Abstract
Neutrophils are the most abundant leukocyte in humans and provide a
critical early line of defense as part of our innate immune system. We
perform a comprehensive, genome-wide assessment of the molecular
factors critical to proliferation, differentiation, and cell migration
in a neutrophil-like cell line. Through the development of multiple
migration screen strategies, we specifically probe directed
(chemotaxis), undirected (chemokinesis), and 3D amoeboid cell migration
in these fast-moving cells. We identify a role for mTORC1 signaling in
cell differentiation, which influences neutrophil abundance, survival,
and migratory behavior. Across our individual migration screens, we
identify genes involved in adhesion-dependent and adhesion-independent
cell migration, protein trafficking, and regulation of the actomyosin
cytoskeleton. This genome-wide screening strategy, therefore, provides
an invaluable approach to the study of neutrophils and provides a
resource that will inform future studies of cell migration in these and
other rapidly migrating cells.
Subject terms: Amoeboid migration, Chemotaxis, TOR signalling
__________________________________________________________________
Neutrophils provide a critical early defense as part of our innate
immune system. Here, authors performed a genome-wide assessment of the
molecular factors critical to proliferation, differentiation, and cell
migration in a neutrophil-like cell.
Introduction
Among the cells of our immune system, neutrophils are the most abundant
cell type and provide a vital early response in host defense by
migrating to sites of infection or tissue wounding^[34]1,[35]2.
Paramount to their success is an exquisite sensitivity to chemical
gradients, extremely rapid migration speeds on the order of
5–20 µm/min, and an ability to perform directed migration over long
distances and through a wide variety of distinct tissue
environments^[36]3–[37]5. Work in recent years has begun to reveal
neutrophils as a more heterogeneous cell type than previously
thought^[38]2,[39]6,[40]7, though the mechanisms that support
differentiation and phenotypic diversity remain incompletely
understood. Furthermore, relatively little is known about how specific
molecular players may change or adapt as the context and environment of
cell migration change.
The emergence of CRISPR-based gene perturbation approaches and robust
genome-wide targeted guide libraries now make it possible to perform
unbiased functional genomic screens in human cells^[41]8–[42]10. These
approaches offer significant technical improvements over past
strategies such as RNAi and offer the opportunity to more
comprehensively identify the genes involved in a biological
process^[43]11. However, the use of genome-wide CRISPR-based screens to
study complex and dynamic cellular processes has been more limited,
with only a few notable exceptions where complex enrichment methods
have been applied to identify factors important for phagocytosis and
for cell motility^[44]12–[45]15. The development of new functional
screening strategies is expected to provide new biological insights.
Our current understanding of cell migration has relied heavily on
videomicroscopy to assess behavior, but that is generally limited to
the study of tens or hundreds of individual cells^[46]16,[47]17. To
extend screening tools to perform a comprehensive screen of neutrophil
cell migration, there are several notable challenges. First,
genome-wide screens require millions of cells, demanding that the
assays to assess the relevant biological response be relatively simple
and easily scalable^[48]18. For example, current pooled CRISPR
strategies commonly use simple selections such as survival after a drug
treatment, or enrichments of a cell population with
fluorescence-activated cell sorting, to identify genetic perturbations
of interest. Second, the terminal differentiation status of neutrophils
and their short lifetime, on the order of days^[49]19, limit the use of
primary human cells and the possible time scale of individual migration
experiments.
In this work, we present the results of several pooled genome-wide
CRISPRi screens that provide a comprehensive, genome-wide look at the
molecular factors contributing to proliferation, neutrophil
differentiation and cell migration. Proliferation and differentiation
were assessed by performing pooled dropout assays^[50]18. Separate
migration screens were developed to assess directed migration
(chemotaxis), undirected migration (chemokinesis), and 3D amoeboid
migration through an extracellular matrix. We confirm known molecular
mechanisms contributing to cell proliferation and differentiation and
identify an unexpected role for mTORC1 signaling that alters
differentiation, survival, and cell migration. We also find a
near-perfect correlation between the genes important for chemotaxis and
chemokinesis, suggesting that both modes of migration are
mechanistically identical. Lastly, we use the results from our
different screens of cell migration to distinguish between
adhesion-dependent and adhesion-independent cell migration, ultimately
identifying several hundred genes that are important across these
different migratory contexts. This work demonstrates an invaluable
strategy to study cell migration and provides a resource that will
apply to future studies of migration in neutrophils and other rapidly
migrating cell types.
Results
Pooled CRISPRi screens identify genes that alter cell proliferation,
neutrophil differentiation, and cell migration
To identify novel regulators important for neutrophil biology, as well
as to facilitate our primary goal of identifying genetic factors
critical for cell migration, we used the immortalized HL-60 human
tissue culture cell line, derived from a patient with acute
promyelocytic leukemia^[51]19–[52]21. The proliferating,
undifferentiated HL-60 cells (uHL-60) can be induced to differentiate
into a neutrophil-like cell type (dHL-60) by treatment with the
signaling molecule all-trans retinoic acid (ATRA) or with the organic
solvent dimethylsufoxide (DMSO)^[53]22. After differentiation, the
chemotactic and migratory behaviors of dHL-60 cells closely mimic those
of primary neutrophils, and they are able to clear fungal infections in
neutropenic mice^[54]23–[55]25. We validated the efficiency of
knockdown in uHL-60 cells expressing dCas9-KRAB by targeting the CD4
gene, with immunofluorescence flow cytometry measurements demonstrating
robust knockdown using this construct (Fig. [56]1a). This dCas9-KRAB
construct, which includes a minimal-ubiquitous chromatin opening region
and proteolysis-resistant 80 amino acid XTEN linker^[57]26,[58]27,
provided substantially better efficacy over other constructs tested in
these cells (Supplementary Fig. [59]1).
Fig. 1. Genome-wide CRISPRi screens of proliferation, differentiation, and
cell migration.
[60]Fig. 1
[61]Open in a new tab
a Flow cytometry immunofluorescence shows near-complete loss of CD4
protein (blue) in uHL-60 cells expressing dCas9-KRAB, relative to
normal expression (orange) and an isotype control (gray, shaded). b
Schematic of pooled genome-wide CRISPRi dropout experiments of uHL-60
cell proliferation and differentiation into dHL-60 neutrophils.
Proliferation was assayed by comparing sgRNA abundances following six
days of growth (~24 hr doubling time) (set 2 versus set 1; 4
independent replicates). Differentiation was assayed by comparing sgRNA
abundance between dHL-60 neutrophils and uHL-60 cells (set 3 versus set
1; 8 independent replicates). c Schematic of pooled CRISPRi cell
migration assays. Migration of dHL-60 cells were assayed across three
experiments: chemotaxis (serum gradient), chemokinesis (uniform serum
stimulation), and 3D amoeboid migration in an extracellular matrix (see
“Methods”). For quantification, sgRNA abundance in both migratory
fractions (sets 4i and 5i) and remaining cells (sets 4ii and 5ii) were
compared to our initial dHL-60 library (set 3). Membrane, pores and
cells drawn to scale. d Error bars represent mean values +/− SD of the
migratory fraction across independent experiments (3D amoeboid: 6
replicates; chemokinesis 2 hr and 6 hr: 4 replicates each; chemotaxis
2 hr: 4 replicates; chemotaxis 6 hr: 16 replicates). For 3D amoeboid
experiments, the migratory fraction of cells was collected from the
fibrin layer. e Volcano plots across the screens of proliferation,
differentiation, and cell migration. Data points represent the average
log[2] fold-change from three sgRNAs per gene across independent
experiments (4 replicate screens for proliferation, 8 replicate screens
for differentiation, and 20 migration screens). Cell migration values
represent an average across all migration assays. Controls were
generated by randomly selecting groups of three control sgRNAs.
P-values were calculated using a one-sided permutation test, adjusted
for multiple comparisons using the Benjamini–Hochberg procedure (dashed
line: p = 0.05). f Screen overlap. The number of significant genes are
identified in the left horizontal bar plot using an adjusted p value
cutoff of 0.05, while the intersection of genes across screens is shown
in the vertical bar plot (dot diagram identifies the specific
intersection).
We used a pooled genome-wide CRISPRi library (3 sgRNA per gene^[62]10)
to perform dropout-type assays of proliferation and differentiation
(Fig. [63]1b). For proliferation, we quantified changes in sgRNA
abundance in uHL-60 cells following six days of growth, as compared to
day zero. We identified 2,127 genes that disrupted growth and only 56
genes that enhanced growth (Supplementary Data [64]1). Our results were
well-correlated with those reported by Sanson et al. in HT29 and A375
cell lines using this CRISPRi library (Supplementary Fig. [65]2a). To
derive dHL-60 cells, we induced differentiation by incubating uHL-60
cells with 1.57% DMSO for five days, which provides a near-complete
differentiation of the cell population into CD11b+ neutrophil-like
cells^[66]28. Viable dHL-60 cells were isolated using density gradient
centrifugation to remove dead cells and cellular debris following our
differentiation protocol. We compared sgRNA abundance between dHL-60
and uHL-60 cells to identify gene perturbations that altered the
abundance of cells during differentiation and, therefore, could serve
as indicators of altered differentiation. Here, we identified 989 genes
that were depleted and 869 genes that were enriched relative to our
control sgRNAs (Supplementary Data [67]2). The ratio of enriched sgRNAs
to depleted sgRNAs was strikingly different from our proliferation
screen, where only ~2% of knockdowns led to an enrichment of sgRNAs.
We assessed cell migration using three different experimental
paradigms. For the first two paradigms, chemotaxis and chemokinesis, we
performed scaled-up transwell migration assays (see “Methods”), which
mimic migration through tight cellular junctions during transmigration
across an endothelial layer^[68]16 (Fig. [69]1c, left panel). Here,
cells were added to the top reservoir above a track-etch membrane with
3 µm diameter pores. To assay these two modes of cell migration, we
manipulated the distribution of heat-inactivated fetal bovine serum
(hiFBS), a general stimulant for migration^[70]29. Chemotaxis, or
directed migration, was assayed by including 10% hiFBS only in the
bottom reservoir, resulting in a chemoattractant gradient toward the
bottom reservoir. We separately assayed chemokinesis^[71]30,[72]31,
referring to stimulated migration absent of any directional cue, by
providing a uniform 10% hiFBS environment in both reservoirs. In both
the chemotaxis and chemokinesis assays, cells were collected following
periods of two and six hours (Fig. [73]1d). To assess migratory
success, we separately collected both the migratory cells that made it
to the bottom reservoir and the cells that remained above the
track-etch membrane. To identify significant gene perturbations,
normalized log[2] fold-change values were calculated by comparing sgRNA
abundances in these cell pools relative to a reference pool of dHL-60
cells (see “Methods” for further details).
