=Paper=
{{Paper
|id=Vol-3363/paper.07
|storemode=property
|title=Particle image velocimetry image processing to assess cell distribution within bioreactors
|pdfUrl=https://ceur-ws.org/Vol-3363/paper07.pdf
|volume=Vol-3363
|authors=Michele Pistillo,Margherita Scamarcio,Federica Liguori,Maurizio Mastantuono,Simona Gramazio,Alfredo Ambrico,Rosaria Alessandra Magarelli,Maria Martino,Mario Trupo,Miriam Merco,Mario Ledda,Giuseppe Falvo D’urso Labate
|dblpUrl=https://dblp.org/rec/conf/determined/PistilloSLMGAMM22
}}
==Particle image velocimetry image processing to assess cell distribution within bioreactors==
https://ceur-ws.org/Vol-3363/paper07.pdf
Particle image velocimetry image processing to assess
cell distribution within bioreactors
Michele Pistillo1,*,† , Margherita Scamarcio1,† , Federica Liguori1,† ,
Maurizio Mastantuono1,† , Simona Gramazio2,† , Alfredo Ambrico3,† ,
Rosaria Alessandra Magarelli3,† , Maria Martino3,† , Mario Trupo3,† , Miriam Merco4,† ,
Mario Ledda4,† and Giuseppe Falvo D’urso Labate1,†
1
Cellex Srl, Piazzale delle Belle Arti, 2 00196 Rome, Italy
2
AKKA Technologies Italy, Life Sciences division, via Enrico Tazzoli 215/12b 10137 Torino
3
Italian National Agency for New Technologies, Energy and Sustainable Economic Development (ENEA), Trisaia
Research Centre SS 106 Jonica, Km 419+500, 75026 Rotondella (Matera), Italy
4
Institute of Translational Pharmacology, National Research Council (CNR), Via Fosso del Cavaliere 100, 00133 Rome,
Italy
Abstract
Cell cultures suspended in bioreactors in a fluid environment are the basis for cell expansion and important
medical products manufacturing. Assessing local cell distribution within bioreactors may provide
information to increase cell production efficiency. Hydrodynamics characterizations of bioreactors are
typically performed via Particle Image Velocimetry (PIV) with fluorescent polystyrene microspheres
or Computational Fluid Dynamics (CFD), while local cell distribution is monitored through expensive
sensors or direct sampling. However, PIV and CFD analysis lack of cell behaviour representativity, while
direct sampling give average and local information and may impact cell culture conditions. In this study a
novel non-invasive method, focusing on the optical investigation of suspended fluorescent nanoparticles
(NPs) -labelled Chinese hamster ovary (CHO) cells distribution within SUSPENCE® bioreactor through
PIV image processing, is presented. Our investigation showcases the favourable effect of an innovative
NPs internalisation approach in terms of cellular uptake efficiency and fluorescence brightness. Moreover,
NPs-labelled CHO cells (NP-CHO) PIV image processing and analysis robustness is validated by cell
sampling and sample processing. Furthermore, the turbulent kinetic energy distribution to gain insight
of the impact of hydrodynamic conditions on cell culture is evaluated.
Keywords
Biomedical image processing, Particle Image Velocimetry, Bioreactors, Nanoparticles
DETERMINED 2022: Neurodevelopmental Impairments in Preterm Children — Computational Advancements,
August 26, 2022, Ljubljana, Slovenia
*
Corresponding author.
†
These authors contributed equally.
$ michelepistillo94@gmail.com (M. Pistillo); margherita.scamarcio@cellex.it (M. Scamarcio);
federica.liguori@cellex.it (F. Liguori); maurizio.mastantuono85@gmail.com (M. Mastantuono);
Simona.GRAMAZIO@akka.eu (S. Gramazio); alfredo.ambrico@enea.it (A. Ambrico); rosaria.magarelli@enea.it
(R. A. Magarelli); maria.martino@enea.it (M. Martino); mario.trupo@enea.it (M. Trupo);
Miriam.merco.15@gmail.com (M. Merco); mario.ledda@ift.cnr.it (M. Ledda); management@cellex.it (G. Falvo
D’urso Labate)
0000-0002-8487-9723 (A. Ambrico); 0000-0001-5419-62543 (R. A. Magarelli); 0000-0002-2640-050X (M. Trupo)
© 2022 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073
CEUR Workshop Proceedings (CEUR-WS.org)
88
1. Introduction
Bioreactors have become fundamental devices in the biomedical industry since they are em-
ployed in different fields such as tissue engineering and cell-based therapies [1]. The use of
bioreactor for suspension cells culture gives many potential advantages over the static cell
culture, including homogeneity of cell conditions, possibility of sampling, automated systems
for controlling parameters such as pH, temperature and dissolved oxygen concentration, thus
allowing to obtain very high cell densities [2].
