Online Supplementary Data & Information
a) Automation Concept
a.1) Automation & Process control system
A schematic draft of the automation concept is depicted in figure S1
a.2) Programming of a complete process control system in LabVIEW®
LabVIEW® is a widely used extensive and powerful data-processing and
analyzing tool. So we were able to expand it to a complete PCS. By using the
mentioned hardware in 2.2 we created a chip-based sensor, actor and device
level (SADL) that enabled us to program an ad hoc code for the creation and
analyzing of data-strings for commands and responds to and from the SADL.
That way, we could use LabVIEW as one central processing and analyzing
software for all of our process data. Further we could expand it to a complete
PCS that integrates all components of performance into one single Code.
Data acquisition and the organization of the timing were performed by using
the data flow principle instead of programming via master-slave principle set
the base for a robust PCS. Partly pre-programmed structures for visualization,
parameterizing and data filing could be configured and complemented. This
way, an adequate Human Machine Interface and a memory saving data
storage function could be implemented. Finally we could implement sundry
feedback control algorithms, freely adapted to the single requirements of the
process and incorporate simulation models in parallel, when it seemed to be
usefully.
b) Automation & Control
b.1) pH and pO2 under hypoxic cultivation
To control pO2 and the pH values in a highly variable range we implemented a
combined but independent control by gas mixing. To stabilize the quality of
control we extended the controller with simple computer models, which predict
the interferences of the pO2 control to the pH value and vice versa each into a
circuit similar to the Smith Predictor. So the controller is able to overcome the
destabilizing impact of dead-time delays. Moreover, the effect of usually
neglected pre-filling of gas pipes between the mixing unit and the reactor
volume that exists respectively to the particular opposite process variable of
the pO2 or pH value could be predicted and incorporated into the controller.
Accordingly, we were able to parameterize the proportional–integral–derivative
(PID) Algorithm of the controller aggressively and avoid long times of
cultivating under suboptimal physiological conditions.
Figures S2 – S4 shows the courses of the controlled and correcting variables
for the combined pH and pO2 value with and without the influence of the
controller extension by internal model control (IMC). Here, the example of
cultivating under hypoxia with the setpoints pO2w = 150 mbar and
pHw value = 7.4 is given. Without the impact of the IMC extension, there was a
strong oscillation of the pO2 value, while the pH value was prioritized. After
engaging of the IMC extension, the course of the pO2 value completely
changed its characteristic and got pushed into the range of tolerance, which
held continuously true for almost the entire experiment. Finally, we could even
observe the typical standard course for Smith Predictor control
b.2) Determination of pressure-volume relation
To determine the slope of the pressure to volume relation of the hearts to be
processed as well as to check the impact of the decellularized matrix, we
collected the data for each heart. One exemplary relation is shown in
figure S5. The relation shows a saturation curve that differs from those known
by one-dimensional tensile testing. Most likely, this may be related to
stretching of the superior 3D organization of the matrix.
b.3) Construct Perfusion
The maintenance of the myocardial interstitial vicinity is done by diffusion,
comparable to the situation in the native myocardium. Here, the supply with
nutrition and oxygen depends mainly on concentration, gradients as well as on
pressure-gradients. Because the dynamics of the real control paths in our real
system concerning the variables pO2 and other concentrations and gradients
inside of the construct, strongly differs to those in native myocardium e.g. by
the cell density and their metabolic activity, especially at early process times
we do believe, that a directly recreation of in vivo perfusion pressure profiles
cannot be translated to mimic nature like growth conditions. So we chose that
the perfusion pressure should be able to follow a setpoint profile, beginning
with a constant setpoint, to support the most physiological pattern of supply.
We successfully controlled the perfusion pressure at a setpoint of 70 mmHg.
Resulting construct perfusion could be observed by typical belling of the
construct. Caused by feeding with a volume flow, the coronary system reacted
to the measured pressure as a proportionally and time-delayed system. Here,
we implemented a simple proportional-integral (PI) control algorithm. Figure
S6 shows the combined standard courses of the stimulation and coronary
perfusion pressure together with courses of the correcting variables. The
pressure control algorithm worked properly with a very good control quality, as
the process variables stay strictly inside of the area of tolerance
c) Control of pressure amplitude
To control the amplitude of pressure for the mechanical stimulation of the
ventricle, we implemented a simple two-step control algorithm. Pressures to
volume relations were determined for each heart prior to turning on the
controller in order to obtain the individual characteristics of each heart at
predefined operation points (figure S5). Prior to each process, we had to
adjust the controller for the mechanical stimulation further according to the
stimulation frequency and the desired degree of stretching. The amplitude of
the periodic pressure signal serves as control variable, so we had to observe
the pressure signal adequately. Therefore we used a simple Fourier
transformation together with a 10-fold sampling rate respectively to the
stimulation frequency, directly programmed into the PCS. While pulsating
volume into and out of the pressure pipes, a loss of pressure medium occurs
at the unsterile side of the stimulation device. This appears as the main
disturbance quantity to the controlled system. It could be balanced adequately
by the same system like the steering of 3D stretching. As it can be seen in
figure S6, the amplitude of pressure was pushed into the area of tolerance
rapidly and without any overshooting. The pressure control algorithm worked
properly with a very good control quality, as the process variables stay strictly
inside of the area of tolerance. A screenshot of the software while culturing a
heart under stimulation has already been demonstrated in figure 2 subitem 8
and figures S2.