Our final migration assay focused on probing amoeboid three-dimensional
(3D) migration by embedding cells in a synthetic extracellular matrix
(ECM), more representative of migration through the intercellular
spaces in tissue^[74]32. Cells were embedded at the bottom of a thin
layer (~200 µm) of collagen ECM that they would need to traverse to
reach a second layer of fibrin ECM where they could be recovered
(Fig. [75]1c, right panel). To a first approximation, we expect cells
to perform a random walk, whose mean squared displacement will scale
linearly with time^[76]33. Migration through this complex environment
will therefore require substantially more time compared to migration
through the thin track-etch membranes and we therefore only considered
a longer, nine-hour period prior to cell collection in these
experiments (Fig. [77]1d). The most migratory dHL-60 cells were
collected by degrading the upper fibrin layer using the enzyme
nattokinase, which has protease activity specific to fibrin^[78]34. We
also collected the cells still in collagen, and calculated normalized
log[2] fold-change by comparing sgRNA abundances in these cell
subpopulations relative to a reference population.
Across our entire set of cell migration assays, we identified 344 genes
that reduced the fraction of migratory cells and 31 genes that
increased this fraction, relative to migration of control sgRNAs
(Fig. [79]1e). The results of the pooled CRISPRi screens therefore
revealed a comprehensive set of genes that play crucial roles in cell
proliferation, neutrophil differentiation, and cell migration. We found
that nearly half of the genes identified in each screen of
proliferation and differentiation were unique to those screens, along
with a substantial fraction of genes that affect both of these
processes (Fig. [80]1f). Importantly, a significant number of genes
were identified as unique regulators in either differentiation or cell
migration, or both. In the sections that follow we characterize the
phenotypic changes observed following genetic perturbation across these
subsets of genes.
CRISPRi screens identify genes important for differentiation into migratory
neutrophils
Stimulating the differentiation of leukemia cells using pharmacological
agents remains a key strategy for the clinical treatment of acute
promyelocytic leukemia^[81]35. We were therefore interested in
identifying what gene perturbations influenced the differentiation of
uHL-60 cells into migratory, terminally differentiated
(non-proliferating) dHL-60 cells. Thus, we began by performing gene set
enrichment analysis (GSEA)^[82]36 to identify the pathways and gene
categories that were overrepresented among the differentiation screen
data (Fig. [83]2a). We found a positive enrichment across various
metabolic processes, particularly genes involved in oxidative
phosphorylation (electron transport chain, mitochondrial protein
synthesis). We also found a depletion of key regulatory genes
associated with granulopoiesis including the transcription factors
Fli-1 (FLI1) and the CCAAT/enhancer binding proteins, C/EBPα (CEBPA)
and C/EBPε (CEBPE), suggesting that our screen data is identifying
genes specific to neutrophil differentiation.
Fig. 2. Identification of genes and pathways important for neutrophil
differentiation.
[84]Fig. 2
[85]Open in a new tab
a Pathways enriched in our CRISPRi differentiation screen (dHL-60 cells
relative to uHL-60 cells). Pathways that were associated with genes
whose knockdown predominantly led to an enrichment of target sgRNAs in
the dHL-60 cell population are identified in blue, while those that
decreased in abundance are in red. P-values estimate the statistical
significance of gene set enrichment, calculated using a one-sided
permutation test and adjusted for multiple comparisons using the
Benjamini–Hochberg procedure. b Comparison of log[2] fold-changes
across the CRISPRi screens of proliferation, differentiation, and cell
migration. Several gene sets identified through our pathway enrichment
analysis and several known regulators of neutrophil differentiation are
identified. The cell migration data points represent normalized log[2]
fold-change values, calculated by averaging across all individual
migration screen replicates. c Brightfield microscopy of uHL-60 CRISPRi
knockdown lines targeting ATIC and a control sgRNA. Control uHL-60
cells exhibit an expected round morphology, while sgRNA targeting of
ATIC resulted in many cells that exhibited a migratory capability
(white arrows). Images are representative of acquisitions across three
fields of view. d Schematic of mTORC1/mTORC2 signaling pathway, color
coded by signed statistical significance values (log[10] p[adj] value)
from the differentiation screen results. Blue indicates gene targets
whose sgRNA were enriched in the dHL-60 cells, while red indicates
those that were depleted.
To further distinguish genes whose knockdown specifically affected
differentiation per se from those that generally perturbed basic
cellular processes, we also plotted the differentiation screen’s log[2]
fold-change values against both our screens of proliferation and the
migratory data, averaged across all migration screens (Fig. [86]2b).
PU.1 (SPI1), another transcription factor that is highly expressed in
neutrophils^[87]37,[88]38, was identified across all screens and was
among the strongest perturbations to migration (normalized log[2]
fold-change of −1.2, Fig. [89]2b right panel). Strikingly, knockdown of
genes associated with oxidative phosphorylation and mitochondrial
translation showed systematic effects in all three screens, with sgRNAs
associated with this process mostly enriched following differentiation,
while mostly depleted in the screens for both proliferation and
migration (Fig. [90]2b, red and green points). In contrast, sgRNAs
targeting genes associated with mammalian target of rapamycin (mTOR)
signaling were enriched following cell differentiation and depleted in
the migration assays, but knockdown of these genes had no consequence
on proliferation (Fig. [91]2b, yellow points).
One unexpected gene identified in both our differentiation and
migration screens was ATIC (Fig. [92]2b), which codes for an enzyme
that acts on the adenosine monophosphate analog AICAR, an intermediate
in the generation of inosine monophosphate in the purine biosynthesis
pathway^[93]39. We constructed a stable cell line expressing a sgRNA
targeting ATIC and found a substantial number of polarized migratory
cells prior to induction of differentiation, a notable phenotype that
is not observed in unperturbed uHL-60 cells (Fig. [94]2c and
Supplementary Fig. [95]2b). AICAR is capable of stimulating
AMP-dependent protein kinase (AMPK) activity^[96]40,[97]41 and it is
possible that ATIC knockdown changes the basal concentration of AICAR,
which may alter the cell’s metabolic and energy state in a way that
drives differentiation. Consistent with these observations, RNA
transcriptome analysis showed that these cells had a transcriptional
profile more similar to that of dHL-60 cells expressing a control sgRNA
than to control uHL-60 cells (Supplementary Fig. [98]2d). More work
will be needed to understand the broader impact of this gene
perturbation on cell migration.
Considering our differentiation screen results more broadly, we found
that many of the identified genes were also enriched in recent
genome-scale efforts to characterize mouse embryonic stem cell
differentiation^[99]42,[100]43. Notably, of the roughly 500 genes
reported as important for exit from pluripotency in mouse embryonic
stem cells, half were present in our differentiation data set (adjusted
p value <0.01 when applying gene set enrichment analysis, Supplementary
Fig. [101]2b). While embryonic stem cells and uHL-60 cells represent
distinct developmental cell stages, this commonality suggests shared
processes that may be important as mammalian cells change their
proliferative status and undergo state transitions in differentiation.
Disruption of folliculin and Ragulator-Rag signaling pathways potentiate
survival of dHL-60 neutrophils
Given the enrichment of sgRNAs associated with mTOR signaling during
both differentiation and migration, we wanted to further understand the
consequence of these gene perturbations. In humans, the protein kinase
mTOR is a component of two distinct complexes, mTORC1, and mTORC2.
While mTORC2 has previously been implicated in cell migration and
chemotaxis^[102]44–[103]48, we were surprised to find many genes
associated with mTORC1 signaling in our screens of differentiation and
migration (Fig. [104]2d). mTORC1 coordinates cell growth through its
activity on the surface of lysosomes by mediating cellular changes in
translation regulation, metabolism, and autophagy^[105]49. The genes
most enriched in the differentiation screen were directly upstream of
mTORC1 and included the Rag guanosine triphosphatase (GTPase) A/B:C/D
heterodimer, which recruits mTORC1 to the lysosomal membrane via
binding to the Ragulator complex (LAMTOR1-5)^[106]50. Knockdown of
folliculin (FLCN), a GTPase-activating protein that targets RagC/D and
promotes an active state of the Rag heterodimer^[107]51, led to a
similar enrichment of dHL-60 cells in the differentiation screen and
depletion in the migration screen.
To characterize how these gene knockdowns might affect differentiation,
we generated individual stable cell lines expressing mTORC1-related
sgRNAs targeting LAMTOR1, FLCN, TSC1, and also a cell line expressing
an sgRNA targeting RICTOR, a key subunit of the mTORC2 complex^[108]47.
Following the initiation of neutrophil differentiation, all knockdown
cell lines showed a higher cell density compared to a control sgRNA
cell line after comparable time periods, consistent with their
overrepresentation in our differentiation screen results
(Fig. [109]3a). Normally, cell densities stopped increasing by about 4
days after initial DMSO-induced differentiation and began to decline,
presumably due to apoptosis of the terminally differentiated dHL-60s
(Fig. [110]3b). Interestingly, LAMTOR1 and FLCN knockdown lines showed
a distinct increase in cell lifetime (Fig. [111]2e, pink and purple
symbols). To further confirm a functional role for mTORC1, we treated
these two knockdown lines with 10 nM and 100 nM rapamycin, dosages
expected to abolish mTORC1 kinase activity^[112]52. Rapamycin treatment
resulted in a decrease in the survival of these cells following
differentiation, restoring their survival characteristics to the lower
levels associated with the sgRNA control cell line (Fig. [113]3c).
Fig. 3. Characterization of cellular growth and survival following knockdown
of mTOR-related genes.
Fig. 3
[114]Open in a new tab
a Differentiation screen results were confirmed across a number of
sgRNA targets. Cell density was monitored at 5-days following the
initiation of neutrophil differentiation. The dashed lines represent
the values obtained for dHL-60 cells with a control sgRNA. Error bars
represent mean values +/− SD across independent experiments (8
experiments for screen and 3 experiments for cell density
measurements). b Change in dHL-60 cell density following
differentiation. Cell density measurements were normalized to day 4,
following initiation of cell differentiation of uHL-60 cells. Error
bars represent mean values +/− SD across 3 independent experiments.
Note that media was replenished every three days, with cell density
measurements corrected to account for changes in media volume and
evaporation. c Rapamycin treatment in dHL-60 CRISPRi knockdown lines
targeting FLCN, LAMTOR1, and a control sgRNA. Treatment of dHL-60 cells
was begun on day 4 post-initiation of differentiation. Cells were
either untreated (solid line), or treated with rapamycin at 10 nM
rapamycin (dashed lines) and 100 nM rapamycin (dash-dot lines). Error
bars represent mean values +/− SD across 3 independent experiments.
