Slide-seq: A Scalable Technology for Measuring Genome-Wide Expression at High Spatial Resolution

Abstract

The spatial organization of cells in tissue has a profound influence on their function, yet a high-throughput, genome-wide readout of gene expression with cellular resolution is lacking. Here, we introduce Slide-seq, a highly scalable method that enables facile generation of large volumes of unbiased spatial transcriptomes with 10 µm spatial resolution, comparable to the size of individual cells. In Slide-seq, RNA is transferred from freshly frozen tissue sections onto a surface covered in DNA-barcoded beads with known positions, allowing the spatial locations of the RNA to be inferred by sequencing. To demonstrate Slide-seq’s utility, we localized cell types identified by large-scale scRNA-seq datasets within the cerebellum and hippocampus. We next systematically characterized spatial gene expression patterns in the Purkinje layer of mouse cerebellum, identifying new axes of variation across Purkinje cell compartments. Finally, we used Slide-seq to define the temporal evolution of cell-type-specific responses in a mouse model of traumatic brain injury. Slide-seq will accelerate biological discovery by enabling routine, high-resolution spatial mapping of gene expression.

One Sentence Summary
Slide-seq measures genome-wide expression in complex tissues at 10-micron resolution.

Full text

!
!
!
!
!
!
!
!
Title: Slide-seq: A Scalable Technology for Measuring Genome-Wide Expression
at High Spatial Resolution
Authors: Samuel G. Rodriques1,2†, Robert R. Stickels3,4,5†, Aleksandrina Goeva5, Carly A.
Martin5, Evan Murray5, Charles R. Vanderburg5, Joshua Welch5, Linlin M. Chen5, Fei Chen*5††,
Evan Z. Macosko*5,6††
Affiliations:
1Department of Physics, Massachusetts Institute of Technology, Cambridge, MA, 02139.
2MIT Media Lab, Massachusetts Institute of Technology, Cambridge, MA, 02139.
3Graduate School of Arts and Sciences, Harvard University, Cambridge, MA, 02138.
4Division of Medical Science, Harvard Medical School, Boston, MA, 02115.
5Broad Institute of Harvard and MIT, Cambridge, MA, 02142.
6Department of Psychiatry, Massachusetts General Hospital, Boston, MA, 02114.
*Correspondence to: chenf@broadinstitute.org, emacosko@broadinstitute.org
†These authors contributed equally to this work.
††These authors contributed equally to this work.
Abstract:
The spatial organization of cells in tissue has a profound influence on their function, yet a high-
throughput, genome-wide readout of gene expression with cellular resolution is lacking. Here,
we introduce Slide-seq, a highly scalable method that enables facile generation of large volumes
of unbiased spatial transcriptomes with 10 µm spatial resolution, comparable to the size of
individual cells. In Slide-seq, RNA is transferred from freshly frozen tissue sections onto a
surface covered in DNA-barcoded beads with known positions, allowing the spatial locations of
the RNA to be inferred by sequencing. To demonstrate Slide-seq’s utility, we localized cell types
identified by large-scale scRNA-seq datasets within the cerebellum and hippocampus. We next
systematically characterized spatial gene expression patterns in the Purkinje layer of mouse
cerebellum, identifying new axes of variation across Purkinje cell compartments. Finally, we
used Slide-seq to define the temporal evolution of cell-type-specific responses in a mouse model
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
of traumatic brain injury. Slide-seq will accelerate biological discovery by enabling routine,
high-resolution spatial mapping of gene expression.
One Sentence Summary: Slide-seq measures genome-wide expression in complex tissues at 10-
micron resolution.
Main Text: The functions of complex tissues are fundamentally tied to the organization of their
resident cell types. However, unbiased methods for exploring, genome-wide, spatial
distributions of gene expression in tissues are lacking. Recently developed multiplexed in situ
hybridization and sequencing-based approaches measure gene expression within cells and tissues
with subcellular spatial resolution (1, 2), but can be laborious, and require specialized knowledge
and equipment. In addition, most in situ approaches require the upfront identification and
selection of specific target genes for measurement, which may limit de novo discovery of
spatially varying genes. By contrast, previous technologies for spatially encoded RNA-
sequencing using barcoded oligonucleotide capture arrays are presently limited to resolutions in
the hundreds of microns (3), which is insufficient for detecting many important tissue features.
