STOCHASTIC BARCODING FOR SINGLE-CELL TRACKING
Marc Castellarnau, Gregory L. Szeto, Darrell J. Irvine, J. Christopher Love and Joel Voldman*
Massachusetts Institute of Technology, USA
ABSTRACT
We present stochastic barcoding (SB), a method for tracking cell identity across analytical platforms. SB uses a
randomly generated code based on number, color and position of beads encapsulated together with a set of cells of
interest. We demonstrate SB use in an application where cells are transferred from a microwell array into a microtitre
plate while keeping their identity, and obtain an average identification accuracy of 96% for transfer of 100 blocks.
Finally, we model scaling of the method up to 1000 blocks and show that SB is able to achieve ~90% accuracy.
KEYWORDS: barcoding, single-cell tracking, cell encapsulation
INTRODUCTION
Acquiring multiparametric data
from single cells is critical for
assessing phenotype in heterogeneous
populations, and is increasingly used
across biology. As a result, a diverse
set of academic and commercial
platforms have been developed to
obtain single-cell data (e.g., cytokine
secretion, gene expression, function,
etc.) (Figure 1). However, because
the academic platforms do not
necessarily interface with standard
microtitre plates, transferring cells
between platforms while maintaining
cell identity is challenging. Here, we present a simple, scalable method for tracking cell identity across assay platforms.
Current methods for tracking identity across assay platforms mainly rely on direct tracking or deterministic codes.
Direct tracking by manually picking cells with micromanipulators [1] has limited throughput and requires open access to
the cells, while deterministic fluorescent labeling is limited by multiplexing depth (i.e., number of colors that can be
detected [2],[3]) and the specificity of the label [4],[5]. Instead, we developed stochastic barcoding (SB), a method that
uses bead location and color within a block photo-polymerized around cells to enable high multiplexing depth (1000’s)
without needing physical access to cells.
EXPERIMENTAL
Stochastic barcoding uses a randomly generated code determined by the number, color, and position of beads added
to and polymerized in a hydrogel block around a set of cells (Figure 2). Because we use these three parameters, we get
high coding depth using a modest number of beads, with little likelihood of overlapping codes. For cell and code
encapsulation, we prepare a polymer solution consisting of 20% PEGDA (MW1000) and 1% Irgacure 2959 as photo-
initiator, to obtain fast polymerization. We use a simple UV direct-writing approach through a microscope objective to
photopolymerize the regions of interest. Imaging of the hydrogel blocks after photopolymerization assigns the code, and
imaging after transfer to the recipient container reads the code. We implemented a custom Matlab script to identify the
codes from the images and find the best candidate to match images of pre- and post-transferred blocks. Finally, we also
developed a model for the stochastic barcoding method.
Figure 1: Overview. Many platforms are being developed for assaying single
cells, but tracking cells across platforms is challenging because academic
platforms do not often interface with microtitre plates.
Figure 2: Stochastic barcoding (SB). A) Addition of PEGDA polymer solution with fluorescent beads into single-cell
microwell array. B) Encapsulation of all or selected (shown) cells and beads by photopoylmerization of the hydrogel.
Encapsulated beads constitute a random code based on their color, number, and relative positions. C) Imaging of
blocks to assign the code. D) Transfer of blocks into a microtitre plate. E) Imaging blocks after transfer to read the
code.
978-0-9798064-6-9/µTAS 2013/$20©13CBMS-0001 690 17th International Conference on Miniaturized
Systems for Chemistry and Life Sciences
27-31 October 2013, Freiburg, Germany