Cell speed as phenotypic signature in drug discovery
arez Korsnes1,2& Reinert Korsnes2
1Norwegian University of Sience and Technolgy (NTNU), Department of Biotechnology and Food Science, NO-7491
2Korsnes Biocomputing (KoBio), Trondheim, Norway
Single-cell tracking throughout several cell cycles allows to trace
kinships of cells in lineage trees and ﬁnd correlations among pheno-
types. It allows to utilize the fact that related cells bear information on
the underlying mechanisms behind single cell phenotypes. Combined
or contextual analyses can therefore help to extract more information
from noisy data on cells as compared to independent analyses for each
cell. Cell speed is so far poorly analysed, however, it gives information
on inherent properties of cells since they move with various speed and
sister cells tend to move similarly. Cell speed therefore deserves to be
called a phenotype. The present results are produced using the software
KoBio Celltrack (https://korsnesbiocomputing.no/). It is
under active development as a robust and lightweight software to visu-
alize and track cells from label-free long term recordings produced by
various instruments. The purpose is to provide data of direct biological
interest as well as ground-truth for BIG data analyses.
Single and collective cell motility modes have been conserved
through evolution and they depend on active reorganization of
the cytoskeleton . Cell motility plays a crucial role in many
vital processes as well as in cancer invasion and metastasis. A
metastatasic cascade can enable some initial cells to migrate and
create their own migration tracks .
Quantitative analysis of migration tracks can help to discover
biological functions or processes involved in diseases. This
can be useful in evaluation of drug treatment, detection of rare
sub-populations and discovery of drug-tolerant persister states.
Two-dimensional (2D) in vitro experiments including single-cell
tracking can provide data on individual cell motility behaviour
under drug treatment. Data from such experiments often show
individual variation in motility in the same clonal population.
The following factors can affect cell motility:
•States/conditions of the cytoskeleton.
•Interplay between cell adhesion and contractility.
•Deformability of the nucleus.
•Plasticity of migration mechanisms.
•Intracellular cascade of signalling events.
Cell speed variation in clonal populations allows to correlate
speed of sister cells. Such covariation can indicate that cell
speed is an inherent property.
Aim of study
The aim of this study is to explore cell speed to obtain biolo-
gical relevant information from cells and their response to drug
exposure. Cell speed estimates are one of the easiest available
parameters to derive from single-cell tracking. Speed is there-
fore a natural candidate to explore cell inherent properties. Un-
covering the utility of speed data will therefore presumably have
a signiﬁcant impact on phenotypic screening for drug discovery.
Materials and Methods
A549 cell lines were cultured in RPMI 1640 (Lonza, Nor-
way), supplemented with 9 % heat inactivated fetal calf serum
(FCS, Bionordika, Norway), 0.02 Hepes buffer 1M in 0.85 %
NaCl (Cambrex no 0750, #BE17-737G) and 10 ml 1X Glutamax
(100X, Gibco #35050-038), 5 ml in 500 ml medium. Cells were
maintained at 37 ◦Cin a humidiﬁed 5 % CO2atmosphere.
A549 cells were plated onto 96 multiwell black microplates
(Greiner Bio-One GmbH, Germany) for time-lapse imaging.
Cells were imaged into Cytation 5 Cell Imaging Reader (Biotek,
USA) with temperature and gas control set to 37 ◦Cand 5 % CO2
atmosphere, respectively. Sequential imaging of each well was
taken using 10X objective. The bright and the phase contrast
imaging channel was used for image recording. Two times two
partly overlapping images were stitched together to form images
of appropriate size. A continuous kinetic procedure was chosen
where imaging was carried out with each designated well within
an interval of 6 min for a 94 h incubation period. Exposed cells
were recorded simultaneously subject to three different YTX
concentrations (200; 500; 1000 nM).
The following results give arguments that cell speed can be con-
sidered as a phenotype to help characterize cells. Figure 1 shows
two examples of tracks of sister cells with similar speeds af-
ter their birth. A simple way to quantify similarity between
speed of different cells is to compare their track lengths (or aver-
age speed) for given time periods. This parameter signiﬁcantly
varies among cells. Figure 2 illustrates this variation showing
distributions of track length of the speed of the ﬁrst generations
sister cells in the present recordings using this simple concept of
similarity where the actual time period is between 5 h and 15 h
Figure 3 displays joint distributions between these parameters
for the actual sister cells. Note the general variation of speed
(cf Figure 2) and the correlation between the speed of the sister
cells. This is an argument for considering cell speed as a proper
phenotype. Korsnes and Korsnes  similarly used max speed
as deﬁnition of “speed” where track length is deﬁned as length
of track subject to a smoothing operation. This work applies
similar smooting of thrack to deﬁne track length or speed.