To develop Slide-seq, we first asked whether barcoded oligonucleotides could be arrayed
randomly on a surface at high spatial resolution, with locations determined post hoc. We packed
uniquely barcoded 10 µm microparticles (‘beads’)—similar to those used by the Drop-seq
approach to scRNA-seq (4)—onto a rubber-coated glass coverslip forming a monolayer we
termed a “puck” (Fig. S1, 92.9% ± 2.15% packing). We found that the bead barcode sequences
on the puck could be uniquely determined via SOLiD sequencing-by-ligation chemistry (Fig.
1A, Fig. S1). Moreover, pucks could be stored for extended periods of time prior to use,
allowing them to be produced in large batches and used as needed.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
Next, to capture RNA from tissue with high resolution, we developed a protocol wherein
fresh-frozen tissue sections (10 µm thickness) were transferred onto the dried bead surface via
cryosectioning (4). We observed efficient hybridization of tissue-extracted mRNA to polyT
capture sequences, after which 3’-end digital expression libraries could be prepared and
sequenced (5). The simplicity of the protocol enabled fast, facile generation of libraries in a
highly multiplexed fashion (Fig. 1B). To highlight its generalizability, we performed Slide-seq
across a range of samples, generating data from three different organs (mouse brain, kidney,
liver). Expression measurements by Slide-seq agreed well with those from bulk mRNAseq (r =
0.89) (Fig. S1D), similar to standard single-cell profiling technologies (5), and average mRNA
transcript capture per cell was consistent across tissues and experiments (Fig. 1F).
Unbiased clustering of individual bead profiles using single-cell analysis approaches (4)
yielded cluster assignments reflecting known positions of cell types in the assayed tissues (Fig
1C). Specifically, in analyses of three different regions of brain—cerebellum, hippocampus, and
olfactory bulb—the different neuronal cell types that form the layered tissue architecture were
immediately detectable, as were populations of resident glial types. In kidney, we identified
clusters of cells representing podocytes, surrounded by proximal and distal tubule populations.
In liver, the lobule architecture was visible, with genes known to manifest zonation patterns
driving the distinction between two main clusters (6). To quantify diffusion in the tissue, we
compared the width of mRNA transcript density in hippocampal CA1 observed in Slide-seq to
that observed in an adjacent, DAPI-stained tissue section (Fig. 1D). We estimated the length-
scale of lateral diffusion of transcripts during hybridization to be 1.4um ± 1.3um (Fig. 1E),
implying that mRNA is transferred from the tissue to the beads with high spatial resolution.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
Positioning cell types defined by large-scale scRNA-seq datasets represents an important
application of spatial technologies. To map scRNA-seq cell types onto Slide-seq data, we
developed a computational approach called Non-negative Matrix Factorization Regression
(NMFreg). NMFreg reconstructs expression of each Slide-seq bead as a weighted combination
of metagene factors, each corresponding to the expression signature of an individual cell type,
defined from scRNA-seq (Fig. 2A). Application of NMFreg to a Slide-seq puck yields a
quantitative estimate of the contriubution of each scRNA-seq-defined cell type to each. Bead,
allowing the locations of cell types in space to be inferred. Using cell type definitions from a
recently acquired scRNA-seq dataset of the mouse cerebellum (7), NMFreg on a mouse
cerebellar puck recapitulated the spatial distributions of classical neuronal and non-neuronal cell
types, such as granule cells, Golgi interneurons, unipolar brush cells, Purkinje cells, and
oligodendrocytes (Fig. 2B). We found that 65.8% +/- 1.4% of beads showed significant
representation (defined by at least 25% loading) of only one cell type (4), whereas 32.6% +/-
1.2% showed significant representation of two cell types (mean ± std, N=7 cerebellar pucks)
(Fig. 2C left, Fig. S3). The high spatial resolution of the method was found to be key for
assigning beads to individual cell types with high confidence: upon artificially aggregating the
data into larger feature sizes, we failed to confidently map cell types in heterogeneous regions of
tissue, while homogenous regions such as the granular layer of the cerebellum remained
reasonably cell-type-specific (Fig. S4). Importantly, the representation of cell types in Slide-seq
more accurately represented the spatial distribution of cell types expected from
immunohistochemistry than single-cell sequencing, thus allowing for better identification of rare
cell types with distinct spatial organization: whereas Purkinje neurons make up only 0.7% of
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
cerebellar single-cell atlas data, they make up 7.8% ± 1.3% (mean ± std, N=7 pucks) of the area
of a cerebellar puck (Fig. 2C, right).
