brief communications
Real-time fluorescence Label-free approaches provide quantitative information about
and deformability
cytometry
Philipp Rosendahl1,6   , Katarzyna Plak1,2,6,
Angela Jacobi1, Martin Kraeter1,3   , Nicole Toepfner1,4,
© 2018 Nature America, Inc., part of Springer Nature. All rights reserved.
Oliver Otto1, Christoph Herold1, Maria Winzi1,
Maik Herbig1, Yan Ge1, Salvatore Girardo5,
Katrin Wagner1, Buzz Baum2 & Jochen Guck1   
The throughput of cell mechanical characterization has recently
approached that of conventional flow cytometers. However, this
very sensitive, label-free approach still lacks the specificity
of molecular markers. Here we developed an approach that
combines real-time 1D-imaging fluorescence and deformability
cytometry in one instrument (RT-FDC), thus opening many
new research avenues. We demonstrated its utility by using
subcellular fluorescence localization to identify mitotic cells
and test for mechanical changes in those cells in an RNA
interference screen.
Flow cytometry (FCM) is the gold standard for single-cell
characterization in biological research and numerous clinical
applications1. As cells flow through the cytometer, they are illu-
minated by lasers, and detectors collect emitted fluorescence and
forward- and side-scattered light1. The scattered light provides
information about cell morphology in the absence of a molecular
label. Recently, the assessment of cells’ mechanical properties has
emerged as a means to obtain orthogonal information about the
functional state of whole cells without labels. Importantly, cell
mechanics is tightly regulated and serves as a quantitative readout
for the state of the cytoskeleton2,3. A cell’s stiffness is influenced
by its progression through the cell cycle4,5, differentiation5,6, and
pathophysiological processes such as malignant transformation7,8
and immune-cell activation9–11. With the recent advent of several
microfluidic techniques and their massive increase in through-
put (~100–10,000 cells per second), mechanical phenotyping can
now be performed at rates of conventional FCM5,12,13. Among
many other applications, this opens up the possibility of large-
scale screening for genes that regulate cell mechanics, which was
previously limited by the low throughput (~10–100 cells per hour)
of techniques such as atomic force microscopy.
1Biotechnology Center, Center for Molecular and Cellular Bioengineering, Technische Universität Dresden, Dresden, Germany. 2MRC Laboratory for Molecular and
Cellular Biology, University College London, London, UK. 3Medical Clinic I, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany.
4Department of Pediatrics, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany. 5Microstructure Facility, Center for Molecular and
Cellular Bioengineering, Technische Universität Dresden, Dresden, Germany. 6These authors contributed equally to this work. Correspondence should be addressed to
J.G. (jochen.guck@tu-dresden.de).
Received 11 September 2017; accepted 27 February 2018; published online 2 April 2018; doi:10.1038/nmeth.4639
cells but lack molecular specificity, which can be achieved in prin-
ciple through the use of fluorescent probes1. The total amount of
a certain fluorescent probe in a cell is what is usually quantified
in standard FCM. More advanced variants of FCM, such as fluo-
rescence pulse-shape analysis14 and slit-scanning15, and the use
of line-scan cameras16 and nonlinear optics17 can even provide
the 1D and 2D intracellular distribution of fluorescence, respec-
tively. The combination of the molecular specificity of such FCM
analysis with mechanical phenotyping, which is very sensitive to
changes in cell state, promises to open up entirely new possibili-
ties for scientific exploration.
Here we present RT-FDC, which combines 1D-resolved, fluo-
rescence-based FCM with mechanical phenotyping for the con-
tinuous analysis of large cell populations (Fig. 1a). The principle
of operation is based on real-time deformability cytometry (RT-
DC)5. Up to 100 cells per second are flowed through a microflu-
idic chip (Supplementary Fig. 1) and deformed without physical
contact by hydrodynamic interactions in a constriction zone18,19
(Fig. 1b). Stroboscopic high-speed bright-field microscopy with
real-time image processing captures and evaluates the cell contour.
Simultaneously, three diode lasers excite fluorescence in a light
sheet perpendicular to the channel axis (Fig. 1a, Supplementary
Fig. 2a–c), and avalanche photodiode detectors measure emitted
fluorescence intensity in the region. We checked the synchronic-
ity of the different fluorescence channels, using custom-made
two-color fluorescent agarose beads (Supplementary Fig. 2d).
If an object is detected in the bright-field image, the custom
algorithm extracts the cross-sectional area and deformation
(Fig. 1b), which can be related to the object’s elastic modulus18,19,
and analyzes the respective fluorescence signal for peak maxi-
mum, width, and area. We confirmed that fluorescence detection
by RT-FDC is comparable to that of standard FCM with calibra-
tion beads (Supplementary Fig. 3). To test the utility of RT-FDC
for biological and biomedical research, we assessed the mechani-
cal properties of fluorescently labeled subpopulations of cells in
a heterogeneous mixture without prior sorting.
Surface marker labeling is the standard approach for identifica-
tion of many different cell types, including hematopoietic stem
and progenitor cells (HSPCs)5. Previous studies have shown that
mechanical properties of HSPCs are linked to their circulation and
migration abilities6,20—essential aspects of successful HSPC hom-
ing when the cells are used for transplantation after chemother-
apy. To test whether cell mechanics can be used as a phenotypic
marker for human CD34+ HSPCs, we measured the mechani-
cal properties of cells from granulocyte colony-stimulating
nature methods  |  VOL.15  NO.5  |  MAY 2018  |  355
 brief communications
factor (G-CSF)-mobilized peripheral blood. After classifying the a LED (460 nm)
cells into CD34+ and CD34− subpopulations on the basis of fluo-
Lasers Detector array
rescence intensity by RT-FDC (Fig. 1c), we obtained the mechani-
cal fingerprint (deformation versus area; Fig. 1d,e), which showed 640 nm