We further assessed the role of mTORC1 on differentiation by
quantifying key molecular markers of neutrophils. Here we used flow
cytometry to measure the induced surface expression of CD11b, an early
differentiation marker also known as integrin α[M] (ITGAM), and the
fMLF receptor (FPR1) that recognizes chemoattractant N-formylated
peptides^[115]53. Both markers showed little to no expression in uHL-60
cells but were strongly induced in our dHL-60 cells (Fig. [116]4a). As
positive controls for disruption of neutrophil differentiation, we
constructed stable cell lines expressing sgRNAs targeting the two
essential differentiation genes, SPI1 and CEBPE. To assess induction of
differentiation markers, we used principal component analysis to
quantify the axis associated with co-induction of the two surface
markers ITGAM and FPR1 (principal mode 1, Fig. [117]4b). As expected,
sgRNAs targeting SPI1 and CEBPE showed reduced induction of
differentiation markers. In contrast, we found that stable cell lines
expressing sgRNAs targeting LAMTOR1 and FLCN exhibited higher induction
of the differentiation markers as compared to controls (Fig. [118]4c
and Supplementary Fig. [119]3a). These results show that cells
expressing these sgRNAs are still undergoing terminal differentiation
but with altered survival characteristics.
Fig. 4. Knockdown of FLCN and LAMTOR1 alters differentiation trajectory and
results in cells with poorer chemotactic sensitivity.
Fig. 4
[120]Open in a new tab
a Flow cytometry immunofluorescence measurements of CD11b (ITGAM) and
fMLP receptor 1 (FPR1) cell surface expression in uHL-60 and dHL-60
cells. b The two-dimensional heatmap shows the induced expression of
CD11b and fMLP receptor 1 in dHL-60 cells. The axis associated with
induction of these surface markers were identified by applying
principal components analysis. Measurements using isotype control
antibodies are shown in gray. c The first principal component
identified in (b) was used to compare changes in expression induction
in different gene knockdown lines. Black lines indicate the 99%
confidence interval for the log expression mean along the first
principle component, calculated by bootstrapping across single-cell
flow cytometry measurements. A two-sided Mann-Whitney U test was
applied to the bootstrapped log expression values from each knockdown
cell line and the control cell line (***p < 0.001). d Transcriptional
changes following knockdown of FLCN, LAMTOR1, and SPI1 were assayed by
RNA-seq pre-differentiation (undiff.), 1-day, 5-day, and 7-day
post-differentiation. Dimensionality reduction using UMAP was applied
to transcription data and pseudo-plotted using a spline to show
temporal trajectory. Individual data points represent an average across
6 independent RNA-seq samples. e The acute chemotaxis response of
dHL-60 cells was assayed by photo-uncaging fMLP during migration of
agarose-confined cells on BSA passivated coverslips. Average
instantaneous speed (i), angular bias (ii), and the directed speed
(projected speed along direction of fMLP gradient) (iii) are shown.
Data points indicate mean across 5 independent experiments (~3500 cells
per cell line, per experiment), with the shaded regions showing the
distribution across measurements. A two-sided Mann–Whitney U-test
indicated a significant difference in angular bias (*p = 0.03) and
directed speed (**p = 0.008).
Disruption of folliculin and Ragulator-Rag signaling pathways results in
altered but active mTORC1 signaling in dHL-60 neutrophils
In order to identify the specific changes in mTOR signaling associated
with knockdown of LAMTOR1 and FLCN, we began by checking for altered
phosphorylation of known targets of mTORC1 and mTORC2 protein kinase
activity. Western blot analysis of ribosomal S6 kinase and Akt, which
are well-characterized targets of mTORC1 and mTORC2, respectively,
showed that kinase activity of mTORC1 and mTORC2 remained active
following knockdown of LAMTOR1 and FLCN (Supplementary Fig. [121]3b,
c). It is also known that mTORC1 can regulate the activity of
transcription factors TFEB and TFE3, which depend more specifically
on the activity of RagC/D^[122]49,[123]54. Notably, knockdown of TFEB
and TFE3 resulted in a modest but significant decrease in sgRNA
abundance in our differentiation screen (Fig. [124]2b). We, therefore,
turned to whole-transcriptome sequencing (RNA-seq) to identify global
transcriptional changes that might provide more insight into the
changes in mTORC signaling associated with knockdown of LAMTOR1 and
FLCN.
Focusing on the stable cell lines expressing sgRNAs targeting LAMTOR1,
FLCN, and SPI1, we performed RNA-seq on uHL-60 cells, and dHL-60 cells
at days 1, 5, and 7 following the initiation of differentiation. Using
dimensionality reduction (UMAP^[125]55) to take a broad look at the
entire data set, we found similar changes in transcription in our
LAMTOR1 and FLCN knockdown lines, which were distinct from both the
line expressing our control sgRNA and the line expressing SPI1 sgRNA
(Fig. [126]3d). Also consistent with our screen results, the
transcriptional profiles across our knockdown lines were fairly similar
to the controls in undifferentiated cells and one day after induction
of differentiation, only diverging later in the differentiation
process, suggesting that these gene perturbations are specifically
involved in differentiation and neutrophil function.
To better understand why the FLCN and LAMTOR1 knockdown cells enjoyed a
prolonged lifespan, we delved more deeply into the gene expression
data. Relative to control cells, we found enrichment for genes
associated with lysosomes, autophagy, and transcription of ribosomal
genes (Supplementary Fig. [127]3d, e). Importantly, these match the
reported roles of TFEB and TFE3 as master regulators of lysosomal
biogenesis and autophagy^[128]49, further confirming that altered
mTORC1 signaling is indeed along the RagC/D-FLCN axis. We found a
substantial decrease in expression of the autophagy-activating kinases
ULK1 and ULK2, and increased expression of the anti-apoptotic gene BCL2
(Supplementary Fig. [129]3e), which may support the extended survival
of these cells. Direct inhibition of autophagy has been shown to affect
both neutrophil differentiation and effector function^[130]56,[131]57
and our data suggest that FLCN and LAMTOR1 knockdown play a similar
inhibitory role through altered mTORC1 activity. These knockdown lines
also showed changes in the expression of genes associated with
neutrophil degranulation, including an increase in mpo and a decrease
in mmp9 (Supplementary Fig. [132]3d and Supplementary Data [133]3 and
[134]4). This is also observed following the inhibition of autophagy
during differentiation and may be a reflection of incomplete
differentiation^[135]57.
Intriguingly, with respect to differentiation, our characterization of
cell surface markers showed that the FLCN and LAMTOR1 knockdown lines
expressed normal or slightly elevated levels of fMLF receptor as
compared to controls (Fig. [136]4c and Supplementary Fig. [137]3a). We
were therefore interested in whether these cells maintained their
sensitivity to fMLF as a chemotactic agent, and explored this further
in our FLCN knockdown line. Here we used photo-activation of caged fMLF
to generate spatial gradients of the small chemotactic peptide, using a
standard assay where migratory cells are sandwiched between a
BSA-coated coverslip and an agarose overlay to minimize requirements
for adhesion^[138]30. While FLCN knockdown cells migrated with similar
speeds as our control sgRNA line in this context (Fig. [139]3h (i)),
knockdown cells were much less responsive to fMLF. Across cells, we
observed a reduced average angular bias of about 10°, compared to 15°
for our control sgRNA line (Fig. [140]3h (ii)). This also resulted in a
reduced directed speed (i.e. projected speed along the spatial fMLF
gradient) (Fig. [141]3h (iii)), showing that FLCN knockdown cells were
not as responsive and migratory toward the fMLF gradient.
In our differentiated FLCN and LAMTOR knockdown lines, we also observed
an increase in expression of several genes associated with macrophage
differentiation, including ccr5, cd163, cd64, cd71. uHL-60 cells are
multipotent cells that can be differentiated into other cell types,
including macrophages^[142]58. Given the observed induction of CD11b
and fMLF receptor in dHL-60 FLCN and LAMTOR1 knockdown lines
(Fig. [143]4c and Supplementary Fig. [144]3a) and that these cells are
longer lived, we reasoned that FLCN and LAMTOR1 knockdown might be
skewing their differentiation trajectory away from a purely
neutrophil-like character and towards a more macrophage-like state. To
test this possibility, we mined an available RNA-seq dataset that
measured gene expression changes during differentiation of uHL-60 cells
into both neutrophil-like cells and macrophage-like cells^[145]38, we
identified a variety of genes including cd52 that show higher
differential expression in macrophage-like cells (Supplementary
Table [146]1). Using flow cytometry, we found that both FLCN and
LAMTOR1 knockdown cell lines exhibited higher expression of CD52 than
control lines after DMSO-triggered differentiation, consistent with
this hypothesis (Supplementary Fig. [147]4). While further work is
needed to fully dissect the possibility that FLCN and LAMTOR1 knockdown
alters the trajectory of cell fate, it does help to explain the
observed pattern of enrichment of dHL-60 cells in our differentiation
screen.
Overall, we found that disruption of mTORC1-related genes alters (but
does not eliminate) mTORC1 signaling. This results in an apparent
inhibition of autophagy and reduction in apoptotic signals that extend
cell survival, increasing their abundance within the population, but
also perturbs their differentiation into fully chemotactic neutrophils.
Migration through track-etch membranes is dominated by genes associated with
cell adhesion
To identify genes important for cell migration through the narrow pores
of track-etch membranes, we compared the results of our screens for
chemokinesis (with serum present in both top and bottom reservoirs) and
chemotaxis (with serum present in the bottom reservoir only). The gene
perturbations most detrimental to migration in both our chemotaxis and
chemokinesis screens were associated with inside-outside ɑ[M]β[2]
integrin signaling (Fig. [148]5a). Integrins facilitate cell-substrate
binding and indeed, when examined migration on a fibronectin-coated
coverslip of a stable knockdown line expressing a sgRNA targeting
ITGB2, cells exhibited a polarized morphology but were only loosely
adherent (Fig. [149]5b, Supplementary Fig. [150]4a, and Supplementary
Movie [151]1). We also directly measured the adhesion phenotype by
performing an adhesion assay where cells were allowed to adhere to the
surface of a plastic culture dish. We found a substantial reduction in
the fraction of adherent cells in our ITGB2 knockdown line relative to
our control cell line (Supplementary Fig. [152]5c).
Fig. 5. Cell migration CRISPRi screen identifies genes important for adhesion
and migration on 2D surfaces.