To demonstrate the scalability of Slide-seq, we applied it to 66 tissue slices from a single
dorsal mouse hippocampus, covering a volume of 39 cubic millimeters, with roughly 10 µm
resolution in the dorsal-ventral and anterior-posterior axes, and ~20 µm resolution in medial-
lateral axis. This region contained approximately 1 million beads that could be confidently
assigned to single cell types. We computationally co-registered pucks along the medial-lateral
axis, allowing for visualization of the cell types and gene expression in the hippocampus at high
resolution in three dimensions (Fig. 2D, E, Supplementary Video 1). We plotted metagenes
comprised of markers—defined by a recent large-scale single-cell study (7)—for the dentate
gyrus, CA2, CA3, a subiculum subpopulation, an anteriorly localized CA1 subset (exemplified
by the marker Tenm3) and cells undergoing mitosis and neurogenesis. The metagenes were
highly expressed and specific for the expected regions (Fig. 2F), confirming the ability of Slide-
seq to localize both common cell-types as well as subtler cellular subpopulations. The entire
experimental processing for these 66 pucks (excluding puck generation) required roughly 40
person-hours (4), and only standard experimental apparatus associated with cryosectioning and
next-generation sequencing, making Slide-seq readily scalable to measure gene expression in
large tissue volumes.
One key advantage of Slide-seq’s high-resolution, genome-wide approach is the ability to
identify spatially variable genes within specific cell types. To find genes with non-random spatial
patterns, we developed a nonparametric, kernel-free algorithm to identify genes with spatially
non-random distribution across the puck (Fig. S5) (4). Application of this algorithm to a
coronally sliced cerebellum puck identified Ogfrl1, Prkcd and Atp2b1 as possessing highly non-
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
random patterns of expression localized just inferior to the cerebellum (Fig. 3A). We found
Ogfrl1 in particular to be a highly specific marker for PV interneurons in the molecular and
fusiform layers of the dorsal cochlear nucleus (Fig. 3B), likely the cartwheel cells of the dorsal
cochlear nucleus that are thought to be involved in the generation of feedforward inhibition (8,
9). The identification of marker genes for this population may assist in future optogenetic and
labeling studies of the cochlear nucleus.
Our algorithm also identified Rasgrf1 as having significant nonrandom spatial
distribution (p<0.001, N = 3 coronal cerebellar pucks) within the granule cell layer of the
cerebellum (Fig. 3C, cyan), a pattern previously identified using ISH data (10) (Fig. 3D, left).
We asked whether Slide-seq could discover any novel spatially patterned genes in the granular
layer, which constitutes the main source of input to the cerebellar Purkinje cells. We divided the
puck in Fig. 3C into anterior and ventral regions (Fig. S6) (4) and identified four genes (Gprin3,
Cemip, Syndig1l, and Mab21l2) that showed strong and specific localization to the granule layer
on the ventral side of the puck, especially lobules VIII through X (Fig. 3C, yellow). Gprin3, in
particular, showed a distinct pattern, validated by the Allen Brain Atlas (Fig. 3D, right), which
was largely the opposite of Rasgrf1.
The cerebellum is marked by parasagittal bands of gene expression in the Purkinje layer
which are known to correlate with Purkinje cell physiology and projection targets (11–14). The
most well-known of these genes is Aldoc (also known as the antigen of the Zebrin II antibody);
several genes show similar or complementary parasagittal expression (13, 15, 16), but a
systematic classification of banded gene patterns is lacking. We applied our spatial gene
significance algorithm to the beads marked by NMFreg as Purkinje cells or Bergmann glia (the
two resident cell types within the Purkinje layer) (4), identifying 669 candidate genes, including
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
known markers of Purkinje banding, such as Plcb4 and Nefh. Amongst these 669, we found 57
that correlate more with Aldoc than with Plcb4, a gene that is known to be expressed only in
Zebrin II-negative bands (17), and 69 genes that correlate more with Plcb4 than with Aldoc (Fig.
3E). Among the Plcb4-associated genes were 4 ATPases and 4 sodium channels: Atp1a3,
Atp1b1, Atp2b2, Atp6ap1l, Kcnab1, Kcnc3, Kcng4 and Kcnma1. Of these, Kcng4 has previously
been associated with increased firing rate in fast motor neurons (18), suggesting that its
expression contributes to the faster spiking measured in Zebrin II-negative Purkinje neurons (12,
19), while the calcium-dependent channel Kcnma1 is known to regulate the timing of dendritic
calcium burst spiking in Purkinje cells (20), suggesting that it contributes to differences in
bursting activity previously observed between lobules III-V and X (21).