that CD34+ cells had a mean size of 61.1 ± 1.3 µm2 and low defor- 700/75
mation (0.02–0.05), corresponding to an elastic modulus of 561 nm
~1 kPa; we confirmed these findings with cells from three differ-
z
ent donors. For outliers in the CD34+ population there was no x 593/46
488 nm 40×/0.75-NA
correlation between CD34 expression and deformation or pro- Camera
jected area (Supplementary Fig. 4). Compared with the CD34+
525/50
fraction, CD34− cells were either smaller (lymphoid cells)5,11 or
larger (monocytes and granulocytes)5,11 and had a wider distribu-
tion in deformation. An exemplary multicolor analysis excluding
T cells and monocytes and using specific lineage markers is
shown in Supplementary Figure 5, and validates antibody-based b Sample Outlet c 10,000 112,000 events
N = 611
identification of HSPCs. We also tested the potential for sorting CD34+
© 2018 Nature America, Inc., part of Springer Nature. All rights reserved.

Sheath
cells by deformation and cell size, without relying on fluorescence. flow 1,000
Assuming the fluorescence readout represented the ground

Count
truth, we found that classification by mechanical fingerprint 100

(Fig. 1d) yielded sensitivity of 69.1% and specificity of 90.9%

(Supplementary Table 1). The prevalence was 0.55%, as con- 10
N = 111,389
firmed by FCM (Supplementary Fig. 6). Importantly, this would CD34
–

be sufficient to enrich for HSPCs in G-CSF mobilized blood on Deformation = 1 –

2 π Area 1
Perimeter 1 2
the basis of their morphological and mechanical phenotype alone. CD34–APC (log a.u.)
Circularity
As a second test case, to demonstrate the applicability of RT-FDC 10 µm

with cell-permeant fluorescent dyes, we used RT-FDC to deter- d 0.2

611 events e 0.2
111,389 events

mine the mechanical properties of mature red blood cells (RBCs) CD34+ Max CD34–

Density
compared with those of immature reticulocytes (Supplementary
Min
Fig. 7, Supplementary Table 2). Reticulocytes were slightly larger
Deformation

and less deformed compared with RBCs.
In addition to detecting total fluorescence intensity, the setup
also permits 1D fluorescence imaging. When cells in the channel
pass through the light sheet (3-µm thickness; Supplementary
Fig. 2a–c) at a constant speed, the temporal shape of the fluores-
cence peak provides a measure of the subcellular distribution of
fluorophores in the channel direction (Supplementary Fig. 2a).
This enables RT-FDC to provide spatial information about the
localization of fluorescence signal within cells (Supplementary
Fig. 8). We exploited this possibility to screen for changes in cell
mechanics that accompany entry into mitosis in the Drosophila
cell line Kc167. Mitotic cells, which represent ~5% of an asyn-
chronous population of proliferating cells, were detected on
the basis of nuclear envelope breakdown—characterized by the
entry of nuclear localization signal (NLS)–GFP into the cyto-
plasm, which equalized the GFP and tdTomato peak widths
(Fig. 2a, Supplementary Fig. 9)—and their specific mechani-
cal fingerprint was analyzed (Fig. 2b). This enabled us to use
RNA interference in combination with RT-FDC to investigate
the mechanical effects of 42 genes chosen for their function as
regulators of the cortical actin cytoskeleton in mitotic cells. The
detailed analysis pipeline and all screening results can be found in
Supplementary Figures 10 and 11, Supplementary Table 3, and
the Supplementary Discussion. This screen identified a set of
both known and novel genes involved in the regulation of mitotic
cortical mechanics, revealing important roles for Rho GTPase-
activating proteins in opposing the activity of the mitotic Rho
guanine nucleotide-exchange factors (primarily Ect2/Pbl). Thus,
RT-FDC makes it possible to screen for regulators of cell mechan-
ics in small subpopulations of cells. Of note, the conventional,
Deformation
0.1 0.1
0 0
0 25 50 75 100 125 150 0 25 50 75 100 125 150
Area (µm2) Area (µm2)
Figure 1 | Real-time fluorescence and deformability cytometry.
(a) Experimental setup. Cells are pumped through a microfluidic
channel and imaged by bright-field microscopy, while three lasers excite
and avalanche photodiodes measure fluorescence. (b) In the narrow
constriction zone, cells deform as a result of hydrodynamic interaction,
and deformation is determined by image processing. Focused lasers excite
fluorescence in the region marked by the yellow line. (c–e) HSPCs were
surface-labeled with anti-CD34 conjugated to allophycocyanin (APC).
(c) Log histogram of CD34–APC fluorescence, and gate used for
classification. (d) CD34+ cells form a homogeneous population. The
color-coding corresponds to a linear density scale, as indicated by the key.
The black outline around the cell cluster represents the gate used to
calculate the confusion matrix. (e) The plot of deformation versus area for
CD34− cells shows more spread than that for CD34+ cells. Experiments were
repeated three times, with similar results each time.
two-color FUCCI cell cycle probe also can be used with RT-FDC.
However, only through the 1D fluorescence-localization capabili-
ties of RT-FDC can G2- and M-phase cells be distinguished and
mechanical differences between mitotic and interphase cells be
identified (Supplementary Figs. 12 and 13).
We were also able to identify intracellular structures smaller
than the nucleus, such as separating chromatids during mitosis