[153]Fig. 5
[154]Open in a new tab
a Left, components of inside-out
[MATH: α :MATH]
[M]
[MATH: β :MATH]
[2] integrin signaling. Right, normalized log[2] fold-changes values
for most significant sgRNA in the chemotaxis and chemokinesis screens.
Error bars represent mean values +/− SEM across n = 28 measurements
from 14 independent experiments. The gray shaded region shows the
histogram of control sgRNAs. b Representative phase images of cell
migration on fibronectin-coated coverslips (ITGB2 sgRNA and control
cells). Three fields of view were collected for each cell line. c
Representative phase images of cell migration on fibronectin-coated
coverslips (FLCN sgRNA, LAMTOR1 sgRNA, and control cells). Two fields
of view were collected for each cell line. d Characterization of cell
migration phenotypes. Speed was calculated by tracking cell nuclei
during migration on fibronectin-coated coverslips. Persistence was
inferred from the cell velocity data as described by Metzner et al.
(see “Methods”). Measurements represent experiments performed over 2–3
days, acquired across 32 (sgControl), 10 (sgFLCN), and 14 (sgLAMTOR1)
fields of view. Differences were identified using a two-sided
Mann–Whitney U-test (***p < 0.001). e Comparison of normalized log[2]
fold-changes across the pooled CRISPRi cell migration screens of
chemotaxis and chemokinesis. ITGB2, FERMT3 and TLN1 genes are
identified in red. f Comparison between chemotaxis screen normalized
log[2] fold-changes and measurements of migration fraction of
individual knockdown lines exposed to a serum gradient with 10% hiFBS,
for 2 h. The gray data point and dashed lines represent the values
obtained for control cells. Error bars represent mean values +/− SEM
across independent experiments (8 screen replicates and 4 replicates
for stable knockdown lines). g Characterization of cell migration speed
following knockdown of GIT2 from nuclei tracking during migration on
fibronectin-coated coverslips. Measurements represent experiments
performed over 3 days, acquired across 29 fields of view. The data was
compared using a two-sided Mann–Whitney U-test (**p < 0.01). For (d–g),
individual data points represent average values for cells across a
single field of view, with the shaded regions showing the distribution
of all measurements.
Interestingly, earlier we noted that our LAMTOR1 and FLCN knockdown
lines had slightly elevated levels of CD11b (integrin ɑ[M])
(Fig. [155]4c), suggesting that these cells may also exhibit an altered
adhesion phenotype. Indeed, when placed on a fibronectin-coated
coverslip, these cells appeared to make more extensive contact with the
substrate and often lacked a normal front-back polarity (Fig. [156]5c
and Supplementary Movie [157]2). We performed tracking of cell nuclei
as cells migrated on fibronectin-coated coverslips for 30 min. Here we
found that the LAMTOR1 and FLCN knockdown cells moved at about half the
speed of our control sgRNA line, with an average speed of 0.13 µm/s as
compared to 0.21 µm/s (Fig. [158]5d, top). This effect was almost as
detrimental as directly knocking down ITGB2 (Fig. [159]5d, top). These
observations are consistent with previous results in other cell types
indicating that there is an optimum degree of cell-substrate adhesion
for efficient migration, with either increasing or decreasing adhesion
causing decreased cell speed^[160]59–[161]62.
Efficient directional migration for motile cells depends on directional
persistence as well as cell speed. We employed a Bayesian inference
algorithm based on a model for a heterogeneous random walk^[162]63 to
calculate a migratory persistence metric for each cell. In this model,
a persistence value of zero corresponds to a non-persistent, diffusive
movement. Persistence values closer to −1 indicate anti-persistent,
reversive movement, while values closer to +1 indicate more persistent
migration. In line with the reduced front-back polarity, FLCN and
LAMTOR1 knockdown lines showed a reduction in persistence
(Fig. [163]5d, bottom) compared to control cells. These results
highlight the importance of adhesion as a key prerequisite to entry and
migration through the pores of the track-etch membrane.
Chemotaxis and chemokinesis migration screens are strongly correlated
One of the most notable observations from the chemotaxis and
chemokinesis datasets was a strong correlation between their normalized
log[2] fold-change values (Fig. [164]5e, left panel;
[MATH: ρ :MATH]
= 0.99). This suggests there are no distinct molecular pathways that
neutrophils require for directed migration as opposed to random
migration, in the context of serum stimulation.
In order to confirm the measured log[2] fold-change values for cells
reaching the lower reservoir in the large-scale screen with a more
direct measurement of transmigration, we assayed the fraction of cells
that migrate across the track-etch membrane in individual knockdown
cell lines. We generated individual lines expressing one of the sgRNAs,
chosen to span the range of observed log[2] fold-change values from the
screen (ITGB2, APBB1IP, TLN1, VPS29, ARHGAP30, FMNL1, ATIC, GIT2).
Quantifying the fraction of cells that migrated through the track-etch
membrane after two hours with 10% hiFBS added to the bottom reservoir,
we found a strong correlation with the log[2] fold-change values from
our chemotaxis screen (Fig. [165]5f,
[MATH: ρ :MATH]
= 0.87). More specifically, the strongest perturbation, targeting
knockdown of ITGB2, led to only 6% of the cells in the bottom reservoir
versus 30% of the cells with our control cell line. Knockdown of GIT2
showed the largest positive increase, with 34% of the cells collected
in the bottom reservoir.
Since most gene knockdowns decreased the fraction of cells migrating
through the track-etch membrane, we were intrigued by the subset of
genes that exhibited a positive log[2] fold-change; that is, those
whose knockdown enhanced cell migration. Among the most positively
enriched genes was GIT2 (Fig. [166]5e), which encodes a protein that
binds to the p21-activated kinase-interacting guanine nucleotide
exchange factors α-PIX (ARHGEF6) and β-PIX (ARHGEF7). α-PIX and β-PIX
both enhance the activity of the Rho GTPases Cdc42 and
Rac1^[167]64,[168]65 that act as master regulators to enhance actin
assembly at the leading edge of motile cells^[169]66. Along with GIT2,
knockdown of α-PIX and another α-PIX binding partner, PPM1F, also
exhibited positive enrichment in our migration screens (Fig. [170]5e).
To better understand how knockdown of GIT2 influenced cell migration,
we directly examined the motility behavior of our stable cell line with
a sgRNA targeting GIT2. Analyzing cell tracks as cells migrated on
fibronectin-coated coverslips, we find that GIT2 knockdown cells
migrated with an average speed of 0.25 µm/s, or about 20% faster than
our control cell line (Fig. [171]5g, left). Migration otherwise
appeared similar to control cells, exhibiting similar migratory
persistence (Supplementary Fig. [172]5d).
In sum, the most substantial gene perturbations across our chemotaxis
and chemokinesis screens are those that impact cell–substrate adhesion
or play a role as regulatory components governing the behavior of the
actomyosin cytoskeleton.
CRISPRi screen identifies genes important for 3D amoeboid cell migration
To explore how 3D amoeboid migration differs from 2D migration in our
track-etch membrane assays, we began by comparing our 3D amoeboid
screen with the chemokinesis screen results. We find that many of the
adhesion-related genes did not exhibit strong phenotypes in the 3D
migration assay (Fig. [173]6a). This observation is consistent with the
expectation for integrin-independent migration of cells embedded in
fibrous ECM^[174]67. Interestingly, knockdown of talin 1 (TLN1), which
mediates the linkage between integrins and the actin cytoskeleton,
still inhibited migration into the upper fibrin layer of the 3D
amoeboid screen. This suggests additional roles for talin 1 beyond the
interaction between the actin cytoskeleton and ɑ[M]β[2] integrins in
dHL-60 cells.
Fig. 6. Cell migration CRISPRi screen identifies genes important for 3D
amoeboid migration.
[175]Fig. 6
[176]Open in a new tab
a Comparison of normalized log[2] fold-changes across the pooled
CRISPRi cell migration screens of 3D amoeboid migration and
chemokinesis. b Comparison of 3D migration normalized log[2]
fold-changes with measurements of cell speed and migratory persistence
via single-cell nuclei tracking. Cells from individual sgRNA knockdown
lines were tracked during migration in collagen for 60 min (1 min frame
rate). The median cell speed (left) and inferred migratory persistence
(right, see “Methods”) are plotted against their measured normalized
log[2] fold-change from the pooled screen. Error bars represent mean
values +/− SEM across independent experiments (6 screen replicates and
4 for experiments using stable knockdown lines). Individual cell line
measurements represent experiments performed over 2–5 days, acquired
across 10 (sgControl), 7 (sgFLMN1), 7 (sgCORO1A), and 4 (sgITGB2)
fields of view. c, d show immunofluorescence localization of FMNL1 and
CORO1A in amoeboid-migrating cells in collagen. F-actin was labeled by
phalloidin, while DNA was stained by DAPI. Left images are maximum
projection composite images; right images show grayscale localization
of formin-like 1 (c) and coronin 1A (d). Red arrows indicate the
approximate direction of cell migration based on cell shape and a more
intense phalloidin intensity expected at the cell front.
Of the many interesting candidates, we chose to further characterize
two actin regulatory proteins identified in the 3D amoeboid screen,
formin-like 1 (FMNL1) and coronin 1A (CORO1A). These proteins have
previously been implicated in cell migration^[177]68–[178]70, but their
role during 3D migration remains less well-characterized. We generated
stable cell lines with sgRNA targeting each of these genes and
quantified migratory speed and persistence as cells migrated in 3D
collagen gels. As expected, knockdown of ITGB2 showed no effect on
speed or persistence in 3D, but migration speed was decreased in the
FMNL1 and CORO1A knockdown lines, consistent with the results of our 3D
screen (Fig. [179]6b, left panel,
[MATH: ρ :MATH]
= 0.83). For both knockdown lines, cells migrated with an average
speed of approximately 0.07 µm/s, roughly half as fast as our control
sgRNA or ITGB2 knockdown lines, which have average speeds of
0.11–0.12 µm/s.
While knockdown of either FMNL1 or CORO1A resulted in reduced speed,
only FMNL1 knockdown showed a significant reduction in migratory
persistence (Fig. [180]5b, right panel). This may reflect different
roles during 3D migration. We used immunofluorescence to determine the
localization of these proteins in wild-type cells. In contrast to the
expected leading edge localization of well-characterized formins like
mDia1/2^[181]71, FMNL1 was rear-localized and often directly behind the
nucleus (Fig. [182]6c and Supplementary Fig. [183]6a). This is
consistent with recent work in T cells^[184]72, who found similar
localization and hypothesized that formin-like 1 may support actin
polymerization to aid in squeezing the nucleus through tight
endothelial barriers. Indeed, knockdown of FMNL1 was also identified in
our track-etch membrane-based screens and may more specifically support
movement as cells squeeze through small pores. In the 3D context,
formin-like 1 may help to support more persistent movement as cells
move through the complex fibrous network.