Assessing cell and nutrients distribution and fluid flow field is important to evaluate bioreactors
performances. In this regard, in the current study turbulent kinetic energy (TKE), a fluid dy-
namic parameter, and cell concentration distribution have been determined. A homogeneous
distribution of all components inside bioreactors, including cells, is a key property to promote
nutrients exchange and to avoid cell deposition and local cell accumulation, thus increasing cell
growth [3]. On the other hand, TKE is a fluid parameter often employed to evaluate bioreactors
mixing mechanisms [4]. As regards to evaluate fluid flow field inside bioreactors, Particle
Image Velocimetry (PIV) is a technique used to investigate fluid parameter such as flow pattern
information, velocity field and local energy dissipation rates. PIV experimentations are usually
conducted using a fluid model composed by water and specific fluorescent tracking particles.
Also, through PIV it is possible to study the fluid dynamic characteristics inside bioreactors for
cell culture [4]. However, even though tracking particles may possess the same size and density
of living cells they not have the same biological properties, so they are unable to mimic the cell
behaviour inside cell culture chambers.
Premature new-born children strongly risk to develop health problems in childhood and ado-
lescence, early diagnostics are fundamental to prevent and ensure the welfare of the child
[5]. Different cell types can be analysed in newborns, for instance studies have shown the
role of nasal epithelial cells in neonates with asthma and allergic rhinitis and so their impact
on respiratory disease in adulthood [6]. Another example is represented by peripheral blood
mononuclear cells from umbilical cord blood of premature born babies, that can be cultured in
bioreactors for suspended cell culture and that can be employed to investigate inflammatory
processes that are a cornerstone of pathophysiology in the developing organs of preterm born
children [7][8]. As it is reported in the mentioned studies, the improvement in cell culture
through the employment of bioreactors will lead to obtain a higher number of cells and a
higher amount of biomolecules production compared to static cell culture, thus facilitating the
downstream early diagnosis processes for preterm new-borns. In the current study, CHO cells
were tested with the possibility of extending the same analysis to other cell types.
In this study a novel non-invasive method has been developed, focusing on the optical investiga-
tion of suspended fluorescent nanoparticles (NPs) -labelled Chinese hamster ovary (CHO) cells
distribution within SUSPENCE® bioreactor, through PIV image processing. PMMA-FluoRed-
COOH NPs were internalized by CHO cells and the experiments were conducted using cell
culture medium as the liquid phase. Moreover, NPs-labelled CHO cells (NP-CHO) distribution
analysis robustness was confirmed by comparing it to direct cell sampling measurements.
The current work is structured as follows. Section 2.1 describes how cell culture was performed,
both in static and in dynamic conditions. In section 2.2 cell label techniques and NPs inter-
nalisation inside cells methods are presented (results reported in Section 3.1). In section 2.3
89
the experimental setup used for PIV images acquisition of NP-CHO, including bioreactor set
up and PIV system description, are reported. Section 2.4 describes the techniques used for
PIV-images post processing: firstly the algorithm used for background noise reduction and then
the methods employed for cell and TKE distribution estimation (results reported in Section 3.2).
2. Materials and Methods
2.1. Cell culture
In recent years, NPs have gained increasing interest in various research fields, especially in the
biotechnological and biomedical fields [9]. Moreover, fluorescent NPs can be internalised by
living cells, and their uptake can be assessed via fluorescence microscopy, at the same light
wavelength emitted by tracer particles used for PIV analysis [10] [9]. Methods to obtain NPs
internalisation within CHO cells growth in adherence can be easily found in literature [11]. For
these reasons, and since the aim of this work was to study cell distribution within a bioreactor
for cell growth in suspension via PIV, we choose to use NPs to label CHO cells, firstly starting
from CHOs growth in adherence and subsequently with CHOs growth in suspension.
2.1.1. Adherence cell culture
CHO cells (American Type Culture Collection) were grown in Corning® TC-Treated Multiple
Well Plates using Ham’s F-12K (Kaighn’s) medium supplemented with 10 % heat-inactivated
fetal bovine serum (FBS, Euroclone), 2 mM L-glutamine (Sigma), 1.0 unit ml1 penicillin (Sigma),
and 1.0 mg-1 streptomycin (Sigma). The cell line was cultured on a plastic Petri dish at 37°C in
a humidified incubator containing 5% CO2.