0 5 10 15 20 25 30 35 40
Figure 1: Examples of tracks for two couples of sister cells. The top ﬁgure
shows their tracks superimposed on an image from the actual video record-
ing. The lower left part shows graphs for the corresponding track lengths as
function of time after their birth. The lower right part shows the actual 3D
tracks (position versus time) for the whole pedigree tree.
0 50 100 150 200 250 300 350 400
Track length (micron)
Distribution of track length during 5-15 hours after ﬁrst cell division
Figure 2: Distribution of track length 5-15 h after ﬁrst cell division of record-
ing. The graphs illustrate increased variation for exposed cells as compared
to the control.
Figure 3: Joint distribution of track length for ﬁrst generation sister cells
during 5-15 hours after cell division. The cells were subject to YTX ex-
posure at concentrations 200 nM,500 nMand 1000 nMas well as control (no
exposure). Each distribution results from observations of 100 initial cells nor-
mally dividing the ﬁrst day of recording. Note the rich structure and change
in distributions depending on exposure.
Sister cells tend to move more similarly as compared to random
cells in a clonal population. This strongly indicates that cell
speed reﬂects varying inherent properties. Speed therefore de-
serves to be called a phenotype. This understanding conforms
with the idea of stereotypical behaviour as explored by Luke et
al.  as well as Mencattini et al. .
The Figures 2 and 3 apply track length during 5 h -15 h after
the actual cell division. This is an arbitrary choice. The graph of
Figure 1 (lower left part) can indicate that speed similarity may
be formalized in more complex ways which more precisely re-
ﬂect innate properties. Such alternative deﬁnitions may assume
to include analyses of data from all cells in the whole available
pedigree tree in a combined analysis to deﬁne and identify dis-
tinct inheritable phenotypes of a cell . This idea has con-
ceptual similarities to using data from relatives when making
diagnosis in cancer research.
Data from trajectories of single cells contain information that
help to predict behavior in heterogeneous cell populations. Ob-
servations of heritable traits are currently difﬁcult to perform
using single cell tracking based on 3D platforms. Trajectory
data from 2D in vitro experiments may therefore give proxy data
for 3D situations since the physical underlying mechanisms for
movements in 2D and 3D are presumably much the same.
This presentation elaborates speed data to illustrate its poten-
tial as a phenotype reﬂecting varying inherit properties among
cells. Data on velocity (vector) certainly provides additional in-
formation reﬂecting coordination, persistence and duration of
processes. However, focusing on speed only in the 2D situation
may have an advantage as proxy data for 3D as compared to the
vectorial quantity velocity.
 C. D. Paul, P. Mistriotis, and K. Konstantopoulos, “Cancer cell motility: lessons from mi-
gration in conﬁned spaces,” Nature Reviews Cancer, vol. 17, no. 2, pp. 131–140, 2017.
 A. Mencattini, D. Di Giuseppe, M. Comes, P. Casti, F. Corsi, F. Bertani, L. Ghibelli,
L. Businaro, C. Di Natale, M. Parrini, et al., “Discovering the hidden messages within
cell trajectories using a deep learning approach for in vitro evaluation of cancer drug
treatments,” Scientiﬁc reports, vol. 10, no. 1, pp. 1–11, 2020.
 M. S. Korsnes and R. Korsnes, “Single-cell tracking of A549 lung cancer cells exposed
to a marine toxin reveals correlations in pedigree tree proﬁles,” Frontiers in Oncology,
vol. 8, p. 260, 2018.
 L. Tweedy, P. Witzel, D. Heinrich, R. H. Insall, and R. G. Endres, “Screening by changes
in stereotypical behavior during cell motility,” Scientiﬁc reports, vol. 9, no. 1, pp. 1–12,
 L. Andrei, S. Kasas, I. O. Garrido, T. Stankovi´
c, M. S. Korsnes, R. Vaclavikova, Y. G.
Assaraf, and M. Peˇ
c, “Advanced technological tools to study multidrug resistance in
cancer,” Drug Resistance Updates, vol. 48, p. 100658, 2020.
This work was supported by the Norwegian University
of Science and Technology (NTNU), Department of Bio-
technology and Food Science.