Beyond the classic parasagittal bands, previous studies have noted that the lobules of the
cerebellum have distinct cognitive functions. In particular, lobules VI and VII have been
associated primarily with cognitive tasks (20), whereas lobules IX and X have been associated
with the vestibular system (21, 22), so we asked whether there might also be differences in gene
expression between these two regions. We therefore examined all genes with significantly
enriched or depleted expression in each of these lobules (Fig. S6). Most of the genes thus
identified displayed highly correlated expression in lobules IX/X and VI/VII (Fig. 3F),
consistent with standard Zebrin II staining (Fig. 3G), but several, such as Prkci, Prkcd (13) and
Hspb1 (23), were restricted just to lobules IX and X. By contrast, other genes, which included
H2-D1, Cops7a, and Kmt2c, displayed relatively uniform expression except in lobule X (Fig.
3H), confirming that lobules IX and X have their own independent program of gene expression.
We found other Aldoc-associated genes, exemplified by B3galt5 (4, 23), which showed
exclusive expression in lobules IX/X and VI/VII (Fig. 3I), suggesting that these regions might
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
share a third pattern of gene expression, despite the highly disparate cognitive roles typically
associated with them. Consistent with the hypothesis of a third gene pattern, we also found
several Plcb4-associated genes, including Gnai1, which had ubiquitous expression except in
lobules IX/X and VI/VII (Fig. 3J). Remarkably, most of the genes that we found along with
B3galt5 are not Purkinje markers, suggesting that some spatial patterns in the cerebellum may
extend to other cell types in addition to Purkinje cells. Indeed, the most prominent example we
found is Mybpc1, a little-studied Bergmann cell marker that appears both in Slide-seq data (Fig.
3K) and in ISH data (Fig. 3L) to have a pattern of expression similar to Aldoc, Kctd12, and
Car7. We thus conclude that the Purkinje cells, Bergmann glia, and granule cells of the
cerebellum are all subdivided into many spatially segregated subpopulations. This finding is
particularly surprising, since none of these distinctions was previously observed in single-cell
sequencing studies (7, 24). Although differences in Purkinje cell physiology across the
cerebellum have been well-studied, we expect that future studies may reveal an as-yet-
undiscovered, region-specific role for Bergmann glia as well.
Finally, to illustrate the utility of Slide-seq for studying biological responses to
pathology, we applied it to a model of traumatic brain injury. Cortical injury was induced with an
intracranial injection needle, and mice were sacrificed 2 hours, 3 days, or 2 weeks following
injury. The puncture site was visualized in Slide-seq data by the presence of hemoglobin
transcripts (at 2 hours, Fig. 4A), or transcripts of Vim, Gfap, and Ctsd (at 3 days and 2 weeks),
and the regions of the puck thus identified matched the location of the injection site as
determined by histology (Fig. 4B,C). Using the overlap algorithm developed for the sagittal
cerebellum analysis, we identified all genes that correlated spatially with those transcripts. At the
2 hour timepoint, the only gene we found in this way was Fos, although we also found that
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
rRNA correlated significantly with the hemoglobin genes (Fig. 4A) only at the injury site (Fig.
S7).
At the 3-day timepoint, in contrast to the 2-hour timepoint, we observed many genes at
the injection site marking a robust astrocytic and immune response. Running NMFreg on the
injection pucks, we observed a structured distribution of injury-associated cell types around the
injection site at both the 3-day and 2-week timepoints. At the 3-day timepoint, we observed
beads assigned to be microglia/macrophages localized to the lesion, bordered by a distinct layer
of cells expressing mitosis-associated factors, followed by a layer of astrocytes (Fig. 4D),
suggesting that mitosis occurs primarily in contact with the lesion itself. In contrast, at the 2-
week timepoint, we observed microglia/macrophage-assigned beads filling the lesion,
surrounded by consecutive layers of astrocytes and neurons, suggesting that mitosis had ceased
by two weeks (Fig. 4E). Quantifying the thickness of these features (Fig. 4F), we found the
mitotic layer at 3 days to be 92.4 µm ± 11.3 µm (mean ± sterr, N=3), suggesting several cell-
widths of mitotic cells. By contrast, at 2 weeks, the astrocytic scar layer, defined as the distance
between the half-maximum of the astrocyte layer and the half-maximum of the neuron layer, was
36.6 µm ± 13.4 µm (mean ± sterr, N=6), suggesting only one or two cell widths between the
lesion and the nearest neurons. In addition, we also observed penetration of beads assigned to the
microglia/macrophage through the astrocytic scar and into neuron-rich regions (39 µm ± 17.8
µm, N=6). Finally, we asked whether macrophages were localized differently from microglia.
