HyAc_5.After hydration of 42 h, HoHs were 115 AE 6 J g TCP À1 for HyAc_0, 91 AE 5 J g TCP À1 for HyAc_3, and 111 AE 7 J g TCP

À1
for HyAc_5.Hence, the values were practically identical for HyAc_0 and HyAc_5, while they were slightly lower for HyAc_3.Though the final HoHs after 42 h slightly differed, the reactions of all three samples can be considered reproducible.Calculations of the estimated reaction rates, as described in Experimental Section, Isothermal calorimetry, resulted in values of 90% AE 6% for HyAc_0, 67% AE 4% for HyAc_3, and 86% AE 7% for HyAc_5.Hence, no significant difference was recorded between HyAc_0 and HyAc_5, while a remarkably lower reaction rate was achieved for HyAc_3.

Imeter Measurements
The Imeter hardness H i20 of HyAc_0 showed a pronounced increase during the first 0.5 h of hydration and a slower, but continuous increase afterward (Figure 3).In contrast, the hardness of HyAc_5 continuously increased during the measurement time of around 3 h, the H i20 values were always far below those of HyAc_0.Restart of the measurement of one sample of HyAc_5 resulted in H i20 values strongly exceeding those reached after around 3 h in the same sample.IST was 9.1 min (0.15 h) for HyAc_0 and around 100 min (1.7 h) for HyAc_5.Since none of the samples reached a hardness of 63 MPa mm À1 , no FST data could be obtained.The measurements of both HyAc_0 and HyAc_5 were well reproducible.

Quantitative Phase Composition after Hardening
After 7 days of hydration at 37 °C, all three samples investigated (HyAc_0, HyAc_3, and HyAc_5) were composed of CDHA and residual α-TCP, no other crystalline phases were detected.An exemplary plot is presented in Figure 4a, the patterns of the other samples are practically identical.Quantitative analysis by Rietveld refinement and G-factor method resulted in CDHA quantities ranging from 51 to 54 wt%, with no significant differences between the three samples (Figure 4b), and 4-5 wt% of crystalline α-TCP were left.Accordingly, the amorphous fraction, which included the residual water after cement setting, was around 42-44 wt%.Hence, no significant differences in quantitative phase composition were observable among the three samples with varying HyAc content.
The true crystallite size (True CS) parameters of CDHA were 8.46 AE 0.03 nm for HyAc_0, 8.1 AE 0.2 nm for HyAc_3, and 8.1 AE 0.1 nm for HyAc_5 (Figure 4c), thus slightly higher for the sample without HyAc and nearly identical for both HyAc-containing samples.Aspect ratios were around 3.1-3.2for all samples, with no significant differences related to the HyAc content.

Extrudability of the α-TCP Cement Systems
To examine whether the cement systems can be used for 3D-extrusion-printing method, the extrudability was investigated.The extrudability as a function of the HyAc content at an extrusion speed of 20 mm s À1 can be seen in Figure 5a, indicating mean values from three measurements.When the syringe plunger reached the bottom of the syringe, 100% extrudability was assumed and the material remaining in the needle was neglected.At extrudabilities below 100%, filter pressing occurred, meaning a phase separation (within the paste) into a liquid and a solid phase.Figure 5 indicates that all samples were extrudable with the 1.2 mm cannula, but the extrudability clearly decreased when using the 0.8 mm cannula.The extrudability as a function of the HyAc (2-2.5 MDa) concentration in the solution was measured at 20 mm s À1 and the results can be seen in Figure 5b.The significantly higher extrudability of the system with HyAc_1 solution (2-2.5 MDa) compared to the same composition in Figure 5a was achieved due to improved mixing by using a planetary vacuum mixer, which was necessary for the systems with >1 wt% HyAc due to the higher viscosity.
The reason for the low extrudability at HyAc_3, HyAc_4, and HyAc_5 might be the high viscosity of these cement systems causing limited extrudability.The high viscosity leads to a significantly greater resistance during extrusion, resulting in leakage of the cement paste at the connection between the tip and the syringe.The poor extrudability of the HyAc_3 and HyAc_4 systems might be further caused by the fast extrusion speed used, as these systems were 100% extrudable at a speed of 10 mm s À1 .