2.1.2. Suspension cell culture
The content of one vial from the cryopreserved CHO cell bank was thawed and suspended
in 80 mL pre-warmed CD FORTI CHO medium (Thermo Fisher) supplemented with 4 mM
L-glutamine (Biowest, France) and 1% Penicillin/Streptomycin (GE Healthcare Bio-Sciences,
Sweden). The CHO suspension was grown in ventilate Erlenmeyer flask and kept in a shaker
incubator (80-85 RPM) at 37°C and with 5-7% CO2 for 96 hours.
2.2. Cell labelling and nanoparticles internalisation inside cells
2.2.1. Fluorescent nanoparticles cell internalisation
Red-fluorescent monodisperse polymethylmethacrylate carboxylated particles (microParticles
Gmbh, Germany) with a mean diameter of 286 nm (PMMA-FluoRed-COOH, SD = 7nm, abs/em
= 530/607 nm, COOH >30 µmol/g) were used to stain cells. The internalisation of NPs in CHO
cells grown in adherence was performed as follow: firstly, cells were seeded in a Corning®
TC-Treated Multiple Well Plates, where each well contained 3mL of suspended cells at a cell
concentration of 8,33 · 105 cells/mL. A 0,0025 g/mL NPs starting solution was obtained by
diluting PMMA-FluoRed-COOH stock solution at room temperature with cell culture medium.
90
0,1 mL of starting solution were added to each well to obtain a NPs concentration of 68000
NPs/cell, expressed also as 1,7 · 1012 NPs/mL and 83,3 µg/mL, the latter comparable to NPs
concentration values for cell internalisation found in literature [12]. Cells were incubated
for different time periods (1 hour, 3 hour and overnight) under growth conditions in order to
identify proper incubation periods to maximize the number of internalised NPs. Then, cell
medium containing residual NPs was discarded and cells were washed thrice with Dulbecco’s
phosphate buffered saline (DPBS). DNA staining was performed by using Hoechst dye (Thermo
Scientific™ Solution Hoechst 33342) 20 nM. Cells were harvested with 0.05% trypsin/EDTA 1 ×
and via centrifugation (980 RPM, 5 minutes) and suspended in fresh cell culture medium. Finally,
NPs-labelled cells (NPs-CHO) were examined using a fluorescence microscope (Olympus IX51,
RT Slider SPOT—Diagnostic Instruments, Sterling Heights, MI, USA), equipped with a 20 ×
objective and with a cooled CCD camera (Spot RT Slider, full frame; Diagnostic Instruments).
Moreover, the effect of fluorescence microscope exposition time (10, 100 and 1000 ms) on images
quality, intended as the easiness for the user to recognise NPs inside cells, was qualitatively
evaluated.
Regarding suspended CHO cells, a different method was developed and used to ensure NPs
internalisation. Firstly, PMMA-FluoRed-COOH NPs working solution at the same concentration
used for adherent cells internalisation (as mentioned above: 68000 NPs/cell; 1,7 · 1012 NPs/mL
NPs/mL; 83,3 µg/mL) was obtained by diluting PMMA-FluoRed-COOH NPs stock solution in
fresh cell culture medium. CHO cells were seeded and expanded in suspension cell growth
conditions, and when the required cell number was reached, they were harvested via centrifuga-
tion (700 RPM - 6 minutes) and re-suspended in PMMA-FluoRed-COOH NPs working solution,
transferred into a T175 Corning® cell culture flasks (Falcon) and incubated at different time
periods in static condition. Then, cells were washed thrice via centrifugation (700 RPM - 6
minutes) and re-suspended in fresh cell culture medium. Images of fluorescent nanoparticles
labelled – suspended CHO cells (NPs-SCHO) were collected using a fluorescence microscope
(Nikon Eclipse E400) with B-2A filter (Ex=450-490, DM=500, BA=515).
In order to compare the effect of NPs encapsulation in terms of fluorescence intensity to different
cell staining procedures, NPs-SCHO fluorescence intensity was compared to the one of CHO
cells stained with CellTracker®Orange probe (Invitrogen, UK), a largely used cell dye [13] [14].
2.2.2. CellTracker® staining
Cell staining was performed, according to manufacturer’s protocols, by using CellTracker®
Orange probe (Invitrogen, UK). 10 mM CellTracker® Orange (Invitrogen, UK) stock solution
was prepared by dissolving CellTracker® in high-quality, anhydrous dimethylsulfoxide (DMSO).