d) 3D inspection of the constructs
We performed a couple of z-stacks with the confocal microscope to gain
insight into the three-dimensional distribution of the orientation, alignment and
arrangement of the cells inside of the hearts ECM. The figure S7 contrasts
exemplary 3D representations for mechanically stimulated and non stimulated
hearts. They reflect the benefit of the 3D mechanical stimulation to the
orientation and alignment of cultivated cells inside of the hearts ECM and
depict it in a 3D graphical impression. The movie M1 shows a slowly spinning
z-stack representation of a highly oriented cell-bundle inside of the ECM of a
mechanically stimulated heart.
e) Image processing
We used the image J (NIH, USA) software to detect and analyze the nuclei,
that we illustrated exemplary in figure 6. For that reason we used the “make
binary”, “outline” and finally “analyze particles” functions. The corresponding
processed images are depicted in figure S8. We used the raised data to
approximate the relative area of the suspected smallest and biggest nuclei that
we could manually identify from the original pictures (fig. 6) and applied this as
the upper and the lower cut-off before further analyzing of the data as
described in the main text.
f) Details of varying cellular alignment
With respect to the described quantification of cellular alignment we illustrate
observed variances by the combined histograms of the circularity and the
orientation angle of detected nuclei. Figure S9 reflects the determined
histograms and contrasts varying distributions for stimulated vs. non
stimulated constructs side by side.
Figure Legends
Fig. S1: Draft of the composition of automation. 1) Computer with LabVIEW®
Software. 2) Serial Network as one part of the SADL containing analog to digital and
vice versa converters. 3) USB based Data Acquisition for very fast data input. 4)
Device for 3D Laser Scanning connected via a serial port as well.
Fig. S2: Screenshot of the PCS, showing the human machine interface (HMI) for the
pO2 and pH value control. The flow charts indicate the courses of the pH value
(upper) and the pO2 (lower). Left to the charts: 3 bars, representing O2, N2 and CO2
gas-flow, indicate the composition of the actual gassing.
Fig. S3: Courses of the process and actuating variables of the pO2 and pH control
according to the settings depicted in figure S2. One can see the strong oscillation of
the pO2, what is observable but not as strong for the pH value. The normal gas flows
Fni for O2, CO2 and N2 show how the variables set each other to oscillation.
Fig. S4: Courses of the process and actuating variables of the pO2 and pH control
according to the settings depicted in figure S2 but with additional engaging the
model- predictive circuit. The course of the process variables present themselves
clearly smoothed and stay inside of the tolerance band (dashed lines) almost
completely.
Fig. S5: Diagram of the resulting measured static pressure in bar inside of the
pressure-pipes and the ventricle-balloon according to volume added to the native
volume of the ventricle (+V) in µl. The striated area indicates the static pressure
inside of the ventricle, that feeds any stretching of the matrix and represents the
equality to each resulting tension of the walls.
Fig. S6: Diagram showing the courses of the process variables perfusion pressure
pPerf. and stimulation pressure amplitude pAmpl. Together with those of the actuation
variables pump speed (NPump) and added or subtracted Volume Increments of the
ventricle balloon. After manual setting of the intrinsic pressure, there occurs a loss of
pressure until the level of the non-stretched balloon is reached again. After the 2-step
control algorithm is active, the set value can be preserved with a good quality of
control. The perfusion pressure swings into the band of tolerance rapidly with only a
small overshot. Spikes in the course indicate occasional agglutination of the pipe
inside of the pump-head.
Fig. S7: Representation of confocal recorded Z-stacks of the cultivated constructs
LVWs. Nuclei and cytoplasmatic nucleic acids are presented in green by acredin
orange staining, while the cytoskeleton is presented in red by F-Actin staining.. A)
Exemplary representation for mechanically stimulated hearts. B) Exemplary
representation for non stimulated hearts.
Fig. S8: Pictures of fig.6 processed via image J software. The figure displays the
processed pictures of fig. 6, which represent the prearrangement for analyzing and
quantification of the cellular alignment as described in the main text. The single
pictures are labeled according to the corresponding pictures of fig.6. I A), III A) & V
A) non stimulated; II A), IV A) & VI A) stimulated.
Fig. S9: Combined histograms of the determined circularity and orientation angle for
the analyzed pictures of fig.6. The combined histograms are depicted as 3D graphs.
Bars represent the counted nuclei accordant to the represented coordinates
(circularity; orientation angle). Each diagram is labeled according to the analyzed
picture of fig.6. I A), III A) & V A) non stimulated; II A), IV A) & VI A) stimulated.
M1: Slowly spinning z-stack representation of a highly oriented cell-bundle inside of
the ECM of a mechanically stimulated heart. Scale marks are depicted at the border
lines of the stack.