3D-Printing of the α-TCP Cement System
After investigating the extrudability of the different cement pastes containing HyAc, 3D printability was tested.Therefore, cubic or rectangular scaffolds with 0°-90°strand orientation were fabricated (Figure 6a).With a HyAc concentration of 1 wt%, the printed strands of the individual layers started to fuse, losing their roundish shapes and flattening onto the collector substrate 10 min after processing, as indicated in Figure 6b.Therefore, to prevent the flattening and fusion of cement paste, the fibers were directly printed into the Na 2 /Na activator solution, enabling a fast hardening of the 3D-printed cement paste and more defined strands within the construct.However, printing directly into the Na 2 /Na solution resulted in less accurate fiber placement and clogging of the nozzle tip due to hardening of the cement paste already within the nozzle tip.
Another approach to improve the form stability of the fibers after 3D printing is to adapt the HyAc content within the cement paste by increasing it up to 5 wt%.HyAc_2 did not show any changes compared to the cement system using HyAc_1.In contrast, 3D printing of HyAc_5 was still possible; however, the resulting fibers were not uniform and extrudability was only possible using the larger nozzle tip with a diameter of 1.2 mm.Therefore, using HyAc_4 resulted in accurate fiber placement within the 3D-printed construct and sufficient fiber morphology and uniformity, as shown in Figure 6c.
In Figure 6d,e, overview and magnified scanning electron microscope (SEM) images show CDHA crystals on the 3D-printed fibers after incubation in Na 2 /Na activator solution.The crystals cover the whole fiber surface (Figure 6d) showing  their roundish shape in the overview image and their platelike structure within the magnified SEM image in Figure 6e.

Compressive Strengths after Hardening
To be suitable for use as bone substitute material, the cement systems must be able to withstand mechanical stress to a certain extend.The strengths of manually fabricated samples as a function of the molecular weight of the HyAc are shown in Figure 7a, indicating that after 1-day setting time, a larger molecular weight of the HyAc had a negative effect on the stability of the cement system.The cement system with the largest HyAc had a strength of only of %2.1 MPa after 1 day, whereas the system with the smallest HyAc reached a strength of 3.6 MPa.However, after 7 days of setting, systems with larger HyAc were significantly more stable than systems with small HyAc.Particularly large are the differences for HyAc larger than 1 MDa.The peak value after 7 days is achieved by the cement system with a HyAc of 2-2.5 MDa with an achieved strength of 7.9 MPa.In comparison, the system with the 8-15 kDa HyAc only reached a strength of 5.0 MPa.
In addition to the change in molecular weight, the proportion of HyAc in the paste was also varied.The compressive strength of the cement systems with varying 2-2.5 MPa HyAc proportions is described in Figure 7b.After 1-day setting time, HyAc had a negative effect on the stability of the test specimens for all concentrations.While the HyAc free system shows a strength of 4.3 MPa after 1 day, the HyAc-containing systems range between  2.1 and 3.5 MPa.After 7 days, the HyAc-containing systems, except for HyAc_2, are above the strength of 7.7 MPa provided by the cement system without HyAc.However, only the systems HyAc_4 and HyAc_5 in the liquid are significantly higher, reaching strengths of 9.1 and 9.8 MPa.In comparison, the compressive strength of 3D-printed samples from HyAc_4 was found to be 8.6 AE 0.6 MPa, despite the microporous character of the sample.