356  |  VOL.15  NO.5  |  MAY 2018  |  nature methods


a 1,078/6,120 events
120
tdTomato width (µs)
Deformation
40
0 0
0 40 120
GFP width (µs)
c d
© 2018 Nature America, Inc., part of Springer Nature. All rights reserved.
mCherry intensity (a.u.)
Meta- 10 µm
mCherry intensity (a.u.)
phase
0 150
Time (µs)
Figure 2 | 1D fluorescence imaging for identification of mitotic cells.
(a) Nuclear envelope breakdown leads to equal peak widths for NLS–GFP
and tdTomato. Mitotic cells along the diagonal can be gated by the red
outline (Supplementary Fig. 9). Interphase cells are shown in gray,
whereas mitotic cells are color-coded according to a linear density scale.
(b) Deformation versus area of the mitotic subpopulation. Results in a,b
are from one experiment representative of three repeated experiments.
(c,d) Subcellular localization of H2B–mCherry in HeLa cells. Cells in (c)
metaphase and (d) anaphase exhibited one or two fluorescence peaks
corresponding to one metaphase plate (arrow in c) or two separated
chromatids in the bright-field images (arrows in d), respectively.
Experiments in c,d were performed five times independently, with similar
results each time. a.u., arbitrary fluorescence units.
in cells carrying fluorescently labeled histone 2B (Fig. 2c,d).
In this way HeLa cells could be separated into metaphase and
anaphase populations on the basis of the presence of single and
double peaks of mCherry fluorescence signal, respectively. An
important aspect for the detection of separated chromatids is the
proper alignment of cells, which are elongated along the spindle
axis, with the channel axis, which is ensured by the hydrodynamic
forces acting on the elongated cells.
The combination of the molecular specificity of fluorescent
probes with the sensitivity of the measured cell mechanics as a
functional readout in RT-FDC enables correlation of both, or gat-
ing for rare cells in large samples for mechanical analysis. In the
future, the possibility to enrich for cells by using mechanics alone
might aid or even replace the current label-based purification of
cells, which might be particularly interesting for clinically relevant
cells such as HSPCs. Our results further show that the technique
is ideal for screening for regulators of cell mechanics. Screens
focused on mechanical properties will benefit from the high
throughput per sample (100 cells per second), as well as from the
short time needed to complete one experiment (15 min), which
brief communications
b 0.07 1,078/6,120 events allowed us to carry out more than 500 individual measurements in
the presented screen for candidate genes involved in the regulation
of cell mechanics during mitosis. Future investigations of mitotic
cell mechanics could take into account information encoded in
the fluorescence peak with greater detail, as demonstrated for H2B
HeLa cells for resolution of meta-, ana-, and telophase. This aspect
highlights the need for simultaneous fluorescence and mechanical
readout: if cells had to be presorted before mechanical measure-
0 40 60 80 100 120
ment by any given method, time-sensitive phases, such as mitosis,
Area (µm2) could not be resolved because of the experimental delay. Only
simultaneous readout retrieves the full information at the single-
cell level, going beyond the state of the art of what is achievable
with FCM or deformability cytometry alone.
Methods
Methods, including statements of data availability and any associ-
ated accession codes and references, are available in the online
Ana-
10 µm
phase version of the paper.
Note: Any Supplementary Information and Source Data files are available in the
online version of the paper.
Acknowledgments
We thank D. Klaue, M. Urbanska, F. Rosendahl, and I. Richter for helpful
0 150
discussions and technical support. We thank M. Schürmann for refractive index
Time (µs)
measurements of agarose. We thank M. Piel (Institute Curie, Paris, France)
for donation of the HeLa H2B–mCherry cell line. We thank M. Bornhäuser
(Department of Hematology and Oncology, University Clinic Carl Gustav Carus,
Technische Universität Dresden, Dresden, Germany) for providing the apheresis
blood samples. We received support from the Light Microscopy Facility for
confocal microscopy, and from the Microstructure Facility for microfluidic chip
production; both are core facilities of the Center for Molecular and Cellular
Bioengineering (CMCB) at Technische Universität Dresden. We gratefully
acknowledge financial support from the Alexander-von-Humboldt Stiftung
(Humboldt Professorship to J.G.), Sächsisches Ministerium für Wissenschaft und
Kunst (TG70 grant to O.O. and J.G.; European Fund for Regional Development–
EFRE to S.G. and J.G.), the ERC (starting grant “LightTouch” 282060 to J.G.),
the DFG Center for Regenerative Medicine of the Technische Universität Dresden
(seed grant to J.G.), the DFG KFO249 (GU 612/2-2 grant to J.G.), DKMS
(‘Mechthild Harf Research Grant’ DKMS-SLS-MHG-2016-02 to A.J.), and Cancer
Research UK (C1529/A17343 grant to B.B.).
AUTHOR CONTRIBUTIONS
J.G. conceived the project. P.R. built the experimental setup. O.O. and C.H.
supported the optics design. P.R. programmed the real-time data-evaluation
software. K.P. designed and performed experiments on Kc167 and HeLa cells
and analyzed the data with B.B.’s support. P.R. and N.T. performed experiments
with RBCs. A.J., M.K., and P.R. performed experiments on HSPCs and analyzed
the data. P.R. performed and analyzed experiments for device characterization.
M.H., O.O., and Y.G. developed the linear mixed model algorithm for statistical
evaluation. K.W. and S.G. produced fluorescent beads for device characterization.
M.W. performed reference measurements with calibration beads on FCM. S.G.
produced masters for microfluidic chips. J.G. and B.B. supervised the project.
P.R., K.P., and J.G. wrote the manuscript. All authors reviewed the manuscript.
COMPETING INTERESTS
P.R., O.O., and C.H. are shareholders and employees of the company Zellmechanik
Dresden GmbH, which sells devices based on RT-DC and RT-FDC technology. All
other authors declare no competing interests.
Reprints and permissions information is available online at http://www.nature.