25µM staining solution was obtained diluting stock solution to CD FORTI CHO medium. The
working solution was added to suspended cells after harvesting them via centrifugation (700
RPM, 6 minutes), and cells were incubated for 45 minutes under growth conditions. Then,
the working solution was replaced with fresh media and the cells were examined using two
fluorescence microscope, depending on the laboratory availability: (i) Nikon Eclipse E400 with
B-2A filter (Ex=450-490, DM=500, BA=515) and (ii) Olympus IX51, RT Slider SPOT equipped
with a 20 × objective and with a cooled CCD camera Spot RT Slider, full frame, Diagnostic
Instruments USA).
91
2.3. Experimental set up for PIV images acquisition
2.3.1. Bioreactor set up
Suspence® (Cellex, Italy) was the bioreactor used in this work: it consists of a device in which
cell culture fluids are continuously pumped in its semi-transparent vessel from the bottom and
exit from an outlet in the upper part. Cell suspension in CD FORTI CHO medium (Thermo
Fisher) supplemented with 4 mM L-glutamine (Biowest, France) and 1% Penicillin/Streptomycin
(GE Healthcare Bio-Sciences, Sweden) was inoculated in the bioreactor with a starting volume of
1500 mL, at a starting density of 5·105 cells/mL, where the 5% of the total were NPs-SCHO cells.
Medium perfusion was guaranteed by using a peristaltic pump (MASTERFLEX® L/S® 07522-20)
setting a flow of 80 mL/min. Figure 1 illustrates the inner geometry of Suspence® vessel. Cell
samples were withdrawn at different time points after the cell seeding (30, 180 and 300 minutes)
in three different areas inside the bioreactor: the "Bottom" area, located approximately near the
bottom inner surface of the vessel, the "Centre" area, at approximately 70 mm from the bottom,
and the "Top", located near the liquid free surface.
Figure 1: 3D rendering of the inner geometry of the vessel of Suspence® bioreactor. Cells are seeded
inside the vessel and they move from the bottom part to the top.
2.3.2. PIV acquisition system
Suspence® bioreactor was positioned inside a parallelepipedal glass tank filled with distilled
water, to reduce errors caused by refractive and diffractive light phenomena on the cylindrical
surface of the bioreactor. PIV analysis were carried out by employing planar PIV system (Dantec
Dynamics, Denmark) with a green laser (Dantec Dynamics, Denmark) providing a light beam
of 532 nm and a Flow Sense USB camera (Dantec Dynamics, Denmark). The camera was
92
positioned perpendicularly to the laser beam. Figure 2 illustrates the PIV acquisition system
and the bioreactor configuration for PIV images acquisition employed during this work.
Figure 2: PIV system and the bioreactor configuration for images acquisition employed during this
work. The Suspence® bioreactor was positioned inside a glass tank filled with distilled water to reduce
refractive and diffractive light phenomena. A green laser and a camera (Dantec Dynamics, Denmark),
positioned perpendicularly to the laser beam, were also used.
Planar images for cell distribution were acquired at nine different levels of frontal plane
of the bioreactor, with the first level located at the zone in direct contact with the bottom
surface of the bioreactor and the last level covering the sections next to the free surface of
the cell suspension and the bioreactor outlet. A Cartesian coordinate system was used, with
the vertical and horizontal coordinates indicated respectively by y and x. The system origin
was located at the left end of the bioreactor bottom inner surface. The laser was oriented to
acquire the horizontal and vertical direction of cells’ speed. Figure 3 illustrates the spatial
distribution of the nine investigated sections. Images length along the y-axis was approximately
6 mm, while images width along the x-axis ranged from 45,7 to 49,1 mm, corresponding to the
laser-illuminated area dimensions. Laser beam thickness was approximately 1 mm.
93
Figure 3: Spatial distribution of the nine planar sections investigated via PIV. These sections were
analysed in order to evaluate cell distribution and flow speed along all the vessel height. Section vertical
length is approximately 6 mm, while horizontal width ranges from 45,7 to 49,1 mm. Dimension on the
right are expressed in millimeters.
2.4. PIV images post processing
2.4.1. PIV images background noise removal
A Dantec Dynamic PIV system was used to acquire the images in the vertical plane; nine sections,
as shown in Figure 3, were recorded and at different time points. For each condition 100 images,
subsequentially used for TKE distribution estimation, were taken and imported into Python to
estimate the percentage of cells in suspension. Two images for each section were analysed at a
time. Background noise removal is fundamental for biomedical image processing [15]. During
the PIV images post processing the background noise, caused by the presence of refractive
and diffractive light phenomena, was reduced by using a Python script. In our application the
MedianBlur value, obtained computing the median of all pixels under the kernel window and
the central pixel, represented the background noise, since noise corresponded with a diffuse
background halo. Different kernel size, consisting in the product between the filter mask width
and height in pixels, were used (7x7, 11x11, 15x15, 21x21) [16]. Thus, the background value was
subtracted from the starting images. The subtraction was performed because the significant
content of PIV images was assumed to be the single pixel value, that represented the absence
94
or the presence of single or aggregate cells. The threshold binary function was also used to
separate even more the objects (NPs-SCHO) considered as a foreground from its background.