Setting Reaction of HyAc-Modified α-TCP Cements
To interpret the setting reactions recorded by isothermal calorimetry, it should be first considered that the α-TCP starting powder contained 13 AE 2 wt% of ATCP.Hence, the hydration model established for α-TCP powders with amorphous content (ATCP) is relevant here: rapid hydration of the ATCP content, indicated by a sharp, early heat flow maximum, is followed by a slower reaction of crystalline α-TCP, visible as a comparably low, broad maximum.The reactions of ATCP and α-TCP are not clearly separated: α-TCP dissolution starts during the declining part of ATCP dissolution. [7]Based on this information, it can be concluded that the initial heat flows observed in the samples in this study most likely result from the reaction of ATCP.Accordingly, the following, continuously declining heat flow observed in HyAc_0 is proposed to result from α-TCP hydration, likely overlain by the decelerating reaction of ATCP.Reaction of the ATCP fraction present in the starting powder would result in an HoH of 33 J g TCP À1 .This amount of heat release was only reached after about 3 h in both HyAc_3 and HyAc_5.Hence, it can be concluded that in accordance to ref.
[7b], the reaction of ATCP is actually not restricted to the sharp initial maximum, but proceeds during the second maximum, here overlaid by α-TCP reaction.
In a similar cement system investigated in ref. [8], in situ XRD measurements indicated that the first, sharp maximum actually resulted from ATCP reaction, while a second maximum was produced by hydration of α-TCP.The major difference between both studies is that the two maxima were clearly separated in ref. [8], while there was a more continuous transition in this studyindicating the reaction was generally more rapid here.These differences might result from the usage of different α-TCP starting powders, varying in their grain size and hence their reactivity.Indeed, the calorimetry curve of HyAc_3 exhibited a clearly separated, second maximum.This can be interpreted as retardation of α-TCP hydration, induced by HyAc, separating its reaction from the initial reaction of ATCP.In HyAc_5, the continuously declining heat flow after the initial heat flow maximum was remarkably reduced, compared to HyAc_0.However, heat flow proceeded over a prolonged time.These observations suggest that the reaction of the crystalline α-TCP was indeed even stronger retarded than in HyAc_3.To sum it up, there are clear indications that HyAc retarded the hydration of both ATCP and α-TCP, while the extent of this effects increased with increasing HyAc concentration.However, despite this retarding effect, it was evident that the heat flow started to increase earlier in HyAc_3 and HyAc_5, compared to HyAc_0.This suggests that at the very beginning, HyAc even promotes the onset of reaction.
Comparison of the heat flow curves with the Imeter data (Figure 8) pointed out that in both HyAc_0 and HyAc_5, the initial heat flow was accompanied by a pronounced increase of the hardness H i20 .In the following, the continuously decreasing heat flow, resulting in flattening of HoH increase, went parallel to a slower, but continuous increase of H i20 data.In accordance with the isothermal calorimetry data, placing the HoH of HyAc_5 significantly below that of HyAc_0 in the time range considered here, the H i20 values of HyAc_5 were far below these of HyAc_0.Hence, both Imeter and isothermal calorimetry unambiguously demonstrated that the setting reaction was strongly retarded by HyAc.This effect is so pronounced that the degree of hydration reached at the end of the Imeter measurements was still far below that of HyAc_0 after around 3 h.
From qualitative observation of the mixed cement pastes, it was evident that the HyAc-modified samples had a rubberlike consistency, while HyAc_0 was rather liquid.This effect was even more pronounced at both higher HyAc concentration as well as higher molecular weight.Hence, it is likely that HyAc absorbed parts of the water, thus reducing the amount of water available for cement hydration, and accordingly resulting in retardation.As it was further noticed that the consistency of the HyAc-modified pastes turned more liquid after storage over 24 h at 37 °C, it is possible that the HyAc liberates parts of the water after some time, thus allowing further proceeding of the setting reaction.

3D Printability of HyAc-Modified α-TCP Cement
With the addition of HyAc to the cement paste, 3D printability into defined strands with accurate deposition was enabled, while keeping the roundish strand shape in xy-, as well as z-direction (Figure 6c-e) at a concentration of 4% HyAc (HyAc_4).Furthermore, the filter pressing effect occurring when processing the cement paste without HyAc, as well as leakage of the paste between the syringe and the nozzle tip during 3D printing was decreased and extrusion of pastes with different HyAc contents was proofed.Suitable extrudability was enabled for all the tested cement pastes containing HyAc with the 1.19 mm cannula.However, when using the smaller cannula with a diameter of 0.84 mm, filter pressing effects and leakage of the pastes at the connection points were more likely to happen, especially for the more viscous pastes containing higher HyAc contents.However, the extrudability was improved even for the higher viscous pastes by decreasing the extrusion speed.Furthermore, HyAc_4 enabled sufficient 3D printability resulting in uniform strands with good shape fidelity.Although HyAc has been shown to be susceptible to ionic cross-linking (e.g., by 1 M Ca 2þ in ref. [15]), the low calcium concentration in the cement liquid (solubility of the cement component α-TCP is %2.5 mg L À1 [16] ) is not expected to result in significant cross-linking.Indeed, we did not observe any change in rheological behavior of the pastes over a course of several days.