We plotted counts of the Lyz2 gene, a specific marker for macrophages and granulocytes (7, 25),
and found that the distance between the half-maximum of the Lyz2 distribution and the half-
maximum of the microglia distribution was 49.7 µm ± 5.9 µm (mean ± sterr, N=6), suggesting
that although microglia inhabit both the astrocytic layer of the scar and the lesion location itself,
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
macrophages only inhabit the lesion. Thus, Slide-seq enables precise dissection of the spatial
relationships amongst different cell-types, with resolution on the order of individual cells.
In order to investigate other changes in gene expression between the 3-day and 2-week
timepoints, we ran spatial overlap analysis (4) to identify genes correlating with Vim, Gfap, and
Ctsd at the 3 day timepoint and the 2 week timepoint. Applying gene ontology analysis to the set
of genes correlated at one timepoint but not the other identified annotations relating to chromatid
segregation, mitosis, and cell division at the 3 day timepoint (Fig. 4G), annotations relating to
the immune response (Fig. 4H), gliogenesis (Fig. 4I) and oligodendrocyte development (Fig. 4J)
at the 2 week timepoint. The degree to which oligodendrocyte progenitor cells (OPCs)
differentiate into oligodendrocytes following a focal gray matter injury is controversial (26),
however, we confirmed that both Sox4 and Sox10 localize to the region surrounding the injury,
indicating the presence of immature oligodendrocytes (Fig. S8). Thus, we conclude that cell
proliferation occurs on the order of days following injury, and that the period between 3 days and
2 weeks is critical for fate determination of newly born cells following traumatic brain injury.
We finally asked whether there might be other changes in gene expression in the
surrounding tissue that are not immediately localized to the injection site. Remarkably, we
noticed that several immediate early genes, including Fos, Arc, Npas4, and Junb, are upregulated
in a large region around the injection site at both the 3-day and the 2-week timepoints (27–29).
Running the overlap analysis on these genes at the 2-week timepoint revealed a number of other
genes that also localized near the wound, including Egr1, Egr4, Lmo4, Nr4a1, Slc16a13, Rgs4,
Grin2b, and C1ql3 (Fig. 4K). This increase in gene expression decreased with distance away
from the injection site and reached its half maximum 0.722 mm ± 0.191 mm away from the
injection site (mean ± sterr, N=4 measurements). The neural specificity of these genes leads us to
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
conclude that acute intracranial injury leads to alterations in the pattern of gene expression and
possibly activity in nearby neurons for weeks following the injury.
Slide-seq enables the spatial analysis of gene expression in frozen tissue with high spatial
resolution and easy scalability to large tissue volumes. The unbiased capture of transcripts
enabled facile integration with large-scale scRNA-seq datasets, and discovery of novel spatially
defined gene expression patterns in normal and diseased brain tissue. We anticipate that Slide-
seq will play important roles in positioning molecularly defined cell types in complex tissues,
and defining new molecular pathways involved in neuropathological states.
Author Contributions: F.C. and E.Z.M. conceived of the idea and supervised the work. S.G.R.
and R.R.S. developed the puck fabrication methods. S.G.R. developed the puck sequencing and
base-calling pipeline. S.G.R. and E.M. made the sequenced pucks. R.R.S. developed the tissue
processing and library preparation pipeline. S.G.R and R.R.S. performed experiments generating
data for the manuscript, with assistance from C.A.M., L.M.C., and C.V. A.G. developed the
NMFReg method and performed the associated analyses. J.W. performed some initial analyses
to help motivate using NMF for spatial mapping of cell types. S.G.R. developed the analysis
pipelines for identifying spatially non-random and significantly correlated genes. S.G.R., R.R.S.,
F.C., and E.Z.M. wrote the manuscript.
Acknowledgments: We would like to acknowledge Edward Boyden for his support. We
would like to thank Daniel Goodwin and Shahar Alon for useful discussions relating to the
identification of rRNA in single-cell sequencing data. We would like to thank Jamie Marshall
and Anna Greka for helpful discussions regarding kidney tissues. We would like to thank the
McCarroll lab for the use of their Zamboni Drop-seq analysis pipeline. We acknowledge Jon
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
Bloom for help with development of methods, and Velina Kozareva and Tushar Kamath for
assistance with algorithm implementation.
This work was supported by an NIH New Innovator Award (DP2 AG058488-01), an NIH
Early Independence Award (DP5, 1DP5OD024583), the Schmidt Fellows Program at the Broad
Institute and the Stanley Center for Psychiatric Research. S.G.R. acknowledges funding through
the Hertz Graduate Fellowship and the National Science Foundation Graduate Research
Fellowship Program (award #1122374).