In brief, the images were modified by this function so that all pixel intensity values higher than
the threshold were assigned the maximum value (white), or the minimum value (black). More
specifically, when pixel values were greater or equal than the set threshold value (50 or 100),
they were set to 255, in the other case they were set to 0 (black) [17]. Basic Python libraries
have been used, such as “cv2” to read images, “numpy” for scientific computing and “matplotlib”
to plot images.
2.4.2. Cell distribution estimation
For each of the nine sections, the number of nonzero pixels of the two images was determined,
and the result of the mean between these numbers gave an estimation of the amount of cells
in suspension. PIV image processing and analysis robustness was assessed by comparing it to
cell distribution analysis carried out via cell sampling and sample processing. Since three cell
sampling points were used, three mean areas close to the point where cells were withdrawn
were investigated. These three areas, which account for cell distribution in the bottom, centre
and upper part of the bioreactor, were named “Bottom”, “Centre” and “Top” and represented
respectively the mean between values measured in section (1, 2, 3), (4, 5) and (6, 7, 8, 9). For
each area, values were normalised to the total value, equal to the sum of the total values of
the tree areas, calculated at time 0 (cell seeding into the bioreactor). The results were shown
as bar graphs and compared with values obtained from the 3 cell samples. Total cell counts
were determined by hemocytometer, as Bürker chamber, using a fluorescence microscope. The
cell samples and the images acquired by the PIV system and post-processed with the above-
mentioned Python Script were taken at different time points after bioreactor seeding (30, 180
and 300 minutes).
2.4.3. Ensemble-averaged turbulent kinetic energy distribution analysis
TKE is an index that can be employed to understand mass transfer and cell viability within
bioreactors [18] [19] [20]. TKE distribution was obtained by re-adapting a previous PIV data
processing method developed by Odeleye et. al. [10]. Measurements were carried out with
a bioreactor fill volume of 1500 mL and at a flow rate of 80 mL/min. For each frontal plane
section of the bioreactor, 100 image pairs were collected to obtain 100 instantaneous NPs-SCHOs
velocity components (along the x and y axis) vector maps. The two components along the x
and y axis of NPs-SCHOs velocity from the PIV measurements were calculated employing the
adaptive PIV processing algorithm provided within Dynamic Studio software (Dantec Dynamics,
Denmark). Interrogation areas were set from 16x16 to 32x32 pixels, while a universal outlier
detection in neighbourhoods of 5x5 pixels was used for validating vectors. For velocity gradient
adaptivity, the absolute value of each component of the velocity gradient was limited to 0.1 while
the total magnitude of the gradients (square root of the sum of the squares) was limited to 0.2.
Spatial resolution obtained within vector maps ranged between 0.4 x 0.4 and 0.48 x 0.48 mm. The
mean value between vector maps was calculated in order to obtain vector maps representing the
“Bottom”, “Centre” and “Top” areas respectively between velocity values calculated in section (1,
95
2 ,3), (4, 5) and (6, 7 ,8, 9). These were post-processed by using a second Python script to obtain
the ensemble-averaged turbulent kinetic energy through equations presented by Odeleye et.
al. [10]. Since in the latter work kinetic energy was divided to the square of spinner tip speed
and Suspence® bioreactor does not have an impeller, in this work turbulent kinetic energy has
been divided to the square of the input fluid speed. For simplicity, turbulent kinetic energy
distribution was normalised to the maximum value found in all the investigated frontal plane
levels.
Figure 4 illustrates the steps, from the acquisition one to those executed in the post processing
phase, performed in this work.
Figure 4: Schematics of PIV images acquisition and post processing steps performed in this work.
3. Results and discussion
3.1. Fluorescent cell labelling
PMMA-FluoRed-COOH NPs uptake in CHO was monitored via fluorescence microscope (Figure
5). Images show that, at each investigated incubation time, NPs appear to be localised around cell
nuclei, and this may reveal that NPs were able to accumulate within cells. Also, Figure 5 shows
that the number of red spots around cell nuclei tend to increase over incubation time, to such
an extent that after incubating cells overnight, red fluorescence light appears to be predominant
(Figure 5c). In this case, NPs can be considered internalised since some nuclei appear to be
purple due to light overlap. For these reasons, NPs incubation time was subsequently set at 10
hours. These findings may suggest that the internalised NPs number increase over incubation
time. This result is in line with findings shown by dos Santos et. al. [12]. Further studies are
needed to fully understand the internalisation mechanisms.