Compressive Strength of 3D-Printed HyAc-Modified α-TCP Cement
The influence of the HyAc concentration, as well as molecular weight on the compressive strengths of the resulting samples was studied.This indicated that an increasing molecular weight of HyAc decreased the stability of the cement system after 1-day setting time.The cement system with the largest and smallest HyAc molecular weight had strength of 2.1 and 3.6 MPa, respectively, after 1 day.However, after 7 days of setting the systems with the largest HyAc molecular weight were more stable than systems with small HyAc.A possible explanation could be the setting mechanism of hydroxyapatite from α-TCP-forming small needle-shaped crystals (or platelets in the case of CDHA).Cement hardening is then caused by interlocking of the crystals, which is continuously proceeding with setting time.Since the speed of cement setting decreases with increasing concentration of HyAc (see Section 2.3.and 2.4.), samples with larger HyAc molecules (and higher concentration) show initially slower crystal growth and hence lower mechanical stability.After longer setting periods of 7 days, the degree of cement conversion is nearly equal for all sample variations (see Section 2.5) and a reinforcement effect by the HyAc hydrogel phase is clearly visible, especially for higher molecular weight.

Phase Composition of HyAc-Modified α-TCP Cement after Hardening
Despite the remarkable retarding effect of HyAc, as observed in isothermal calorimetry and Imeter measurements, XRD evaluation of the storage samples indicated that there were no differences in final phase composition after 7 days of hardening at 37 °C, as the quantities of both residual α-TCP and precipitated, crystalline CDHA were practically identical.There were also no relevant differences in the size of the crystallites (coherent scattering domains [CSDs]) of CDHA, the information obtained from X-ray diffraction analysis.This means in turn that the crystallite (CSD) growth of CDHA was not affected by HyAc.This is in accordance with results from another study with HEMA-modified cement based on α-TCP, where only minor reduction of CDHA crystallite size was induced by addition of the polymer.However, other studies suggest that slower CDHA formation might result in a pronounced increase of CDHA crystallite sizes. [17]This was obviously not the case in the present study for the retardation induced by HyAc.In addition, typical platelike CDHA crystals completely covering the fiber surface were visible in HyAc-modified cement in SEM images.This means that also the growth of the CDHA crystals (the structures visible under an SEM, which might be composed of several CSDs) was not remarkably affected by HyAc, as the CDHA developed its typical morphology.
Quantitative XRD measurements of storage samples revealed that all samples contained a high fraction of amorphous content after 7 days of hydration at 37 °C (Figure 4b).It should be considered that also the residual water left after hydration accounts to this amorphous content.The water contents of the samples after hydration were 28 wt%, taking into account the water loss during storage.Since water uptake of CDHA is rather low (only 1.9% of its weight), most of this water should still be present as free water after hydration if no other hydration products were formed.However, amorphous fractions of 42-44 wt% were measured in the samples.Hence, it is likely that some kind of amorphous hydration product formed in addition to the crystalline CDHA, in amounts of around 15 wt% or slightly more, considering that also the amorphous phase will probably contain parts of the water not incorporated by CDHA.Though crystalline CDHA with small crystallite sizes is the main hydration product with amounts of 51-54 wt%, it should be considered that the additional amorphous phase might affect the biological performance of the set cements.Amorphous CaP phases are supposed to have a higher solubility than their crystalline counterparts; hence, faster degradation within the body would be expected. [18]ndeed, it was shown that apatite implant coatings with 40 wt% of amorphous fraction were very well resorbed in in vivo studies, even better than completely amorphous coatings. [19]Therefore, the amorphous fraction present in the samples investigated is supposed to have a positive influence on degradability.As there was no significant variation of amorphous fraction between the samples with different HyAc contents, the ratio of crystalline CDHA to the proposed amorphous calcium phosphate phase was unaffected by HyAc addition.It should be further mentioned that the HyAc contained in the samples also accounts to the amorphous fraction.However, since its content in the set cements was only 0.8 and 1.4 wt%, respectively, the observations described earlier are not influenced by this.
Differences in the reaction rates obtained in isothermal calorimetry and XRD storage samples were observed.The reaction rates in the storage samples were higher, as only around 4 wt% of crystalline α-TCP were left, while maximum reaction rates of 90 AE 6% were reached in isothermal calorimetry measurements, resulting in around 10 wt% of unreacted α-TCP.There might be two reasons for this effect: first, it should be considered that isothermal calorimetry measurements were stopped after 42 h, while the XRD storage samples were measured after 7 days.As the heat flow did not decrease to zero after ending the calorimetry measurements, it is likely that the reaction further proceeded, resulting in increased degree of hydration after 7 days.Another option is that mixing of the pastes was less effective in the calorimetry setup, resulting in reduced reactivity of the pastes.
Isothermal calorimetry results suggested a reduction of reaction rate in HyAc_3, compared to HyAc_0 and HyAc_5.However, as this was not confirmed by the XRD storage samples, it is proposed that the 3 wt% of HyAc in the sample (HyAc_3) do not generally reduce the extent of CDHA formation, but it is rather an effect related to the mixing procedure, for example, reduced mixing ability of HyAc_3 paste, compared to the others.Nevertheless, as still an estimated reaction rate of 67% AE 4% was reached in HyAc_3, the isothermal calorimetry results can be considered as reliable.