In contrast to formin-like 1, we find coronin 1A predominantly
colocalized with the lamellipodial filamentous actin structures at the
cell front, though more diffuse protein localization was also observed
at the cell rear (Fig. [185]6d and Supplementary Fig. [186]6b). Coronin
1A localization to the lamellipodial projections is consistent with
prior characterization of cells migrating on a 2D
surface^[187]70,[188]73,[189]74, while the rear-localized protein may
relate to a role in actin turnover and disassembly^[190]68. Here, the
lack of change in migratory persistence following CORO1A knockdown may
relate to a more general disruption of actin cytoskeleton dynamics,
rather than alterations to how cells move through the collagen ECM,
though further work will be needed to clarify this point.
CRISPRi screens of cell migration provide a rich resource for studying
rapidly migrating cells
Beyond the genes noted thus far across our migration screens, we
identified a variety of additional genes with expected roles in
actomyosin-based migration. For example, we identified the β subunit of
the filamentous-actin capping protein CapZ (CAPZB), which is known to
cap actin filaments at their barbed ends^[191]75 and the adenylyl
cyclase-associated protein 1 (CAP1), a regulatory protein which
facilitates cofilin-driven actin filament turnover and may also
interact with talin 1^[192]76. With respect to myosin contraction at
the cell rear, knockdown of RhoA, a key regulator of myosin activity,
significantly perturbed migration across all migration screens. Among
modulators of Rho GTPases, knockdown of ARHGAP30 was among the most
significant perturbations in our migration screens and has been
reported to negatively regulate activity of RhoA and Rac1 by enhancing
GTP hydrolysis^[193]77,[194]78.
Our screens also identified many other genes that have less obvious
roles in cell migration (complete gene lists provided in Supplementary
Data [195]5–[196]7). To explore our migration data sets more broadly,
we applied pathway enrichment analysis to the results of our cell
migration screens (Fig. [197]7a). Combining this with our exploration
of the data thus far, in Fig. [198]7b we provide a summary of genes
identified across our cell migration assays.
Fig. 7. Summary of pathways and genes identified across cell migration
CRISPRi screens.
[199]Fig. 7
[200]Open in a new tab
a Pathways enriched in cell migration screens. Since the majority of
gene knockdowns lead to poorer migratory phenotypes (i.e., negative
log[2] fold-changes), disruption of the noted pathways are associated
with poorer migratory success. Due to the correlation across the
chemotaxis and chemokinesis screens, their data was combined in this
analysis (green). Pathways enriched in the 3D amoeboid migration screen
are shown in yellow. p values estimate the statistical significance of
gene set enrichment, calculated using a one-sided permutation test and
adjusted for multiple comparisons using the Benjamini–Hochberg
procedure. b Summary of the genes identified across the cell migration
screens. Genes were identified from the collated
chemotaxis/chemokinesis screens (green) and 3D amoeboid screen
(yellow), with the shading intensity indicating the false discovery
rate threshold that each gene fell into (adjusted p < 0.05 or <0.3).
Empty column entries (i.e. white entries) indicate that the gene was
not identified as significant. Genes associated with transcription,
translation, and gene regulation, or genes that perturbed the processes
of either proliferation or differentiation with absolute log[2]
fold-changes values larger than 0.7 were excluded.
Differential sensitivity to protein trafficking machinery and integrin
expression is observed across all cell migration assays
Among the candidate genes not obviously associated with cytoskeletal
function was VPS29 (Fig. [201]5e, f), a component of the retromer and
retriever complexes. These complexes recycle transmembrane proteins
from endosomes back to the trans-Golgi network and the plasma membrane,
respectively^[202]79,[203]80. Upon further examination, additional
subunits of both the retromer and retriever complexes were identified
as significant hits in our migration screens (Fig. [204]8a, left).
Among other proteins involved in protein trafficking, we also
identified components of the HOPS and CORVET complexes (Fig. [205]8a,
right), which are specifically involved in endosomal–lysosomal protein
trafficking^[206]81 and may also influence mTORC1 signaling^[207]49.
Fig. 8. 2D and 3D migration show context specific sensitivities to integrin
expression and recycling.
[208]Fig. 8
[209]Open in a new tab
a Summary of normalized log[2] fold-changes of the most significant
sgRNA for protein trafficking genes in the chemotaxis and chemokinesis
screens (green), and the 3D amoeboid screen (yellow). Error bars
represent mean values +/− SEM (green: n = 28 measurements from 14
independent experiments; yellow: n = 7 measurements from 6 independent
experiments). Shared gene products for the protein complexes (retromer/
retriever and HOPS/CORVET) are indicated by a solid black line. The
histograms and shaded region identify the distribution of the control
sgRNAs. b Immunofluorescence flow cytometry of CD11b (ITGAM gene; left
column) and CD18 (ITGB2 gene; right column). Histograms show surface
distribution in control sgRNA (black, solid), ITGB2 (blue, solid),
SNX17 (blue, dashed), VPS29 (blue, dash-dot). Shaded histograms
indicate cellular autofluorescence from a non-targeting isotype control
antibody. c Summary of normalized log[2] fold-changes of the most
significant sgRNA for integrin genes in the chemotaxis and chemokinesis
screens (green), and the amoeboid screen (yellow). Integrin genes whose
transcription is not detected in HL-60 cells^[210]28 were excluded.
Error bars represent mean values +/− SEM (green: n = 28 measurements
from 14 independent experiments; yellow: n = 7 measurements from 6
independent experiments). The histogram and shaded region identify the
distribution of the control sgRNAs. d Characterization of cell
migration phenotypes in integrin knockdown lines. Speed was calculated
by tracking cell nuclei during migration in a collagen ECM. Persistence
was inferred from the cell velocity data as described by Metzner et al.
(see “Methods”). Individual data points represent mean values for cells
across a single field of view, with the shaded regions showing the
distribution of all measurements. Measurements represent experiments
performed over 2–3 days, acquired across 10 (sgControl), 4 (sgITGB2),
and 5 (sgITGA1) fields of view. A two-sided Mann-Whitney U test found
the persistence of the ITGA1 knockdown line differed from control cells
(*p = 0.002).
We considered the hypothesis that the perturbations affecting protein
trafficking might be altering integrin recycling and
degradation^[211]82. In further support of this, sorting nexin 17
(SNX17) was also identified in our chemotaxis and chemokinesis screens.
This protein binds to β integrins in conjunction with the retriever
complex, recycling integrins back to the plasma membrane^[212]83. To
test whether integrin expression was altered when genes associated with
membrane recycling were knocked down, we generated additional cell
lines with sgRNAs targeting VPS29 and SNX17 and measured cell surface
expression of ɑ[M]β[2] integrins (CD11b and CD18, for integrin ɑ[M] and
β[2], respectively) using flow cytometry. As a positive control, we
found that the ITGB2 knockdown cell line showed a near complete loss of
β[2] integrin expression, as expected (Fig. [213]8b, right). Only
assembled heterodimer ɑβ integrin pairs are expected to be stably
localized at the cell surface^[214]84, and consistent with this, these
cells also exhibit a near-complete loss of integrin ɑ[M] (Fig. [215]8b,
left). VPS29 and SNX17 knockdown lines had a moderate drop in ɑ[M]β[2]
integrin expression relative to our control sgRNA (Fig. [216]8b), with
measurable decreases in the expression levels of both subunits. These
findings show that the migration defects associated with perturbations
to the membrane recycling pathway may be due, at least in part, to
disruptions in integrin surface presentation.
Interestingly, although 3D amoeboid migration was insensitive to
knockdown of ɑ[M]β[2] integrin (Fig. [217]6b), several genes among the
protein trafficking complexes, including VPS29, still disrupted cell
migration when knocked down in our 3D migration screen (Fig. [218]8a).
Although disruption of these protein complexes may have pervasive
effects beyond altering integrin expression^[219]28, we wanted to
explore the role of integrins more comprehensively across our cell
migration screens. Notably, in our 3D amoeboid screen, sgRNA targeting
ITGA1 led to a significant perturbation to cell migration
(Fig. [220]8c). Integrin heterodimers containing the integrin ɑ[1]
subunit are known to bind collagen^[221]85, which was the primary ECM
component in this screen. We validated this finding by constructing an
individual cell line with sgRNA targeting ITGA1 and tracked its
migration in 3D. In contrast to ITGB2 knockdown, which only affected 2D
migration, our line targeting ITGA1 showed a modest reduction in cell
speed and a significant drop in cellular persistence during 3D
migration (Fig. [222]8d). In summary, we find that protein trafficking,
and related alterations in integrin cell-surface expression, can
substantially alter behavior of neutrophils across both 2D and 3D cell
migration contexts.
Discussion
The rapid migratory characteristics of neutrophils and their early role
in our innate immune response to infection or wounds make them an
important cell type to consider in the context of cell migration. In
this work, we have demonstrated the use of pooled genome-scale gene
perturbations using CRISPRi gene knockdown to study proliferation,
differentiation, and cell migration in human HL-60 neutrophils. This
was made possible through the additional development of scalable assays
that effectively separate cells based on their migratory capabilities.
With roughly 10^11 neutrophils produced in the bone marrow each
day^[223]1, any process that impacts neutrophil abundance will severely
influence their protective capabilities. As such, we chose to identify
genes relevant to differentiation by looking at changes in cell
abundance between differentiated and undifferentiated cells following
gene knockdown. While some of the results from our screen of
differentiation may reflect specific genetic sensitivities of the HL-60
leukemia cell line and our differentiation protocol, we were encouraged
by the identification of key transcriptional regulatory genes,
including CEBPA, CEBPE, and SPI1, which are known to be involved in
neutrophil differentiation. Beyond these genes, we found that
perturbations along the FLCN-RagC/D signaling axis of mTORC1 signaling
impacts differentiation and survival, and also dramatically altered
cells’ migratory phenotypes. We hypothesize that these effects proceed
via a non-canonical mechanism of mTORC1 regulation that has only
recently begun to be elucidated^[224]49,[225]54.