com/reprints/index.html. Publisher’s note: Springer Nature remains neutral
with regard to jurisdictional claims in published maps and institutional
affiliations.
1. Shapiro, M.H. Practical Flow Cytometry (John Wiley & Sons, Inc., 2003).
2. Elson, E.L. Annu. Rev. Biophys. Biophys. Chem. 17, 397–430 (1988).
3. Fletcher, D.A. & Mullins, R.D. Nature 463, 485–492 (2010).
4. Matthews, H.K. et al. Dev. Cell 23, 371–383 (2012).
5. Otto, O. et al. Nat. Methods 12, 199–202 (2015).
nature methods  |  VOL.15  NO.5  |  MAY 2018  |  357
 brief communications
6. Ekpenyong, A.E. et al. PLoS One 7, e45237 (2012).
7. Tse, H.T.K. et al. Sci. Transl. Med. 5, 212ra163 (2013).
8. Guck, J. et al. Biophys. J. 88, 3689–3698 (2005).
9. Worthen, G.S., Schwab, B.I., Elson, E.L. & Downey, G.P. Science 245,
183–186 (1989).
10. Bufi, N. et al. Biophys. J. 108, 2181–2190 (2015).
11. Toepfner, N. et al. eLife 7, e29213 (2018).
12. Gossett, D.R. et al. Proc. Natl. Acad. Sci. USA 109, 7630–7635
(2012).
13. Nyberg, K.D. et al. Biophys. J. 113, 1574–1584 (2017).
© 2018 Nature America, Inc., part of Springer Nature. All rights reserved.
358  |  VOL.15  NO.5  |  MAY 2018  |  nature methods
14. Ramdzan, Y.M. et al. Nat. Methods 9, 467–470 (2012).
15. Gray, J.W., Peters, D., Merrill, J.T., Martin, R. & Van Dilla, M.A. J.
Histochem. Cytochem. 27, 441–444 (1979).
16. Basiji, D.A., Ortyn, W.E., Liang, L., Venkatachalam, V. & Morrissey, P. Clin.
Lab. Med. 27, 653–670 (2007).
17. Lau, A.K.S., Shum, H.C., Wong, K.K. & Tsia, K.K. Lab Chip 16, 1743–1756
(2016).
18. Mietke, A. et al. Biophys. J. 109, 2023–2036 (2015).
19. Mokbel, M. et al. ACS Biomater. Sci. Eng. 3, 2962–2973 (2017).
20. Kräter, M. et al. Sci. Rep. 7, 2549 (2017).
 ONLINE METHODS
Experimental setup. Figure 1 shows the overall experimen-
tal setup used throughout this study. The microfluidic chip,
synchronized pulsed LED illumination (AcCellerator L1;
Zellmechanik Dresden), syringe pump (neMESYS; Cetoni),
and bright-field imaging (MC1365; Mikrotron) were adapted
from RT-DC as described previously 5. Supplementary
Figure 14 provides additional technical details about the excita-
tion and emission light paths too small to see in Figure 1. For
fluorescence excitation, three solid-state lasers (OBIS 488-nm
LS 60 mW; OBIS 561-nm LS 50 mW; OBIS 640-nm LX 40 mW;
Coherent Deutschland) in combination with adjustable mounted
dichroic mirrors (561-594R, 473-491R, and 1064R; Semrock)
create a combined beam. This aspect is particularly difficult to
achieve and requires advanced optics skills; the use of a shearing
interferometer is beneficial. The beam is expanded and colli-
© 2018 Nature America, Inc., part of Springer Nature. All rights reserved.
mated by two achromatic lenses (AC080-010-A-ML and AC254-
040-A-ML; Thorlabs) and then focused to a sheet in the image
plane of the microscope with a cylindrical lens (LJ1695RM-A;
Thorlabs). The position and angle of this cylindrical lens have
to be aligned carefully to focus the light sheet in the object
plane of the objective. Alignment success can be checked best
with a fluorescent sample, such as a text marker line on a thin
microscope slide. Excitation light is coupled into the micro-
scope (Axio Observer.Z1; Carl Zeiss Microscopy) through the
modified backport for epifluorescence illumination. Below the
40× objective (EC Plan-NEOFLUAR 40×/0.75-NA (numerical
aperture); Carl Zeiss Microscopy), a QuadLine beam splitter
(zt405/488/561/640rpc; Chroma Technology) reflects excita-
tion light toward the specimen but transmits light emitted by
the sample to the detector assembly. A second beam splitter
(zt 473 RDCXT; Chroma Technology) separates light from the
LED illumination with 460 nm to the CMOS (complementary
metal-oxide semiconductor) camera (MC1365; Mikrotron) for
bright-field imaging, which has a quantum efficiency of about
30% in this spectral band and operates at frame rates up to 4 kHz
for our application. Light of longer wavelengths is directed to an
adjustable slit (VA100C/M; Thorlabs) that is in the image plane
of the microscope and part of the confocal detector assembly.
The collimated beam is separated into three fluorescence chan-
nels—FL-1 (FF555-Di03, FF03-525/50; Semrock), FL-2 (zt 633
RDC, Chroma Technology Corp; FF01-593/46, Semrock), and
FL3 (700/75 ET; Chroma Technology Corp)—and finally focused
(LA1951-A-ML; Thorlabs) on the three avalanche photodiode
detectors (MiniSM10035; SensL Corporate).
The analog detector signals are digitized by a PCIe (peripheral
component interconnect express) card (NI PCIe-6531; National
Instruments Germany) at a sample rate of 1 MHz (shared between
channels) and stored in a circular buffer of 1,000,000 samples for
real-time analysis. Because the trigger signal for image acquisition
is derived from the same clock as the detector data acquisition, the
image and fluorescence acquisition are synchronized. If an object
is detected in an image, the corresponding part of the detector
signal is fetched from the circular buffer and also analyzed. For
image processing, we used the OpenCV 3.1 computer vision
library (http://opencv.org) wherever possible. Image processing
included the following steps: subtraction of a background image
(rolling average of the last 100 camera images), threshold opera-
tion, finding contours in the binary image via a border-following
doi:10.1038/nmeth.4639
algorithm21, determining particle cross-sectional area, deforma-
tion (1 – circularity) and position of the contour. Additionally,
the mean brightness of pixels inside the contour is determined
along with the s.d. In case these parameters match the gates set
before, the corresponding fluorescence detector signal is fetched
from the circular buffer and analyzed for the maximum value,
the peak width (full-width at half-maximum (FWHM)), the peak
area, and the peak position. These parameters are stored together
with results from image processing, an image of the cell, and the
fluorescence peak on the disk of the computer for later analysis
and are instantly plotted on the screen.
For characterization of the aperture function (sensitivity as a
function of position), 3-µm fluorescent beads (BD FACSDiva