Figure 6 shows PMMA-FluoRed-COOH NPs uptake in CHO cells at three different exposure
times. As can be recognised from these images, an exposure time of about 100 ms was considered
suitable to assess NPs internalisation.
96
Figure 5: Internalisation evaluation of PMMA-FluoRed-COOH NPs inside CHO cells (NPs-CHO) via
fluorescence microscope at different incubation times under cell culture conditions: (A) 1 hour, (B) 3
hours and (C) overnight. Cells nuclei were stained in blue while NPs are visualised in red. Exposition
time: 1000ms. Scale bar = 100 𝜇𝑚.
Figure 6: Internalisation of PMMA-FluoRed-COOH NPs inside CHO cells (NPs-CHO) evaluation
performed via fluorescence microscope at different exposition times: (A) 10 ms, (B) 100 ms and (C) 1s.
Cell nuclei were stained in blue while NPs are visualized in red. Scale bar = 100 µm.
Figure 7A illustrates CHO cells stained with cell tracker CellTracker®. In this case, few cells,
shown as red spots, can be distinguished (e.g in the upper-right corner of the image). Figure 7B
shows PMMA-FluoRed-COOH NPs uptake in CHO cells. In order to compare the results, the
incubation time for NPs encapsulation was set to 1 hour (similar to the incubation time set for
CellTraker dye, 45 min). From these images the difference in the obtained fluorescence intensity
between the two used methods is appreciable, with the NPs-CHO cells fluorescence being more
intense than CellTracker® one. Further quantitative fluorescence intensity analysis are needed
to confirm this result.
97
Figure 7: Comparison of the fluorescence intensity obtained via CellTracker® cell dye staining and via
NPs internalisation. (A) CHO cells stained with CellTracker® (Incubation Time: 45 minutes), cells are
visualized as red spots and (B) PMMA-FluoRed-COOH NPs inside CHO cells (NPs-CHO), cells nuclei
were stained in blue while NPs are visualized in red (Incubation time: 1 hour). Images were acquired via
fluorescence microscope. Exposition time: 100 ms. Scale bar = 100 µm.
3.1.1. Fluorescent nanoparticles internalisation in cells growth in suspension
The effectiveness of the new method developed to ensure NPs internalisation within CHO cells
growth in suspension was demonstrated by comparing fluorescence intensity of CHO cells
growth in adhesion stained with CellTracker® (Figure 8A) to the one emitted by NPs-SCHO
(Figure 8B). These images provide evidence that cells can be clearly distinguished with both
methods. Also, fluorescence intensity of NPs-SCHO appears to be comparable to CellTracker®
one. Further quantitative fluorescence intensity analysis is needed to confirm this result.
Figure 8: Evaluation of the NPs internalisation within CHO cells growth in suspension in terms of
fluorescence intensity. Fluorescence microscope images: (A) CHO cells growth in adhesion stained with
CellTracker®: cells are visualized in green and (B) PMMA-FluoRed-COOH NPs inside CHO cells growth
in suspension (NPs-SCHO), NPs are visualized in green too. Incubation time: overnight. Exposition time:
69 ms. Scale bar = 50 µm.
98
3.2. PIV images post processing
3.2.1. PIV images background noise removal
Section three of one of the nine sections acquired using the PIV system was used to evaluate the
performance of the developed Python filter and is shown in Figure 9. In this case, the kernel size
was set to 21, while the value of the Threshold Binary function was set to 50. Figure 9A, 9B and
9C represent respectively the non-filtered image, the image containing only the background
noise, and the filtered image obtained via the employed Python script processing. As can be
seen from these figures, the reduction of noise is clearly distinguished on the filtered image
(Figure 9C) single fluorescent dot, that can be associated with single cells or cell aggregates, can
be distinguished.
Figure 9: Effect of the background noise removal filter on PIV images of NPs-SCHO within SUSPENCE®.
(A) Non filtered PIV image (B) PIV image containing only the background noise and (C) Filtered PIV
image using Python script. Analysed section number: 3. Scale bar = 5mm.
3.2.2. Cell distribution estimation
Figure 10 shows cell distribution results obtained via PIV images post processing and via
cell sampling. The results of the two different used methods may be considered comparable,
especially in the first two time points considered. In addition, a homogeneous distribution
of the cell distribution inside the vessel is highlighted, indeed values found in different areas
(Bottom, Centre and Top) appear to be similar. After 30 minutes from seeding the percentage
of the cell concentration is about 33%, instead after 180 minutes about 34%, in both systems
described. This highlights how cell suspension is maintained over time. Anyway, a difference
between the two methods can be seen in the last time point considered (300 minutes). In both
systems the percentage of cell concentration in the Top area is about 40%. While the Bottom
area is about 28% and 32% and the Centre area is about 33% and 27%, in the PIV system and cell
samples respectively. This difference may be due to human errors occurring during PIV image
acquisition procedures. In the latter case, the measurements obtained from the cell samples
are considered more truthful, showing a non homogeneity of the cell concentration inside the
vessel.