Conclusion
The pronounced retarding effect of HyAc, as indicated by isothermal calorimetry and Imeter measurements, needs to be considered in the 3D-printing procedure, specifically in the subsequent hardening in Na 2 /Na activator solution.However, the study also demonstrated that formation of CDHA over prolonged time periods is not hampered by HyAc.This confirms the feasibility of fabricating HyAc-modified CDHA-forming cements based on α-TCP.The resulting cement pastes containing HyAc enable sufficient 3D printability, as the filter pressing effect was reduced with the addition of HyAc to the cement paste.Furthermore, the cement paste HyAc_4 resulted in accurately 3D-printed constructs with uniform fibers and improved shape fidelity with a reduced extrusion speed of 10 mm s À1 .After hardening in Na 2 /Na activator solution for 7 days, compressive strengths up to 9.2 to 9.8 MPa were reached for the cement pastes HyAc_4 and HyAc_5, respectively, and typical CDHA platelike crystals covered the whole fiber surfaces.In conclusion, a cement composition with 4 wt% (HyAc_4) of 2-2.5 MDa HyAc is most suitable for 3D printing.Since the whole process chain works at ambient temperature, this opens the possibility to further modify the paste, e.g., by incorporating drugs such as antibiotics within the aqueous solution of tetrasodium pyrophosphate decahydrate (Na 4 P 2 O 7 •10H 2 O).