CS&T Research Beads; Becton Dickinson) were embedded in
1% agarose gel (low gelling temperature A0701-100G; Sigma-
Aldrich) on a microscope slide (Thickness 2; Glaswarenfabrik
Karl Hecht) and moved through the detection volume on the
automated microscope stage while the detector signals were
recorded to obtain the combined fluorescence excitation
and emission efficiency for a field of 30 × 60 µm as shown in
Supplementary Figure 2b. Because the microscope slide was the
same as that used for the microfluidic chips and the refractive
index of 1% agarose gel is very close to that of water (n = 1.34), we
assumed that the light sheet in the microfluidic channel and the
bead test slide had very similar shapes. The resulting thickness
of the light sheet (FWHM) along the channel axis is presented
in Supplementary Figure 2c and was about 3 µm for all fluores-
cence channels. This is also the size of the fluorescent bead and
thus highlights the capability of the device to detect localization
of fluorophores in cell compartments such as the nucleus that
have a size of about 5 µm.
The dynamic range of the fluorescence detection system was
tested with flow cytometer calibration beads (Spectral calibra-
tion PMMA beads; PolyAn; lot FP170601A). As expected, eight
populations showed up in the histogram of the fluorescence
peak heights (Supplementary Fig. 3), in agreement with refer-
ence measurements provided in the manufacturer’s manual and
our own measurements obtained on an LSR II flow cytometer
(Becton Dickinson). Fluorescence peak width detection in dif-
ferent channels is compared in Supplementary Figure 2d, which
shows data from two-color fluorescent agarose beads with het-
erogeneous sizes. These beads were produced from a mixture of
two ultra-low-gelling-point agarose solutions, one functional-
ized with Alexa Fluor 488, the other with Alexa Fluor 633. For
cross-linking we decreased the temperature below the gelling
point, which resulted in stable beads that could be stored for
weeks. The linear fit showed that the peak widths in the green
and the red channels (FL-1 and FL-3) were proportional, with
a slope close to 1 and a small offset of −0.1 µs, demonstrating
that peak-width measurements can be compared across channels
for investigations of fluorophore localization (Supplementary
Fig. 2d). The particle speed in the flow for this measurement
was about 0.4 m s−1, so the particle diameters ranged from
5 µm to 35 µm based on the fluorescence signal peak widths.
All the setup characterization experiments consisted of three
technical repeats.
Experimental procedure. For all experiments, cells or micro-
particles were suspended in phosphate-buffered saline without
nature methods
 magnesium or calcium (PBS−) complemented with 0.5% or 0.6%
(w/v) methylcellulose (Alfa Aesar) to slow down sedimentation
during the experiment. The methylcellulose solution was cali-
brated with a falling-ball viscometer (Haake Typ C; Thermo Fisher
Scientific) to a viscosity of 15.0 ± 0.8 mPa·s and 25 ± 0.6 mPa·s
at 24 °C for 0.5% and 0.6% methylcellulose, respectively. Prior
to the experiments, the cell suspension was aspirated into PEEK
tubing (Upchurch; Thermo Fisher Scientific) connected to a
1-mL plastic syringe (Becton Dickinson) at a flow rate of 1 µl s−1,
avoiding high stress on the specimen. Then the tubing was
connected to the microfluidic chip (Fig. 1b, Supplementary
Fig. 1), where constant flow was generated by a combination of
sheath and sample flow with a ratio of 3:1. After equilibration of
the flow for 2 min, the measurement was started. The camera’s
region of interest (256 × 96 pixels) was positioned at the end of the
300-µm-long channel region of the chip to ensure cell deforma-
© 2018 Nature America, Inc., part of Springer Nature. All rights reserved.
tion had reached a steady state. Laser powers can be adjusted in
the range of 1–60 mW to match the detector’s dynamic range
in an optimal way. Because of the low spectral overlap between
fluorophores used in our experiments, no compensation was
applied to raw fluorescence data. Compensation procedures
may, however, be performed according to standard methods.
Depending on the cell concentration in the sample, meas-
urements ran for one to a few minutes and yielded typically
2,000–100,000 events at rates of up to 100 events per second.
Higher throughput rates were not possible because multiple cells
in the image might impede the correct assignment of fluores-
cence peaks to cells in the image data. Post-experimental data
analysis was performed with the following programs: ShapeOut
(Zellmechanik Dresden; https://github.com/ZELLMECHANIK-
DRESDEN/ShapeOut) for plotting and gating, and OriginPro
9.0 (OriginLab Corporation) for more detailed data inspection,
plotting, and fitting.
Hematopoietic stem cell isolation and preparation. With
approval for the study (EK221102004) from the ethics com-
mittee of the Technische Universität Dresden, we analyzed the
apheresis product from G-CSF-mobilized peripheral blood from
three healthy human donors, obtained with their informed con-
sent in accordance with the guidelines of good practice and the
Declaration of Helsinki. Before measurement by RT-FDC, HSPCs
were stained for 30 min with anti-CD34–APC (555824; Becton
Dickinson), then pelleted by centrifugation (200g, 5 min, 23 °C)
and resuspended in 0.6% methylcellulose solution.
Reticulocyte and red blood cell preparation. With approval
for the study (EK89032013) from the ethics committee of the
Technische Universität Dresden, we obtained blood from three
healthy donors with their informed consent in accordance with
the guidelines of good practice and the Declaration of Helsinki.
Capillary blood was collected by finger prick with a 21-gauge,
1.8-mm safety lancet (Sarstedt). For sample preparation, 2 µL
of blood was diluted in 1 mL of 0.5% methylcellulose solution
complemented with 2.5 µM syto13 nucleic acid stain (Thermo
Fisher Scientific). We measured cells from three healthy donors
after 5 min of incubation; measurements usually took 10 min. To
validate the stability of mechanical phenotype and fluorescence
staining in methylcellulose measurements, we used a 1-h time