99
Figure 10: Bar graphs showing cell distribution results obtained via PIV images acquisition and Python
script postprocessing (PIV) compared to the one obtained via cell sampling and cell counting (Sampling),
at different time points.
3.2.3. Ensemble-averaged turbulent kinetic energy distribution analysis
Figure 11 shows the ensemble averaged turbulent kinetic energy (TKE) distribution of NPs-
SCHO in the tree investigated bioreactors areas. A high normalized TKE area can be found in
the central region of the Bottom area, corresponding to the region close to the bioreactor inlet.
In this area the value of TKE appears to gradually decrease when moving away from the central
zone, and this may suggest that the considered area is well mixed, since TKE is considered the
portion of kinetic energy that provides a mixing mechanism due to turbulent dispersion [4].
Mixing is fundamental to obtain homogeneity of nutrients and oxygen and to reduce gradients
induced by addition of cell culture media and acid/base tirants and to increase mass transfer.
However, an increase in turbulent energy may be related to an increase in hydrodynamic stresses
that can hinder cell viability [18][19][20]. Another high normalized TKE area is found along
the length of the top area, corresponding to the region close to the bioreactor outlet. Differently
from the TKE distribution of the bottom area, TKE appears to be high also in the peripheral
zones. Moreover, TKE distribution in the Centre zone appears to be similar to the one observed
in the bottom zone but with a lower intensity. TKE intensity in this area appears to be higher
than the one observed by Odeleye et. al.[10], and this can be associated to a higher mixing
mechanism in the central area of the bioreactor and also to the presence of an inlet and an outlet
in the Suspence® bioreactor, which are absent in the bioreactor tested in the above-mentioned
study.
100
Figure 11: Ensemble-averaged turbulent kinetic energy (TKE) distribution of NPs-SCHO normalized
to the maximum TKE value found the investigated bioreactors areas (Top, Centre, Bottom). TKE is
considered the portion of kinetic energy that provides a mixing mechanism, fundamental to obtain
homogeneity of nutrients and oxygen. High normalised TKE regions can be seen near the inlet and the
outlet of the bioreactor, where flow speed is higher than the rest of the considered volume.
4. Conclusions
In this work we have demonstrated the effectiveness of the new method developed to ensure
NPs internalisation within CHO cells growth in suspension and we have exploited NPs-SCHO
to study cell and kinetic energy distribution within SUSPENCE® bioreactor via PIV analysis and
data processing. Fluorescent particles are typically employed for fluid dynamic characterization
of bioreactors and are monitored in different conditions and under various stimuli (e.g. chemico-
physical). The ability to internalise fluorescent nanoparticles inside CHO cells allowed the use
of the PIV system as a non-invasive image acquisition tool for bioreactors for cell culture in
suspension. To the extent of our knowledge, in this work for the first time living cells with
encapsuled fluorescent NPs have been employed, instead of tracer particles alone, for PIV
analysis. The quantification of cell distribution and hydrodynamic parameters, in the same
condition in which the cells are used to grow, within a bioreactor using PIV may allow a complete
and strict study upon local cell culture conditions. Regarding PIV images post processing, firstly
images background noise was successfully filtered by using our Python Script. Then, a second
script was used to calculate cell distribution within the bioreactors. Cell distribution values
appeared to be comparable, especially in the first two time points investigated, to the one
obtained via cell sampling. The ensemble averaged turbulent kinetic energy (TKE) distribution
101
of NPs-SCHO was obtained in the three investigated bioreactors areas. High normalised TKE
regions were found near the inlet and the outlet of the bioreactor, where flow speed is higher
than the rest of the considered volume. Also, in the central area of Suspence® TKE intensity
appeared to be higher than the one observed in the same area of other bioreactors. In these
regions, mixing of nutrients and oxygen is promoted, but also high levels of TKE may hinder cell
viability. Future development would be employing PIV cell distribution and TKE distribution
analysis on different types of cells to predict cell distribution and behaviour within bioreactors
during longer-lasting cell culture. In conclusion, PIV cell distribution and TKE distribution
analysis inside bioreactors may represent two new methods to optimize culture conditions
for cells (e.g. Peripheral Blood Mononuclear Cells) used for preterm new-borns diagnosis and
therapy.