Experimental Section
Fabrication of α-TCP Starting Powder: 1000.0 g of CaHPO 4 (Innotere, Germany) was mixed with 341.9 g CaCO 3 (Merck, Germany) with a ploughshare mixer (M5, Lödige) for 1 h.Subsequently, the mixture was sintered in a high temperature furnace (Oyten thermotechnik system Vecstar) for 5 h at 1400 °C followed by quenching in air.The resulting α-TCP was then pulverized and ground in a planetary ball mill PM400 (Retsch, Haan, Germany) with six zirconia balls (d = 25 mm) and approximately 1 mL of isopropanol for 2.5 h at 200 rpm.
The α-TCP starting powder was characterized by powder XRD with a D8 Advance (Bruker AXS, Karlsruhe, Germany), with the following measurement parameters: range 7°-70°2θ; step size 0.0112°2θ, integration time 0.3 s; radiation: copper K α ; generator settings: 40 kV, 40 mA; divergence slit: 0.3°; and sample rotation with 30 min À1 .Measurements were performed in triplicate.Quantitative phase composition was determined by Rietveld refinement combined with the G-factor method, an external standard method enabling indirect quantification of amorphous fraction. [20]he structure ICSD# 923 (α-TCP) [21] was applied for the Rietveld method; scale factor, lattice parameters, and crystallite size (Lorentz contribution) were refined.A Chebyshev polynomial of 5th order was used for the background.A slice of the natural rock quartzite, calibrated with fully crystalline silicon powder (NIST Si Standard 640d), served as external standard.Application of the G-factor method for the investigation of α-TCP cements is described in detail in ref. [7a].Particle size was determined with a laser-scattering particle size distribution analyzer (LA-300, HORIBA) after dispersion in isopropanol.
Fabrication of Cement Pastes: The premixed pastes were composed of the α-TCP starting powder and an aqueous solution of 0.05 wt% tetrasodium pyrophosphate decahydrate (Na 4 P 2 O 7 •10H 2 O), referred to as PP solution afterward.The liquid to powder ratio (L/P) was 0.4 mL g À1 .For controlled activation, an aqueous solution of Na 2 HPO 4 and NaH 2 PO 4 in a weight ratio of 4:1 and an overall concentration of 30 wt% (labeled as Na 2 /Na) was used.All chemicals used for preparation of the solutions were obtained from Merck (Darmstadt, Germany).For activation of the cement pastes, 21 vol% of Na 2 /Na related to the water fraction in the premixed pastes were added. [8]or the addition of the HyAc to the cement system, 1-5 wt% stock solutions of the corresponding HyAc salt were prepared with 0.05% (0.002 M) sodium pyrophosphate.Six HyAcs with different molecular weights (1: 8-15 kDa, 2: 80-100 kDa, 3: 0.1-0.5 MDa, 4: 0.6-1 MDa, 5: 1-2 MDa, and 6: 2-2.5 MDa) were investigated regarding their influence on the extrudability of the cement paste.Mixing with the α-TCP cement powder was achieved by using a planetary mixer (THINKY ARV-310P, THINKY U.S.A).
For isothermal calorimetry, Imeter measurements, and preparation of storage samples for XRD measurements, HyAc with 2-2.5 MDa was used in concentrations of 0, 3, and 5 wt% related to the water content of the premixed cement pastes.Samples were denominated as HyAc_0, HyAc_3, and HyAc_5, respectively.For these measurements, the HyAc powder was added directly to the α-TCP powder.This alternative preparation method was chosen, as the highly viscous solutions with high HyAc contents would not have been workable within the setup used for isothermal calorimetry (Table 1).
Isothermal Calorimetry: Isothermal calorimetry of the samples HyAc_0, HyAc_3, and HyAc_5 was conducted at a thermal activity monitor (TAM) air isothermal calorimeter (TA Instruments) equipped with eight twin-type channels (sample and reference chamber).A temperature of 37 AE 0.02 °C was adjusted by an integrated thermostat.Cement pastes were mixed by internal stirring with InMixErs (injection and mixing device for internal paste preparation, FAU Erlangen, Mineralogy) directly in the measurement chamber to avoid any external disturbances that might affect the initial heat flow.
For preparation of the measurements, the α-TCP/HyAc starting powder was mixed with the PP solution for 1 min directly in the calorimeter crucibles using a spatula.The Na 2 /Na activator solution was inserted into syringes.Reaction was started by injecting the Na 2 /Na solution into the premixed pastes and stirring for 1 min by an external motor with a defined, constant stirring rate of 858 rpm.Measurements were performed in duplicate and evaluated with the software Microcal Origin V 2019.The heat flow curves were corrected for the calibration constant of the InMixEr tools and the time constant. [22]he total heat release (HoH) achieved at the end of the measurements was obtained by integrating the heat flow curves.Hydration enthalpies for the relevant reactions, i.e., the hydration of both ATCP and crystalline α-TCP to CDHA (see Equation (1)), were determined by Hurle et al. [7b] Based on these studies, values of ΔH R(α-TCP!CDHA) = 33 AE 2 kJ mol TCP À1 and ΔH R(ATCP!CDHA) = 78 AE 2 kJ mol TCP À1 were used.As ATCP was shown to be highly reactive, [7a] it can be reasonably assumed that it completely reacted during the initial part of hydration.Hence, the heat released by reaction of all ATCP from the starting powder was subtracted from the measured HoH.As the difference was proposed to result from α-TCP hydration, the fraction of α-TCP needed to provide this amount of heat was calculated.