course with measurements every 10 min for all donors. To validate
nature methods
counting results, we collected blood from venipuncture with
a 21-gauge needle (Multifly; Sarstedt) in EDTA (S-Monovette
1.4 mL 9NC; Sarstedt) for measurements with RT-FDC and
parallel full blood counts on a state-of-the-art device (XE-5000;
Sysmex Deutschland) at the Institute for Clinical Chemistry
and Laboratory Medicine at the Carl Gustav Carus University
Hospital, Technische Universität Dresden.
Cell lines and culture. The stable H2B–mCherry HeLa cell line
was kindly provided by the lab of Matthieu Piel (Institute Curie).
All HeLa cells were grown in standard DMEM supplemented
with 10% FBS and 2 mM l-glutamine. Transfection of HeLa
cells was done with Effectene reagent (Quiagen) according to
the manufacturer’s directions. We allowed cells to grow for 48 h
after transfection, and we removed dead cells by washing away
loosely attached cells. Prior to RT-FDC measurements, adherent
cells were trypsinized, collected by centrifugation (200g, 5 min,
23 °C), and resuspended in 0.5% methylcellulose buffer.
For mitotic synchronization of the H2B–mCherry HeLa line,
cells were allowed to grow to 35% confluence and were then syn-
chronized with 100 nM nocodazole for 5 h. Mitotic cells were
collected by shake-off, and nocodazole was washed away. Cells
were then allowed to progress through mitosis in full media for
15 min (metaphase cells) or 1 h (anaphase cells).
Kc167 Drosophila cells were grown in M3 Shields and Sang
medium supplemented with 10% FBS. For detection of mitotic
cells, we engineered a stable Kc167 cell line that expresses tdTo-
mato cytosolic protein and NLS(GFP)2(GST), a protein that
stably localizes to the nucleus during interphase. At the entry
to mitosis, the nuclear envelope breaks down and the nuclear
GFP signal is redistributed to the cytoplasm, where it colocal-
izes with tdTomato. Analysis of fluorescence peak widths distin-
guished interphase cells (narrow GFP peak and broad tdTomato
peak) from mitotic cells (both peaks of the same width; Fig. 2a).
Owing to progression through mitosis and cytokinesis, cells in
late mitotic stages have nonspherical shapes due to elongation
and furrow constriction before division (anaphase and telo-
phase of mitosis), which can bias the deformation readout of
RT-FDC. To avoid this effect, we focused on prometaphase cells
by synchronization with colchicine, a drug that inhibits forma-
tion of the mitotic spindle. Cells were synchronized with 4 µM
colchicine for 5 h for enrichment of the mitotic cell fraction
(Supplementary Fig. 9). RNA interference experiments were
conducted as described22. The treatment with colchicine led to
an increase in mitotic cell size; however, it did not affect the
mechanical phenotype of mitotic cells (Supplementary Table 4)
and, as expected, it led to enrichment of the mitotic cell fraction
(Supplementary Fig. 9).
RNA interference screening. For each of the 42 candidate genes
(Supplementary Table 5), RNA interference was performed with
two nonoverlapping double-stranded RNA sequences. Data from
three independent experiments performed on different days were
collected for each RNA sequence and compared with data for
negative controls acquired on the same day (Supplementary
Fig. 11). For evaluation of the measured elastic modulus values,
a linear mixed model (LMM)23 comparison was performed with
the software ShapeOut (Zellmechanik Dresden), for which data
for each gene (different days, different sequences) were pooled.
doi:10.1038/nmeth.4639
 We rejected the null model on the basis of a P value < 0.05 as a
hit. Only if the effect was visible for both flow rates (0.04 and
0.06 µL s−1) was the gene considered to play a role in the regula-
tion of mitotic cell mechanical properties.
Statistical analysis. Statistical analysis was done via the LMM
method integrated into ShapeOut. The method is described in
detail in ref. 23. LMM evaluation was necessary to take into
account random variations of the mean values between replicates
caused by variables that are not controlled by the experimenter.
We defined and fitted an LMM using the R package “lme4”24,25.
To test the null hypothesis, we fitted a model with and without the
fixed effect term and compared the two models by a likelihood
test26. From the resulting likelihood ratio, one can compute the
F-value and P value by using Wilks theorem27.
© 2018 Nature America, Inc., part of Springer Nature. All rights reserved.
Code availability. The source code of ShapeOut, the program
used to analyze and plot the RT-FDC data, is open source and
can be found on GitHub (https://github.com/ZELLMECHANIK-
DRESDEN/ShapeOut). The source code for the custom C++
acquisition software is available from the corresponding author
upon reasonable request.
doi:10.1038/nmeth.4639
Life Sciences Reporting Summary. Further information on
experimental design is available in the Life Sciences Reporting
Summary.
Data availability. The data that support the findings of this study,
as well as general information and biological materials, are avail-
able from the corresponding author upon reasonable request.
The raw RT-FDC data are available as TDMS files that can be
read, processed, and analyzed by ShapeOut. Polygon filters used
to select the cell populations of interest are available upon request
and are also compatible with the ShapeOut software. Source data
for Figures 1 and 2 are available online.
21. Suzuki, S. Comput. Vis. Graph. Image Process. 46, 32–46 (1985).
22. Baum, B. & Cherbas, L. Methods Mol. Biol. 420, 391–424 (2008).
23. Herbig, M. et al. in Flow Cytometry Protocols (eds. Hawley, R. & Hawley, T.)
347–370 (Humana Press, 2017).
24. R Core Team. R: A Language and Environment for Statistical Computing
(R Foundation for Statistical Computing, 2015).
25. Bates, D. R News 5, 27–30 (2005).
26. Mood, A.M., Graybill, F.A. & Boes, D.C. Introduction to the Theory of
Statistics 3rd ed. 540–541 (McGraw-Hill, 1974).
27. Wilks, S.S. Annals of Mathematical Statistics 9, 60–62 (1938).
nature methods
 nature research | life sciences reporting summary
Corresponding author(s): Jochen Guck
Initial submission Revised version Final submission