Conflicts of Interest: S.G.R., R.R.S., C.A.M., F.C. and E.Z.M. are listed as inventors on
a patent application relating to the work.
1. S. Shah, E. Lubeck, W. Zhou, L. Cai, seqFISH Accurately Detects Transcripts in Single
Cells and Reveals Robust Spatial Organization in the Hippocampus. Neuron. 94, 752–
758.e1 (2017).
2. K. H. Chen, A. N. Boettiger, J. R. Moffitt, S. Wang, X. Zhuang, Spatially resolved, highly
multiplexed RNA profiling in single cells. Science. 348 (2015).
3. P. L. Stahl et al., Visualization and analysis of gene expression in tissue sections by
spatial transcriptomics. Science. 353, 78–82 (2016).
4. Materials and methods are available as supplementary materials online.
5. E. Z. Macosko et al., Highly parallel genome-wide expression profiling of individual cells
using nanoliter droplets. Cell. 161, 1202–1214 (2015).
6. K. B. Halpern et al., Single-cell spatial reconstruction reveals global division of labour in
the mammalian liver. Nature. 542, 352–356 (2017).
7. A. Saunders et al., Molecular Diversity and Specializations among the Cells of the Adult
Mouse Brain. Cell. 174, 1015–1030.e16 (2018).
8. L. O. Trussell, D. Oertel, (Springer, Cham, 2018; http://link.springer.com/10.1007/978-3-
319-71798-2_4), pp. 73–99.
9. M. T. Roberts, L. O. Trussell, Molecular Layer Inhibitory Interneurons Provide
Feedforward and Lateral Inhibition in the Dorsal Cochlear Nucleus. J. Neurophysiol. 104,
2462–2473 (2010).
10. E. S. Lein et al., Genome-wide atlas of gene expression in the adult mouse brain. Nature.
445, 168–176 (2007).
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
11. C. Gravel, R. Hawkes, Parasagittal organization of the rat cerebellar cortex: Direct
comparison of purkinje cell compartments and the organization of the spinocerebellar
projection. J. Comp. Neurol. 291, 79–102 (1990).
12. J. Xiao et al., Systematic regional variations in Purkinje cell spiking patterns. PLoS One.
9, e105633 (2014).
13. N. H. Barmack, Z. Qian, J. Yoshimura, Regional and cellular distribution of protein kinase
C in rat cerebellar Purkinje cells. J. Comp. Neurol. 427, 235–54 (2000).
14. N. L. Cerminara, E. J. Lang, R. V Sillitoe, R. Apps, Redefining the cerebellar cortex as an
assembly of non-uniform Purkinje cell microcircuits. Nat. Rev. | Neurosci. 16, 79 (2015).
15. A. Demilly, S. L. Reeber, S. A. Gebre, R. V. Sillitoe, Neurofilament Heavy Chain
Expression Reveals a Unique Parasagittal Stripe Topography in the Mouse Cerebellum.
The Cerebellum. 10, 409–421 (2011).
16. G. Brochu, L. Maler, R. Hawkes, Zebrin II: A polypeptide antigen expressed selectively
by purkinje cells reveals compartments in rat and fish cerebellum. J. Comp. Neurol. 291,
538–552 (1990).
17. J. R. Sarna, H. Marzban, M. Watanabe, R. Hawkes, Complementary stripes of
phospholipase Cβ3 and Cβ4 expression by Purkinje cell subsets in the mouse cerebellum.
J. Comp. Neurol. 496, 303–313 (2006).
18. D. Müller et al., Dlk1 Promotes a Fast Motor Neuron Biophysical Signature Required for
Peak Force Execution. Science. 343, 1264–1266 (2014).
19. H. Zhou et al., Cerebellar modules operate at different frequencies. 3, 2536 (2014).
20. C. J. Stoodley, J. D. Schmahmann, Evidence for topographic organization in the
cerebellum of motor control versus cognitive and affective processing. Cortex. 46, 831–44
(2010).
21. K. Maekawa, J. I. Simpson, “Climbing Fiber Responses Evoked in Vestibulocerebellum
of Rabbit From Visual System” (1972), (available at www.physiology.org/journal/jn).
22. D. R. W. Wylie, M. R. Brown, I. R. Winship, N. A. Crowder, K. G. Todd, Zonal
organization of the vestibulocerebellum in pigeons (Columba livia): III. Projections of the
translation zones of the ventral uvula and nodulus. J. Comp. Neurol. 465, 179–194 (2003).