Intriguingly, we found substantial overlap of enriched genes when
comparing our results on differentiation of HL-60 cells with those from
a differentiation screen for exit from pluripotency in mouse embryonic
stem cells^[226]42,[227]43. Pathways regulating stem cell renewal and
differentiation share substantial overlap with stress-response
pathways^[228]86 and our observations suggest these processes continue
to be important as mammalian cells change their proliferative status
during myeloid differentiation. In the context of myeloid cells,
terminal cell differentiation would provide a useful strategy by the
host to minimize the propagation of damaged or cancerous cells, and our
results appear particularly relevant to differentiation therapy as a
treatment for patients with acute promyelocytic leukemia^[229]35. While
the mechanism behind DMSO-mediated differentiation of myeloid
precursors remains poorly understood^[230]19, DMSO can alter membrane
permeability and impair mitochondrial function^[231]87,[232]88 and may
act as a chemical insult that drives cell differentiation by impairing
cellular homeostasis. Mitochondria, in particular, are hubs of
metabolic signaling that produce molecules that modulate cellular
function, gene expression, and can alter differentiation
state^[233]89,[234]90.
Among our cell migration screens, the most striking difference between
the track-etch membrane transmigration screens for chemotaxis and
chemokinesis as compared to the screens for efficient migration in 3D
extracellular matrices was the importance of cell-substrate adhesion.
In particular, genes associated with inside-out
[MATH: α :MATH]
[M]
[MATH: β :MATH]
[2] integrin signaling dramatically reduced migration success in our
screens using track-etch membranes. These results are especially
relevant to the family of leukocyte adhesion deficiency (LAD)
disorders, where neutrophils are unable to effectively extravasate from
blood vessels into tissue and mount an immune response following
infection^[235]91. Notably, both ITGB2 and FERMT3 are hits in our
screens that are also single-gene defects causing the LAD1 and LAD3
subtypes of this disorder, respectively^[236]92,[237]93. The majority
of these adhesion-related genes were not important for 3D amoeboid
migration, highlighting the value in using multiple assays to probe
cellular function in different, but related contexts. This is also
consistent with prior work showing that this type of cell migration is
largely integrin-independent^[238]32,[239]67. However, our 3D screen
did allow us to identify a specific role for integrin ɑ[1] (ITGA1),
where we found that knockdown led to cells with poorer persistence when
monitoring single cells migrating in a 3D collagen matrix.
We observed a strong, indeed a near-perfect, correlation between the
log[2] fold-change values between chemotaxis and chemokinesis
track-etch membrane experiments. As a first attempt to compare these
two processes genome-wide, it was surprising to find a lack of any
notable enrichment for any genes between the two assays, suggesting
that there are no major genetic differences or distinct molecular
pathways required for directed as opposed to random neutrophil
migration. This was true even though our chemotaxis assay resulted in
substantially more cells migrating to the bottom reservoir than the
chemokinesis assay over equivalent time frames. This result speaks to
the ability of neutrophils, as well as more disparate motile cell types
like fish keratocytes^[240]94, to spontaneously polarize their
migratory machinery in the absence of any asymmetric spatial
cue^[241]29,[242]95,[243]96. Chemotaxis, at least in the context of
serum stimulation as explored here, likely acts to more efficiently
guide the movement of cells through the pores of the track-etch
membrane, but our results indicate that the underlying molecular
mechanisms of directed and spontaneous cell motility are essentially
identical. For more specific chemoattractants like fMLF, we would still
expect an increase in migratory activity during both chemokinesis and
chemotaxis^[244]30, and we hypothesize that the similarity between
chemotaxis and chemokinesis may be more broadly applicable.
In summary, our data provides a valuable resource for future study of
proliferation, differentiation, and context-dependent cell migration of
rapidly migrating neutrophils. Further experimental adjustments may
provide additional insights into cell migration. For example, in our 3D
amoeboid screen, changes could be made to the ECM composition and
density^[245]97 or alternative spatial gradients could be
implemented^[246]98,[247]99. These alternative experimental paradigms
could be used to yield new insights into other modes of cell migration
like durotaxis (gradients in ECM rigidity)^[248]100, haptotaxis
(gradients in substrate composition)^[249]101, or galvanotaxis
(directional response to electrical cues)^[250]102,[251]103.
Methods
Cell culture and neutrophil differentiation
Undifferentiated HL-60 cells (uHL-60) were a generous gift from the lab
of Dr. Orion Weiner. These cells were cultured and differentiated into
neutrophil-like cells (dHL-60) as previously
described^[252]22,[253]104. Briefly, cells were maintained at 37 °C and
5% CO2, cultured in RPMI 1640 medium containing L-glutamine and 25 mM
HEPES 1640 (Gibco #22400089) supplemented with 10% heat-inactivated
fetal bovine serum (hiFBS) (Gemini Bio Products #900–108), and 100 U/mL
penicillin, 10 μg/mL streptomycin, and 0.25 μg/mL Amphotericin B (Gibco
#15240). Differentiated HL-60 cells (dHL-60) were generated by
incubating cells in media containing 1.57% Dimethyl Sulfoxide (DMSO,
Sigma, #D2650). Here, confluent cells (approx. 1–1.5 × 10^6 cells/mL)
were diluted by adding two volumes of additional media and DMSO. The
culture media was replenished with fresh media, including DMSO, three
days following the initiation of differentiation. Except for where
noted otherwise, dHL-60 cells were used in cell migration assays five
days following the initiation of differentiation. Experiments involving
rapamycin treatment used rapamycin (Thermo Scientific Chemicals
#AAJ62473MF).
CRISPRi pooled library construction
All genomic integrations involved lentiviral transduction. uHL-60 cells
expressing dCas9-KRAB were first generated and this cell line was used
for all subsequent work (sgRNA genome-wide libraries and individual
sgRNA targeting cell lines).
To achieve reliable gene knockdown in uHL-60 cells, we used dCas9-KRAB
linked by a proteolysis-resistant 80 amino acid XTEN
linker^[254]26,[255]27, driven by an EF1ɑ promoter that was placed
downstream of a minimal-ubiquitous chromatin opening (UCOE) element to
prevent gene silencing^[256]26,[257]105. The dCas9-KRAB construct was
based on a construct originally gifted by Dr. Marco Jost and Dr.
Jonathan Weissman, but modified to include blasticidin resistance
(pHR-UCOE-Ef1a-dCas9-HA-2xNLS-XTEN80-KRAB-P2A-Bls).
The sgRNA library was previously reported in Sanson et al. (Dolcetto
CRISPRi library set A, Addgene #92385). This library contains 57,050
sgRNA, with 3 sgRNA per gene target and 500 non-targeting control
sgRNA. For optimal library design, sgRNA were selected based on their
position relative to annotated transcription start sites, expected
on-target activity, and the presence of off-target matches.
For large-scale lentivirus production (dCas9 construct or pooled sgRNA
library), 15 μg transfer plasmid, 18.5 μg psPAX2 (Addgene #12260), and
1.85 μg pMD2.G (Addgene #12259) were diluted in 3.5 ml Opti-MEM I
reduced-serum media (Gibco #31985070) and then combined with 109 μL
TransIT-Lenti Transfection Reagent (Mirus, MIR6600). Following a
10 minute incubation, this mixture was added dropwise to confluent
HEK-293T cells (ATCC, CRL-3216) in a T175 flask containing 35 mL DMEM
media (Gibco #11965-092) and supplemented with 1 mM sodium pyruvate
(Gibco #11360-070). Lentivirus was recovered by collecting media 48 hr
later, with centrifugation at 500 × g for 10 min to remove any residual
cells and debris. For our dCas9-KRAB construct, we additionally
concentrated the lentivirus approximately 60-fold using Lenti-X
Concentrator (Takara Bio Inc., #631231).
For small-scale lentivirus production (individual sgRNAs), lentivirus
was prepared in 6-well tissue culture plates. Here 1 μg sgRNA transfer
plasmid, 1 μg psPAX2, and 0.1 μg pMD2.G were diluted in 200 μL Opti-MEM
I reduced-serum media and combined with 6 ul transIT. Following a
10 minute incubation, this mixture was added dropwise to confluent
HEK-293T cells, with lentivirus collected as noted above.
Construction of individual sgRNA plasmids for CRISPRi
Individual sgRNA plasmids were constructed for the generation of stable
CRISPRi knockdown cell lines using sgRNA identified from the
genome-wide CRISPRi screens. These were cloned into the same base
vector pXPR_050 plasmid (Addgene #96925) as the pooled library, as
previously described^[258]10. Briefly, pXPR_050 was first linearized
using the restriction enzyme BsmBI (New England Biolabs, #R0739S, which
includes NEB buffer 3.1). Here, 20 µg of pXPR_050, 20 µL NEB buffer
3.1, and 10 µL BsmBI were combined for a 200 µL reaction and incubated
for 5 hours at 55°C. The resulting linear pXPR_050 DNA was gel
extracted using the QIAquick gel extraction kit (Qiagen, #28704) and
resuspended in TE buffer (10 mM Tris·Cl, pH 8.0; 1 mM EDTA) to a
concentration of 10 ng/µL. sgRNA inserted were generated by annealing
complementary oligonucleotides with DNA overhangs compatible for
ligation with the BsmBI-digested pXPR_050 DNA. The oligonucleotides
were purchased from Integrated DNA Technologies (Coralville, IA) as
described below and annealed by combining 1.5 µL of each forward and
reverse oligonucleotide (stock concentration of 50 µM in water), 5 µL
NEB buffer 3.1, and 42 µL water. The mixture was first incubated for
5 min at 95 °C and then allowed to cool by lowering the temperature by
5 °C every 5 min until the sample was at room temperature. Finally to
ligate the sgRNA insert into the pXPR_050 vector, 1 µL of annealed
sgRNA insert was combined with 20 ng of the BsmBI-digested pXPR_050 DNA
and ligated using T4 ligase (New England Biolabs, #M0202S). The ligated
DNA product was transformed into NEB Stable Competent E. coli (New
England Biolabs, #C3040H) following the manufacturer directions and
successfully inserted sgRNA were identified by Sanger sequencing
(performed by Genewiz from Azenta Life Sciences; Burlington, MA).
Forward oligonucleotides were ordered as 5’ CACCG (20 bp sgRNA target
sequence)3’, while reverse complement oligonucleotides were ordered as
5’ AAAC (20 bp reverse complement sgRNA target sequence)C 3’. The
forward 20 bp sgRNA target sequences used for individual CRISPRI
knockdown cell lines are listed below.