Life Sciences Reporting Summary

Nature Research wishes to improve the reproducibility of the work that we publish. This form is intended for publication with all accepted life
science papers and provides structure for consistency and transparency in reporting. Every life science submission will use this form; some list
items might not apply to an individual manuscript, but all fields must be completed for clarity.
For further information on the points included in this form, see Reporting Life Sciences Research. For further information on Nature Research
policies, including our data availability policy, see Authors & Referees and the Editorial Policy Checklist.

` Experimental design
1. Sample size
Describe how sample size was determined. In general sample sizes were 100x higher than in comparable studies with AFM,
optical stretcher or micropipette aspiration. Statistical test applied yielded
significant differences for standard assays with cytoskeletal drugs with the sample
sizes used (Golfier et al., Cytoskeleton 2017)
No statistical methods were used to predetermine sample size.
2. Data exclusions
Describe any data exclusions. - CD34+ HSPCs: only objects smaller than channel (20 μm) were processed
- Reticulocytes: only objects smaller than channel (20 μm) were processed,
doublets were excluded (area def. plot), also images with two or more cells were
excluded
- HeLa H2B NLS: only objects smaller than channel (20 μm) were processed, non-
fluorescent cells were excluded
- KC167 screen: only objects smaller than channel (20 μm) were processed, non-
fluorescent cells were excluded
3. Replication
Describe whether the experimental findings were - CD34+ HSPCs: 3 replications, reliable
reliably reproduced. - Reticulocytes: 3 replications, reliable
- HeLa H2B NLS: >3 replicaitons, reliable
- KC167 screen: 2 dsRNA sequences for each gene tested. 3 replications for each
dsRNA sequence. Altogether 6 different measurement performed on separate
days. Only genes with consistent results for all replications/sequences were
considered a hit
4. Randomization
Describe how samples/organisms/participants were - CD34+ HSPCs: There were no exp. groups
allocated into experimental groups. - Reticulocytes: There were no exp. groups
- KC167 screen: replications were done in randomized order on different exp. days
5. Blinding
Describe whether the investigators were blinded to - dsRNAs for specific gene knock downs were generated and assigned a random
group allocation during data collection and/or analysis. number from range 1-42. The knock down experiments were then performed with
given dsRNAs and only after the final analysis the gene names were retrieved.
- Data analysis of KC167 samples was done automated
Note: all studies involving animals and/or human research participants must disclose whether blinding and randomization were used.
June 2017

1
Nature Methods: doi:10.1038/nmeth.4639
 6. Statistical parameters

nature research | life sciences reporting summary

For all figures and tables that use statistical methods, confirm that the following items are present in relevant figure legends (or in the
Methods section if additional space is needed).

n/a Confirmed

The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement (animals, litters, cultures, etc.)
A description of how samples were collected, noting whether measurements were taken from distinct samples or whether the same
sample was measured repeatedly
A statement indicating how many times each experiment was replicated
The statistical test(s) used and whether they are one- or two-sided (note: only common tests should be described solely by name; more
complex techniques should be described in the Methods section)
A description of any assumptions or corrections, such as an adjustment for multiple comparisons
The test results (e.g. P values) given as exact values whenever possible and with confidence intervals noted
A clear description of statistics including central tendency (e.g. median, mean) and variation (e.g. standard deviation, interquartile range)
Clearly defined error bars

See the web collection on statistics for biologists for further resources and guidance.