Imeter Measurements: The hardening performance of HyAc_0 and HyAc_5 was measured with an IMETER (IMETER/MSB Breitwieser MessSysteme, Augsburg, Germany) using the "Auto-Gilmore-Needle" approach.The IMETER method Nr. 20 was applied, providing the H i20 data as a measure for cement hardness. [23]The initial and final setting times (IST/FST) of the cements were determined according to the definition for cements.The criterion for IST was H i20 = 3.94 MPa mm À1 and H i20 = 63.0MPa mm À1 for FST.The premixed pastes were prepared by mixing the α-TCP/HyAc powder with the PP solution for 1 min with a metal spatula.Then, the Na 2 /Na solution was added and further stirred for 1 min and transferred into a circular sample holder.The sample chamber temperature was adjusted at 37 °C.Measurements were performed in duplicate.For one sample of HyAc_5, the measurement was interrupted after 3 h and restarted after 16 h to check the hardness development over a prolonged time period.The sample was stored in humid atmosphere at 37 °C during the break to prevent desiccation of the paste.
Extrudability and Rheological Properties: To quantify the extrudability of the pastes, a HyAc solution (varying in concentration from 1 to 5 wt% and in molecular weight from 8 to 15 kDa to 2-2.5 MDa) was mixed with 3 g cement powder in a ratio of L/P = 0.4.Each sample was then labeled by their combination of concentration and molecular weight.These pastes were transferred into syringes of 12 mm diameter and fixed in a custom designed mount in the universal testing machine Z010 (Zwick/Roell, Ulm, Germany).The pastes were extruded through a 1.19 and 0.84 mm cannula via a constant protrusion of 20 mm min À1 until either the syringe was empty or a force of more than 350 N was reached.The extrudability was calculated from the residual cement paste in the syringe m residual and the initial loaded weight of the cement paste m full according to Equation (2) where m syringe is the weight of the empty syringe including the cannula.The rheological behavior of the pastes was measured at a shear rate range from 0.01 to 1000 s À1 with a Rheometer (MCR 301 TruGap Ready, Anton Paar) with the PP50 measurement head (Ø 50 mm) and a plate distance of 0.5 mm (0.7 mm for high viscous pastes).
3D Printing: The scaffolds were prepared using a 3D extrusion printer (3D Discovery, RegenHU, Switzerland) with a 0.84 mm cannula.For a smooth printing, the applied pressure was always adjusted with the fresh prepared paste, but it remained in a range of around 0.15 to 3 bar, depending on the paste composition.The 24 Â 24 mm scaffolds with 4 layers, 12 Â 12 mm scaffolds with 4 layers, and 6 Â 12 mm scaffolds with 8 layers were printed.
Mechanical Properties: To address the cements' mechanical properties, 6 Â 6 Â 12 mm (H Â W Â L) samples were prepared by mixing 0.0832 mL of the Na 2 /Na per gram of cement powder followed by a transfer of the paste into silicon molds.The samples were hardened for 1 and 7 days in 100% humidity at 37 °C.For each time point and composition, 12 samples were removed and tested under compression load until failure.In comparison to the traditionally prepared and hardened samples, also dimensionally equal samples were printed and tested.Compression tests were performed in a universal testing machine (Z010, Zwick/Roell, Germany) with a crosshead speed of 30 mm min À1 until failure.The compressive strength was calculated according to Equation (3) where F max is the force at failure and A is the area of the sample in contact with the machine XRD Characterization of Hardened Samples: Storage samples were fabricated to investigate the quantitative phase content of the cements 7 days after activation.Activated cement pastes were prepared with the same procedure as for the Imeter measurements.The freshly prepared pastes were inserted into special plastic containers with an inner diameter of 23 mm and an inner height of 3 mm.The containers were tightly sealed with parafilm to minimize water evaporation during storage.The cements were then allowed to harden in an incubator Heratherm (Thermo Fisher Scientific, Schwerte, Germany) at 37 °C for 7 days.Water loss during storage was determined by weighing the samples before and afterward.For XRD analysis of the samples, the lid was removed and the sample surface was polished using a 120-grit sandpaper to remove any possible surface effects.Samples were covered with a Kapton polyimide film (Chemplex Industries, Cat.No. 440) to reduce evaporation of residual water during XRD measurement.
The samples were analyzed at the D8 Advance used for powder measurements.The integration time was increased to 0.4 s, while the other parameters were identical to the powder measurements.Measurements were performed in duplicate.For Rietveld refinement, the structure of hydroxyapatite (HAp) with ICSD #26 204 [24] was used for CDHA.As the CDHA showed anisotropic crystallinity, i.e., an anisotropic size of the CSDs, a special ellipsoid model was applied for refinement of the CSDs. [25]Due to the constraints of the hexagonal symmetry, rx and ry were set to the same value.rx was aligned parallel to the crystallographic a-axis and rz to the crystallographic c-axis.The cube root of the model ellipsoid volume revealed the "true crystallite size" (True CS).The background contributions of the Kapton film and the residual water in the samples were each modeled by an hkl phase. [26]ble 1.Overview of cement paste preparation approaches used for the different experimental methods in the study.