Control sgRNA: AGGGCACCCGGTTCATACGCNGG;
GIT1 sgRNA: GGCGGCGCTTCCGCTCTAACNGG
FMNL1 sgRNA: GCCCCGTCCGTGGGACCGGGNGG
TSC1 sgRNA: GACTGTGAGGTAAACAGCTGNGG
ATIC sgRNA: CTGGGTTCAGGGCGAGCGGGNGG
RICTOR sgRNA: CGGGCTTACCTCGTACTCGGNGG
ITGA1 sgRNA: CGTGTTTAGGCTAAAGTCCANGG
APBBIIP sgRNA: CCTTAGTCCCTCTTGCGTCGNGG
CORO1A sgRNA: ATCTTCAGCGGGCGAGTCCCNGG
VPS29 sgRNA: CGACGGTGGTGGTGACTGAGNGG
SNX17 sgRNA: TGCGGGGACTCGCTGAGCAGNGG
ITGB2 sgRNA: CGGTGTGCTGGAGTCCTCGGNGG
CEBPE sgRNA: GTAGGCGGAGAGGTCAATGGNGG
SPI1 sgRNA: CCCAGGGCTCCTGTAGCTCANGG
ARHGAP30 sgRNA: CAGGACACAATTTCTTGCCANGG
FLCN sgRNA: GCCCGGGTTCAGGCTCTCAGNGG
TLN1 sgRNA: GGGCGACCCGAGAAGCGGCGNGG
LAMTOR1 sgRNA: GCTGCTGTAGCAGCACCCCANGG.
CRISPRi cell line construction
The dCa9-KRAB and sgRNAs constructs were integrated into uHL-60 cells
using a lentivirus spinoculation protocol. Briefly, lentivirus was
added to 1 mL cells (1 × 10^6 cells/mL) and polybrene reagent (final
concentration of 1 μg/mL) in 24-well tissue culture plates. Cells were
spun at 1000 × g for 2 h at 33 °C. Virus was removed and cells were
placed in an incubator for 2 days prior to antibiotic selection for 6
days (dCas9-KRAB: blasticidin 10 μg/mL; sgRNA constructs: puromycin
1 μg/mL).
For CRISPRi sgRNA library preparations, lentiviral titers were
estimated by titrating lentivirus over a range of volumes (0 μL, 75 μL,
150 μL, 300 μL, 500 μL, and 800 μL) with 1 × 10^6 cells in a total of
1 mL per well of a 24-well tissue culture plate, using the
spinoculation protocol noted above. Two days post-transduction, cells
were split into two groups, with one placed under puromycin selection.
After 5 days, cells were counted for viability. A viral dose that led
to a 12.5% transduction efficiency was used for subsequent pooled
library work. This low efficiency was targeted to ensure most cells
only received one sgRNA integration^[259]10. For library work, roughly
230 million cells were transduced, targeting a final number of roughly
30 million successful sgRNA integrations following puromycin selection
(or about 500 cells per sgRNA). Cells were maintained across multiple
T175 tissue culture flasks with 35 mL of media.
Knockdown was confirmed in the ITGB2 knockdown line by flow cytometry
immunofluorescence, in the FLCN, LAMTOR1, SPI1, and ATIC knockdown
lines by RNA-seq, and in the FMNL1 knockdown line by immunofluorescence
microscopy.
Genome-wide CRISPRi assays
Overview of cell collection and experimental replicates
For CRISPRi proliferation drop-out screens, genomic DNA (gDNA) was
collected from uHL-60 cells at two time points, separated by 6 days of
proliferation in T175 tissue culture flasks. Each cell preparation was
pelleted by centrifugation and frozen for later genomic DNA isolation.
Results from the proliferation screen represent averages across four
independently prepared genome-wide CRISPRi libraries.
For CRISPRi differentiation drop-out screens, gDNA was collected in
uHL-60 and in dHL-60 cells 5 days following the initiation of
differentiation in 15 cm tissue culture dishes. Results from
differentiation screens represent averages across four independently
prepared genome-wide CRISPRi libraries, each performed twice (8
experiments total).
For cell migration experiments, gDNA was collected from three
populations: On the day of each experiment, 5-day differentiated dHL-60
cells were collected, with 3×10^7 cells set aside as a reference sample
(3 in Fig. [260]1b). For chemotaxis and chemokinesis experiments
involving track-etch membranes, the other two populations were the
fraction of cells that migrated through the membrane (4ii in
Fig. [261]1c) and the fraction of cells that remained on top of the
membrane (4i in Fig. [262]1b). For the amoeboid 3D screen, the other
two populations were the fraction of cells that migrated into the
fibrin (5ii in Fig. [263]1c) and the fraction of cells that remained in
collagen (5i in Fig. [264]1c).
Regarding experimental replicates, for assays using track-etch
membranes, 6-h time point chemotaxis experiments were performed across
four independently prepared genome-wide CRISPRi libraries, with eight
experiments in total. For 6-h chemokinesis experiments and all 2-h time
points (both, chemotaxis and chemokinesis), experiments were performed
using two independently prepared genome-wide CRISPRi libraries (two
experiments each). For amoeboid experiments, results represent the
average across two independently prepared genome-wide CRISPRi
libraries, each performed three times (six experiments total).
Chemotaxis and chemokinesis cell migration experiments with track-etch
membranes compared the number of cells that migrated through the pores
(4ii in Fig. [265]1c) with respect to the reference sample, and those
that did not (4i in Fig. [266]1b) with respect to the reference sample.
For the 3D amoeboid migration screen we examined the fraction of cells
that migrated into the fibrin (5ii in Fig. [267]1c) with respect to the
reference sample, and those that remained in the collagen (5i in
Fig. [268]1c) with respect to the reference sample. This resulted in
two separate measurements per migration experiment, except for several
samples where the sgRNA did not PCR amplify properly from the gDNA and
was therefore not sequenced.
Removal of cell debris and dead cells prior to cell migration assays
Cellular debris and dead cells were removed from the dHL-60 cell
suspensions using density gradient centrifugation. Pooled CRISPRi
libraries were differentiated in 15 cm dishes (55 ml cell culture per
dish) and eight plates were combined for a single preparation. Cells
were first spun down (10 min at 300 × g), resuspended in 10 mL
PolymorphPrep (Cosmo Bio USA #AXS1114683) and added to the bottom of a
50 mL conical tube. Using a transfer pipette, 15 mL of 3:1
PolymorphPrep: RPMI media + 10% hiFBS was gently layered on top by
dispensing along the walls of the tube. This was followed by layering
another 14 mL of RPMI media + 10% hiFBS. Cells were centrifuged at
700 × g for 30 min with the centrifuge acceleration and brake set to
half-speed. Live dHL-60 cells were collected between the RPMI media and
the 3:1 PolymorphPrep. RPMI media layers were diluted with one volume
of RPMI media + 10% hiFBS, and spun down once more for 10 min at
300 × g. Finally, cells were resuspended in 10 mL RPMI media and
counted using a BD Accuri C6 flow cytometer (live cells identified by
their forward scatter and side scatter).
Cell migration assays: chemotaxis and chemokinesis using track-etch membranes
Chemotaxis and chemokinesis transwell migration assays^[269]106 used
track-etch membranes with 3 µm pore sizes (6-well plates with 24.5 mm
diameter inserts; Corning, #3414). For each experiment, 24 million
cells were distributed across four 6-well plates. For each track-etch
membrane insert, one million cells were diluted in 1.5 mL media and
added to the top of the track-etch membrane. Note that for chemotaxis
experiments, cells were purified and resuspended in RPMI media without
hiFBS. To the bottom reservoir, 2.6 mL RPMI media with 10% hiFBS was
added and the plates were carefully moved to a 37 °C incubator.
Following incubation for the required time (2 and 6 h time points), the
track-etch inserts were separated from the bottom reservoir. To ensure
more complete recovery of the migratory cells, the bottom side of the
track-etch membrane was gently scraped using a cell-scraper (Celltreat,
#229310) to dislodge any cells remaining on the membrane surface. The
migratory cells (bottom reservoir) and remaining cells (top reservoir)
were separately collected by centrifugation. Following an additional
wash with 1 mL PBS, cells were pelleted and frozen at −80 °C for later
genomic DNA extraction.
Note that in a second set of chemotaxis experiments (6-h time point;
four of the replicates), custom devices were fabricated to house larger
49 mm track-etch membranes (3 µm pore size, Sigma #TSTP04700). Similar
cell densities were targeted as the experiment using multi-well plates
above.
Cell migration assays: amoeboid 3D using collagen and fibrin ECM
Cells were seeded into a multi-layer system of collagen (rat-tail
collagen used throughout; ThermoFisher # A1048301) and fibrin as shown
in Fig. [270]1c. Briefly, these were prepared by first creating a
~50 μm layer of collagen on top of 25 mm glass coverslips, seeding
dHL-60 cells in another layer of collagen ~150 μm thick, and then
overlaying this with fibrin ECM. Each genome-wide CRISPRi screen
involved our pooled dHL-60 library spread across 32 coverslips with
1 × 10^6 cells added to each coverslip.
The glass coverslips were first surface modified to support an adhered
layer of collagen ECM using a silane treatment^[271]107. A 2%
aminosilane solution was first prepared in 95% ethanol/5% water and
incubated for 5 min to allow silanol formation. Coverslips were then
immersed in the solution for 10 min, rinsed with 100% ethanol, and then
cured on a hot plate heated to 110 °C for 5–10 min. The coverslips were
then immersed in 0.25% glutaraldehyde for 15 min and then rinsed in
water for 5 min. This 5 min wash was repeated two more times. We then
prepared the initial collagen layer by pipetting 22.6 μL of a 1.5 mg/mL
collagen mixture (for 3 mL: 1.5 mL 3 mg/mL collagen, 376 μL 0.1 M NaOH,
210 μL 10× PBS, and 923 μL PBS) onto a 15 cm plastic tissue culture
dish and then placing a coverslip on top, causing the mixture to spread
across the entire coverslip. Approximately 16 coverslips were prepared
inside a single 15 cm tissue culture dish. The dish was placed in an
incubator at 37 °C for 18 min to gel.
Coverslips containing the initial layer of collagen were carefully
removed using tweezers and flipped collagen side up to allow dHL-60
cells to be seeded. Here, a 1 mg/mL collagen mixture was prepared and
mixed with dHL-60 cells (recipe for 1.2 ml: 400 μL 3 mg/mL collagen,
100 μL 0.1 M NaOH, 55 μL 10x PBS, 525 μL PBS, and 120 μL hiFBS). For
each collagen treated coverslip, 3 × 11.3 μL aliquots were pipetted on
top of the initial collagen layer, which wetted and spread across the
initial collagen layer. Note that the initial coverslip and collagen
layer will begin to dry out while inside a biosafety cabinet due to the
air flow, so this second collagen mixture must be added relatively
quickly to ensure proper spreading of this second layer. The coverslips
were again moved to an incubator at 37 °C for 15 min to gel, placed
inside a closed tissue culture dish containing a wetted kimwipe to
minimize evaporation. Note that during this time, prior to gelling,
cells will settle down to the initial collagen layer.