` Software
Policy information about availability of computer code
7. Software
Describe the software used to analyze the data in this ShapeOut (0.7.5) for RT-FDC data analysis and linear mixed model analysis (github)
study. custom C++ code for acquisition is available upon request
OriginLab 9.0 for plotting
Excel 2013 for KC167 screening data
For manuscripts utilizing custom algorithms or software that are central to the paper but not yet described in the published literature, software must be made
available to editors and reviewers upon request. We strongly encourage code deposition in a community repository (e.g. GitHub). Nature Methods guidance for
providing algorithms and software for publication provides further information on this topic.

` Materials and reagents

Policy information about availability of materials
8. Materials availability
Indicate whether there are restrictions on availability of Microfluidic Chips and methyl cellulose medium are available from
unique materials or if these materials are only available ZellmechanikDresden GmbH
for distribution by a for-profit company. All other materials used are available from standard suppliers (Sigma,
Thermofisher, ...)
9. Antibodies
Describe the antibodies used and how they were validated Supplier / CatNo. / clone / Lot / dilution
for use in the system under study (i.e. assay and species). BD Pharming / 555824 / 581 (RUO) / 99056 / 1:5
BD Pharming / 555751 / MOPC-21 (RUO) / n.a. / 1:5

Antibodies were used as specified by the manufactures protocol for flow

cytometry. Anti-CD34 APC was validated against manufacturers recommended IgG
control. In brief, 1 x 10^6 cells were diluted in 100 μl PBS and incubated either with
20 μl anti-CD34 APC antibody or 20 μl IgG APC control for 15 at room temperature.
After washing with PBS cells were re-suspended in 300 μl MC-PBS and RT-FDC
measurement started immediately. Positive cells were separated by gating
according to IgG control cell population.
June 2017

2
Nature Methods: doi:10.1038/nmeth.4639
 10. Eukaryotic cell lines
a. State the source of each eukaryotic cell line used.
b. Describe the method of cell line authentication used.
c. Report whether the cell lines were tested for
mycoplasma contamination.
d. If any of the cell lines used are listed in the database
of commonly misidentified cell lines maintained by
ICLAC, provide a scientific rationale for their use.
` Animals and human research participants
Policy information about studies involving animals; when reporting animal research, follow the ARRIVE guidelines
11. Description of research animals
Provide details on animals and/or animal-derived
materials used in the study.
Policy information about studies involving human research participants
12. Description of human research participants
Describe the covariate-relevant population
characteristics of the human research participants.
Nature Methods: doi:10.1038/nmeth.4639
nature research | life sciences reporting summary
HeLa H2B: Received from Mathieu Piel Lab (Institut Curie, Paris)
KC167: received from Buzz Baum Lab (MRC Lab UCL London)
HeLa H2B: authentication was not performed. Missidentification would not affect
the results, since we don't draw conclusions specific to the HeLa cell line.
KC167: authentication was not performed.
This cell line does not belong to the list of frequently misidentified cell lines. Due to
the specific growth conditions (25C and atmospheric CO2) and media
requirements cross contamination with human cell lines used in the same lab is not
possible
All cell lines in our lab tested for mycoplasma infection regularly (every 2 months)
Results were negative for all cell lines in the time these studies were performed.
KC167 cells are not susceptible for contamination with Mycoplasma
Provide a rationale for the use of commonly misidentified cell lines OR state that no
commonly misidentified cell lines were used.
For laboratory animals, report species, strain, sex and age OR for animals observed
in or captured from the field, report species, sex and age where possible OR state
that no animals were used.
Reticulocytes: Healthy volunteer donors, samples were anonymized
CD34+ HSPCs: Healthy volunteer donors, samples were anonymized
June 2017
3
 nature research | flow cytometry reporting summary
Corresponding author(s): Jochen Guck
Initial submission Revised version Final submission

Flow Cytometry Reporting Summary

Form fields will expand as needed. Please do not leave fields blank.

` Data presentation
For all flow cytometry data, confirm that:
1. The axis labels state the marker and fluorochrome used (e.g. CD4-FITC).
2. The axis scales are clearly visible. Include numbers along axes only for bottom left plot of group (a 'group' is an analysis of
identical markers).
3. All plots are contour plots with outliers or pseudocolor plots.
4. A numerical value for number of cells or percentage (with statistics) is provided.

` Methodological details
5. Describe the sample preparation.
G-CSF (granulocyte colony-stimulating factor) mobilized peripheral blood
was obtained from healthy donors after informed consent (ethical
approval no. EK221102004, EK47022007) and stained with CD34-APC
antibody conjugate for 30 minutes as stated in the methods part.

6. Identify the instrument used for data collection. LSR II flow cytometer (BD Bioscience)

7. Describe the software used to collect and analyze Flow cytometry data was collected with BD FACS Diva software and
the flow cytometry data. analyzed with FlowJo

8. Describe the abundance of the relevant cell no sorting was performed with the flow cytometer
populations within post-sort fractions.

9. Describe the gating strategy used. Gating strategy is described in Supplementary Figure 3: Forward scatter vs.
side scatter plot with gate 1 to separate cell events from debris. (b)
Fluorescently labeled CD34+ cells are found in gate 2 in the side scatter vs.
APC intensity plot. (c) Histogram of fluorescence peak heights for cells of
gate 1 (gate 3 = CD34+)
Tick this box to confirm that a figure exemplifying the gating strategy is provided in the Supplementary Information.

June 2017

1
Nature Methods: doi:10.1038/nmeth.4639