References
[1] M. Stephenson, W. Grayson, Recent advances in bioreactors for cell-based therapies,
F1000Research 7 (2018).
[2] K. S. Carswell, E. T. Papoutsakis, Culture of human t cells in stirred bioreactors for cellular
immunotherapy applications: Shear, proliferation, and the il-2 receptor, Biotechnology
and bioengineering 68 (2000) 328–338.
[3] R. Olmer, A. Lange, S. Selzer, C. Kasper, A. Haverich, U. Martin, R. Zweigerdt, Suspen-
sion culture of human pluripotent stem cells in controlled, stirred bioreactors, Tissue
Engineering Part C: Methods 18 (2012) 772–784.
[4] A. Håkansson, D. Arlov, F. Carlsson, F. Innings, Hydrodynamic difference between inline
and batch operation of a rotor-stator mixer head-a cfd approach, The Canadian Journal of
Chemical Engineering 95 (2017) 806–816.
[5] A. R. Hayward, Development of lymphocyte responses and interactions in the human
fetus and newborn, Immunological reviews 57 (1981) 39–60.
[6] D. Miller, S. W. Turner, D. Spiteri-Cornish, N. McInnes, A. Scaife, P. J. Danielian, G. De-
vereux, G. M. Walsh, Culture of airway epithelial cells from neonates sampled within
48-hours of birth, PLoS One 8 (2013) e78321.
[7] J. Baier, A. C. Gwellem, R. Haase, I. Volkmer, B. Bartling, M. S. Staege, Co-culture of
peripheral blood mononuclear cells and endothelial colony forming cells from cord blood
of preterm born babies, in: In Vitro Models for Stem Cell Therapy, Springer, 2021, pp.
107–124.
[8] P. C. Collins, L. K. Nielsen, S. D. Patel, E. T. Papoutsakis, W. M. Miller, Characterization of
hematopoietic cell expansion, oxygen uptake, and glycolysis in a controlled, stirred-tank
bioreactor system, Biotechnology progress 14 (1998) 466–472.
[9] A. Bhattacharyya, P. Datta, P. Chaudhuri, B. Barik, Nanotechnology-a new frontier for
food security in socio economic development, in: Disaster risk vulnerablity conference,
2011, pp. 116–120.
[10] A. Odeleye, D. Marsh, M. Osborne, G. Lye, M. Micheletti, On the fluid dynamics of a
laboratory scale single-use stirred bioreactor, Chemical Engineering Science 111 (2014)
299–312.
102
[11] V. B. Bregar, J. Lojk, V. Šuštar, P. Veranič, M. Pavlin, Visualization of internalization of
functionalized cobalt ferrite nanoparticles and their intracellular fate, International journal
of nanomedicine 8 (2013) 919.
[12] T. Dos Santos, J. Varela, I. Lynch, A. Salvati, K. A. Dawson, Quantitative assessment of
the comparative nanoparticle-uptake efficiency of a range of cell lines, Small 7 (2011)
3341–3349.
[13] O. Uckermann, I. Iandiev, M. Francke, K. Franze, J. Grosche, S. Wolf, L. Kohen, P. Wiede-
mann, A. Reichenbach, A. Bringmann, Selective staining by vital dyes of müller glial cells
in retinal wholemounts, Glia 45 (2004) 59–66.
[14] M. Hsu, T. Andl, G. Li, J. L. Meinkoth, M. Herlyn, Cadherin repertoire determines partner-
specific gap junctional communication during melanoma progression, Journal of cell
science 113 (2000) 1535–1542.
[15] S. R. Sternberg, Biomedical image processing, Computer 16 (1983) 22–34.
[16] B. Chen, J. Liang, N. Zheng, J. C. Príncipe, Kernel least mean square with adaptive kernel
size, Neurocomputing 191 (2016) 95–106.
[17] S. Parveen, J. Shah, A motion detection system in python and opencv, in: 2021 third
international conference on intelligent communication technologies and virtual mobile
networks (ICICV), IEEE, 2021, pp. 1378–1382.
[18] S. Mishra, V. Kumar, J. Sarkar, A. S. Rathore, Cfd based mass transfer modeling of a
single use bioreactor for production of monoclonal antibody biotherapeutics, Chemical
Engineering Journal 412 (2021) 128592.
[19] A. Nienow, Hydrodynamics of stirred bioreactors (1998).
[20] A. R. Lara, E. Galindo, O. T. Ramírez, L. A. Palomares, Living with heterogeneities in
bioreactors, Molecular biotechnology 34 (2006) 355–381.
103