23. C. Chung et al., Heat Shock Protein Beta-1 Modifies Anterior to Posterior Purkinje Cell
Vulnerability in a Mouse Model of Niemann-Pick Type C Disease. PLoS Genet. 12, 1–22
(2016).
24. R. A. Carter et al., A Single-Cell Transcriptional Atlas of the Developing Murine
Cerebellum. Curr. Biol. 28, 2910–2920.e2 (2018).
25. B. E. Clausen, C. Burkhardt, W. Reith, R. Renkawitz, I. Förster, Conditional gene
targeting in macrophages and granulocytes using LysMcre mice. Transgenic Res. 8, 265–
277 (1999).
26. K. L. Adams, V. Gallo, The diversity and disparity of the glial scar. Nat. Neurosci. (2017),
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
doi:10.1038/s41593-017-0033-9.
27. A. M. Kenney, J. D. Kocsis, Peripheral axotomy induces long-term c-Jun amino-terminal
kinase-1 activation and activator protein-1 binding activity by c-Jun and junD in adult rat
dorsal root ganglia In vivo. J. Neurosci. 18, 1318–28 (1998).
28. G. A. Robinson, Immediate early gene expression in axotomized and regenerating retinal
ganglion cells of the adult rat. Mol. Brain Res. 24, 43–54 (1994).
29. J. Honkaniemi, S. M. Sagar, I. Pyykönen, K. J. Hicks, F. R. Sharp, Focal brain injury
induces multiple immediate early genes encoding zinc finger transcription factors. Mol.
Brain Res. 28, 157–163 (1995).
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
Figure 1: High-resolution RNA capture from tissue by Slide-seq. (A) Left: Schematic of
array generation. A monolayer of randomly deposited, DNA barcoded beads (termed a “puck”)
is spatially indexed by SOLiD sequencing. Top Right: A representative puck with called
barcodes shown in black. Bottom Right: A composite image of the same puck colored by the
base calls for a single base of SOLiD sequencing. (Scale bar 500 μm) (B) Top Row: Schematic
of the sample preparation procedure developed for Slide-seq. Total time for library generation is
~3 hrs. Bottom Row: Schematic of a naïve analysis, in which each bead is clustered by its gene
expression, visualized in a tSNE two-dimensional embedding, and by locations in space. (C)
Spatial positions of Slide-seq beads, colored by clusters defined purely by gene expression
relationships amongst beads, across five tissue types (see Fig. S2 for tSNE embeddings and
definitions). (D) Characterization of lateral diffusion of signal on the Slide-seq surface. Top
Left: Digital image of a Slide-seq puck with bead color intensity scaled by total transcript counts.
Top Right: Image of the adjacent tissue section, stained with DAPI (scale bar 500 μm). Boxes
represent regions where an intensity profile was taken across CA1. Bottom left: Profile of pixel
intensity across CA1 in Slide-seq. Bottom right: Profile across CA1 in DAPI stained tissue. Red
dots represent locations of half max of the distribution. (E) Quantification of full width at half
maximum of profiles in (D), from both Slide-seq (red dots) and DAPI-stained tissue (blue dots)
(dotted line, mean; N = 10 profiles) (F) Log ratio of total number of quantified RNA transcripts
on a puck to the number of cells counted on a serially stained DAPI slice of equal area (dotted
line, mean) across five different tissues.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
Figure 2: Localization of cell types in cerebellum and hippocampus using Slide-seq. (A)
Schematic for assigning cell types from scRNA-seq datasets to Slide-seq beads using NMF and
NNLS regression (NMFreg). (B) Loadings of individual cell types, defined by scRNA-seq
cerebellum (7) on each bead (minimum 15 genes per bead) of one 3 mm-diameter coronal
cerebellar puck (red, cell type location, gray, Purkinje loadings plotted as a counterstain). (C)
Left: Number of cell types assigned per bead. A cell type was assigned to a bead if the factors
corresponding to that cell type made up at least 25% of bead loading (Fig. S3). Right: The
number of beads called as each atlas-defined cell type for cerebellar pucks. Error bars represent
standard deviation (N=7 pucks). (D) Projection of hippocampal volume with NMFreg cell type
calls for CA1 (green), CA2/3 (blue) and dentate gyrus (Red). Top left: Sagittal projection. Top
right: Coronal projection. Bottom left: Horizontal projection. Bottom right: axis orientations for
each of the projections. (E) Cell type calls of three representative sections from the dataset with
the position on the mediolateral axis denoted at the bottom of the image. (F) Top: Metagene