Figure 1 .
Figure 1.Mean viscosity as a function of the shear rate for cement systems differing by a) the molecular weight of the hyaluronic acids (HyAc, w/c = 0.4 and 1 wt% HyAc content) and b) the proportion of HyAc (2-2.5 MDa) within the liquid phase in wt% and labeled as followed HyAc_1, HyAc_3, HyAc_4, and HyAc_5, respectively.

Figure 2 .
Figure 2. Isothermal calorimetry of samples HyAc_0, HyAc_3, and HyAc_5; a) initial heat flow; b) high resolution of prolonged measurement time; and c) overview of complete reaction, including heat of hydrations (HoHs); T = 37 °C; n = 2, all single measurements are shown.

Figure 4 .
Figure 4. a) Diffraction pattern of set cement, exemplarily shown for HyAc_0; b) quantitative phase composition of set cements, determined by G-factor, and c) true crystallite size (True CS) and aspect ratio rz/rx of calcium-deficient hydroxyapatite (CDHA); samples were hardened for 7 days at 37 °C; n = 2.

Figure 5 .
Figure 5. Extrudability of α-tricalcium phosphate (α-TCP) cement systems (w/c = 0.4) as a function of the cannula size (1.19 and 0.84 mm) measured at an extrusion speed of 20 mm s À1 for a) different molecular weights of the HyAcs (HyAc solution concentration of 1 wt%) and b) for different HyAc concentrations using the HyAc with the molecular weight of 2-2.5 MDa.

Figure 6 .
Figure 6.The 3D-printed samples using the α-TCP cement system.a) Sample size of 12 Â 6 Â 6 mm used for compression tests.b) W/c = 0.4 with HyAc_1 (1 wt%; 2-2.5 MDa) after hardening in Na 2 /Na solution.c-e) W/c = 0.4 with HyAc_4 (4 wt%; 2-2.5 MDa) after hardening for 7 days in Na 2 /Na solution.Scanning electron microscope (SEM) images showing the surface of the of the 3D-printed HyAc_4 structures after 7 days hardening with the typical platelike crystals of CDHA as overview image (d) and a magnified view (e).

Figure 7 .
Figure 7. Compressive strength of α-TCP cement systems after 1 and 7 days setting time and a w/c ratio = 0.4 and a) with 1 wt%.HyAc in the liquid phase as a function of the different molecular weights of the HyAcs in Da, and b) with different HyAc contents (wt%) in the solution.A sample series without HyAc was used as reference.

Figure 8 .
Figure 8.Comparison of isothermal calorimetry and Imeter results of HyAc_0 and HyAc_5; T = 37 °C; n = 2; the means are presented for heat flows and HoHs, all single measurements are shown for the H i20 data.