profiles on a sagittal hippocampus section representing cell subtypes. Bottom: Composite image
of all metagenes. All scale bars show 250 µm.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
Figure 3: Identification of novel variation in cerebellar gene expression by Slide-seq. (A) A
coronal cerebellar puck is shown, with Purkinje-assigned beads in white, choroid-assigned beads
in green, and beads expressing Ogfrl1 in magenta. Red arrow indicates cluster of Ogfrl1-positive
beads. (B) An Allen ISH atlas image of Ogfrl1, from a similar brain region. Red arrow indicates
Ogfrl1 expression in the cochlear nucleus. (C) A sagittal cerebellar puck showing counts of Pcp4
(gray), Rasgrf1 (blue), and a metagene consisting of Gprin3, Cemip, Mab21l2, and Syndig1l
(yellow). (D) Allen atlas images of Rasgrf1 (left) and Gprin3 (right). Arrows indicate point of
boundary of expression within the granular layer for each gene. (E) A heatmap illustrating the
separation of Purkinje-expressed genes into two clusters on the basis of the other genes they
correlate with spatially. The i,jth entry is the number of genes found to overlap with both gene i
and j in the Purkinje cluster (4). (F) For genes with significant expression (p<0.001) in the
nodulus-uvula region (4), the fraction of reads localized to the nodulus/uvula and to the VI/VII
boundary is shown. Kctd12 and Car7 did not pass the p-value cutoff, but are displayed as squares
to demonstrate their location relative to Aldoc. (G) An Aldoc metagene, consisting of Aldoc,
Kctd12, and Car7 is shown in cyan. A Cck metagene, consisting of Cck, Stmn4, Kcng4, and
Atp6ap1l is shown in red. (H) The H2-D1 metagene consisting of H2-D1, Cops7a, and Kmt2c is
shown in yellow. A Hspb1 metagene, consisting of Prkci and Hspb1, is shown in blue. (I) As in
(G), but only genes with significant expression both in the nodulus (p<0.05) and the VI/VII
boundary (p<0.05) are shown. (J) The Gnai1 metagene, consisting of Gnai1, Nefh, Plcb4, Rgs8,
Homer3, Scg2, Scn4b, and Gm14033 is shown in green, while the B3galt5 metagene, consisting
of B3galt5, Gdf10, Tmem248, Mpped2, and Dpf3 is shown in magenta. (K) Mybpc1, a gene
expressed in Bergmann glia, is shown in orange. (L) An Allen atlas image for Mybpc1. All scale
bars show 250 µm; Pcp4, a ubiquitous marker for Purkinje cells is shown in gray in (C), (H) and
(K).
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
Figure 4: Slide-seq identifies local transcriptional responses to injury: (A) Top: All mapped
beads plotted for a coronal slice from a mouse sacrificed 2 hours after injection. Circle radius is
proportional to the number of transcripts per bead up to a maximum of 500 transcripts. Bottom:
Three genes that mark the injection site. (B) As in (A), but for a mouse sacrificed 3 days after
injection. Top right shows a DAPI image of an adjacent slice, with injury indicated by a red
arrow and inset below. Middle right shows an expanded view of the injection site, with damage
primarily in layer 1. Panels with black backgrounds show cell types as called by NMFReg using
the hippocampal dataset, as density plots, scale bar: 250 µm. (4). Cluster 13 marks hippocampal
neurogenesis, and is interpreted here as marking mitotic cells. (C) As in (B), for a mouse
sacrificed 2 weeks following injection. Bottom, scale bar: 500 µm. (D) The spatial density
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;

!
!
!
!
!
!
!
!
profiles of microglia, mitotic cells, and astrocytes plotted for the 3-day puck in (B) (4). Lyz2 is
taken to represent macrophages. (E) The spatial density profiles of microglia, astrocytes, and
neurons for the 2-week puck in (C). (F) The average thickness of the features shown in (D) and
(E) is shown, in microns. Error bars show standard error (N=6 for scar, N=6 for penetration, N=3
for mitosis layer). (G-J) Gene ontology-derived metagenes are plotted for the indicated
transcriptional program, for two 3-day pucks (left) and two 2-week pucks (right). Red arrows
indicate injection site. (K) The IEG metagene (see text) is plotted for two 2-week pucks. Circular
images in (A), (B), and (C) refer to the 1mm scale bar in (A). All scale bars for images with blue
backgrounds are 500 µm.
.CC-BY-NC-ND 4.0 International licenseIt is made available under a
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint. http://dx.doi.org/10.1101/563395doi: bioRxiv preprint first posted online Feb. 28, 2019;