The final layer, composed of fibrin ECM^[272]108, was then prepared on
top of the collagen. Here, a 1 mg/mL fibrin ECM was generated by mixing
1 μL thrombin (100 U/ml, Sigma-Aldrich #T1063-250UN) per 1 mL
fibrinogen at 1.5 mg/mL (plasminogen-Depleted from human plasma,
Sigma-Aldrich #341578) in 1× Hanks’ buffered salt solution (HBSS;
ThermoFisher Scientific, Gibco #14-065-056). This was carefully added
on top of the collagen by slowly pipetting near the edge of the
coverslip. The fibrin ECM will begin to gel immediately, but the
coverslips were further incubated at 37 °C for 18 min to complete the
gelling process. Finally, the multi-layered gels were covered with RPMI
media containing 10% hiFBS and incubated for 9 h for dHL-60 cells to
migrate.
Recovery of the most migratory dHL-60 cells from the fibrin ECM were
obtained by first incubating coverslip/gels with nattokinase, which
specifically degrades fibrin^[273]34. Here, the RPMI media was first
aspirated from the 15 cm dish and 25 ml PBS containing 2.1 mg/mL
nattokinase (Japan Bio Science Laboratory USA inc.) and 0.5 M EDTA were
added, incubating at 37 °C for 40 min. The coverslips and remaining
collagen layer was then removed, allowing recovery of the released
dHL-60 cells. The remaining coverslips + collagen were rinsed with
25 mL PBS + 10% hiFBS prior to scraping the collagen together for later
gDNA extraction.
Quantification of sgRNA from CRISPRi libraries
Genomic DNA (gDNA) was isolated using QIAamp DNA Blood Maxi
(3 × 10^7–1 × 10^8 cells) or Midi (5 × 10^6–3 × 10^7 cells) kits
following protocol directions (Qiagen, #51192 and #51183). gDNA
precipitation was then used to concentrate the DNA. Briefly, salt
concentration was adjusted to a 0.3 M concentration of ammonium
acetate, pH 5.2 and 0.7 volumes of isopropanol were added. Samples were
centrifuged for 15 minutes at 12,500 × g, 4 °C. Following a decant of
the supernatant, the gDNA was washed with 10 mL 70% ethanol and spun at
12,500 × g for 10 min, 4 °C. The samples were washed in another 750 μL
70% ethanol, spun at 12,500 × g for 10 min, 4 °C, and decanted. The
pellets were allowed to air-dry prior to resuspending them in water.
The gDNA concentrations and purity were determined by UV spectroscopy.
The sgRNA sequences from each gDNA sample were PCR amplified for
sequencing following protocols provided by the Broad Institute’s
Genetic Perturbation Platform. Briefly, gDNA samples were split across
multiple PCR reactions, with 10 μg gDNA added per 100 μL reaction:
10 μL 10× Titanium Taq PCR buffer, 8 μL dNTP, 5 μL DMSO, 0.5 μL 100 μM
P5 Illumina sequencing primer, 10 μL 5 μM P7 barcoded Illumina
sequencing primer, and 1.5 μL Titanium Taq polymerase (Takara,#
639242). The following thermocycler conditions were used: 95 °C
(5 min), 28 rounds of 95 °C (30 s)–53 °C (30 s)–72 °C (20 s), and a
final elongation at 72 °C for 10 min. PCR products (expected size of
~360 bp) were gel extracted using the QIAquick gel extraction kit
(Qiagen, #28704) following protocol directions. After elution, samples
were further cleaned up using isopropanol precipitation. Here, 50 μL
PCR DNA samples were combined with 4 μL 5 M NaCl, 1 μL GlycoBlue
coprecipitant (ThermoFisher Scientific Technologies,# AM9515), and
55 μL isopropanol. Samples were incubated for 30 min and then
centrifuged at 15,000 × g for 30 min. The resulting pellet was washed
twice with 70% ice-cold ethanol and resuspended in 25 μL of Tris-EDTA
buffer. Illumina 150 bp paired-end sequencing was performed by Novogene
Corp. (Sacramento, CA).
Sequence reads were quality filtered by removal of reads with poor
sequencing quality, and reads were associated back to their initial
samples based on an 8 bp barcode sequence included in the P7 PCR
primer. The 20 bp sgRNA sequences were identified and mapped to gene
targets using a reference file for the genome-wide CRISPRi
library^[274]10.
Log[2] fold-change values were calculated from the sequencing counts
between two sets of samples. The specific comparison sets made for each
screen are described in section “Overview of cell collection and
experimental replicates.” Note that each cell migration assay resulted
in two log[2] fold-changes measurements, since two populations of cells
were collected and each compared to a reference set of sgRNA. Since we
expect these two measurements to be inversely correlated, the log[2]
fold-change values from the less-migratory population were multiplied
by −1. For example, if gene knockdown resulted in a negative log[2]
fold-change in the bottom reservoir of our chemotaxis track-etch
membrane experiment, we would expect a positive log[2] fold-change for
that sgRNA in the upper reservoir. The multiplication by −1 allowed the
two sets of log[2] fold-change values to be compared directly and we
averaged across all such measurements.
Reported Log[2] fold-changes represent averages across
median-normalized replicate measurements from the multiple experiments
performed. Here, a pseudocount of 32 was added to the sgRNA counts to
minimize erroneously large fold-change values in cases of low library
representation^[275]109. For the differentiation screen and cell
migration screens, log[2] fold-changes were also scaled to have unit
variance prior to averaging across individual experiments^[276]110. p
Values were determined by performing permutation tests^[277]111 between
the calculated fold-change values for each gene target and our set of
~500 control sgRNA fold-change values. Adjusted p values for multiple
comparisons were determined using the Benjamini–Hochberg
procedure^[278]112.
Gene set enrichment analysis
Enrichment analysis was performed using the GSEA software version
4.1.0^[279]113 following recommended parameters^[280]36. A subset of
significant gene ontology terms are shown. Statistical analyses were
performed by the GSEA software.
Image acquisition
All microscopy-based image acquisition was performed using setups
operated by MicroManager (v. 2.0)^[281]114. Details of the microscope
configurations are provided below for each of the cell migration
assays.
Cell migration assays using individual sgRNA CRISPRi lines
2D migration on fibronectin coverslips
Ibidi µ-Slide I slides (#80106, IbiTreat) were incubated with 10 μg/mL
fibronectin (from human plasma, #2006, Sigma-Aldrich Inc.) in PBS at
room temperature for 1.5 h. The channel slides were then washed once
with 1 mL RPMI media containing 10% hiFBS, once with Leibovitz’s L-15
Medium (ThermoFisher Scientific, Gibco #11415064), and then the media
was removed from the reservoirs of the channel slide. Separately,
2 × 10^5 dHL-60 cells were collected, resuspended in 1 mL L-15 + 10%
hiFBS media containing 1 μg/mL DNA stain Hoechst 33324 and incubated at
37 °C for ~15 min. Cells were then spun down, resuspended in 200 μL
L-15 + 10% hiFBS media, and pipetted into one side of the channel slide
inlet. Following a 30 min incubation at 37 °C to allow cells to settle
and adhere, the slide was rinsed with three washes of L-15 + 10% hiFBS
media to remove any remaining floating cells. We note that the switch
from RPMI to L-15 was made to avoid the need for CO[2] in our
microscope system for pH buffering.
Cells were imaged for nuclei tracking at 37 °C on an inverted
microscope (Nikon Ti Eclipse), using a ×20 objective lens (Nikon ×20
0.75 NA plan apo phase contrast), with sequential phase contrast and
epifluorescence illumination through a standard DAPI filter set. For
each sample, a 30 min time-lapse movie was acquired with 60 s
intervals. Individual experiments were performed over 2–3 days, with
five different fields of view during each acquisition. For the
higher-magnification still images, cells were imaged with a ×100
objective lens (Nikon ×100 1.45NA plan apo) with an additional ×1.5
intermediate magnification. Images were captured on an iXon EMCCD
camera (Andor).
2D migration using agarose overlay and fMLF photo-uncaging
To assess chemotaxis, we stimulated dHL-60 cells with
N-formyl-methionine-alanine-phenylalanine (fMLF) as previously
described^[282]30. Briefly, to reduce adhesion, glass bottom 96 well
plates (Cellvis, #P96-1.5H-N) were treated with 1% BSA (Millipore
Sigma, #A7979) in water for 15 min followed by two washes with water
then dried overnight at 37 °C incubator with lid slightly ajar.
Approximately 1000 dHL-60 cells labeled with Celltracker Orange
(ThermoFisher, #C2927) were plated in each well, in a 5 µL drop of
modified L-15 + 2% hiFBS media. Cells were allowed to adhere to the
glass for 5 min, before a 195 µL layer of 1.5% low-melt agarose
solution in L-15 (Goldbio, #A-204-25) was mixed 1:1 with L-15 + 10%
hiFBS media (warmed to 37 °C) and overlaid on top. We note that the
initial agarose mixture was allowed to cool to approximately 37 °C
prior to preparing the final dilution and then used immediately.
The agarose was allowed to solidify at room temperature for 40 min and
then the plate was transferred to the 37 °C microscope incubator 40 min
prior to imaging. Imaging is done using a Nikon Ti-E inverted
microscope controlled by MATLAB via Micromanager, allowing simultaneous
automated imaging of multiple wells in groups. An environmental chamber
was used to maintain a 37 °C temperature, and wells were imaged at 4X
magnification every 30 s using epifluorescence illumination with the
X-Cite XLED1 LED (Excelitas Technologies, GYX module). The excitation
light was filtered using a Chroma ZET561/10 band pass filter (custom
ZET561/640x dual laser clean-up filter). A caged UV-sensitive
derivative N-nitroveratryl derivative (Nv-fMLF) of fMLF was used at
300 nM final concentration and uncaged on the microscope by exposure to
UV light with a filter cube with 350/50 bandpass filter (max intensity
around 360-365 nm). The initial gradient was generated with a 1.5 s
exposure of UV light and recharged with 100 ms exposure after every
frame. Image processing and statistical analyses of chemotaxis were
performed using custom MATLAB software (Collins et al.^[283]30). Note
that for each experiment, each cell line was added to 24–32 wells. Each
well contained roughly 100 cells, resulting in about 20,000 cells
quantified for each cell line across five experiments.
3D migration in collagen ECM
Cells were prepared as previously described^[284]104. Briefly, 2 × 10^5
dHL-60 cells were collected, resuspended in 1 mL L-15 + 10% hiFBS media
containing 1 μg/mL DNA stain Hoechst 33324 and incubated at 37 °C for
~15 min. During incubation with Hoechst stain, a 200 μL collagen
aliquot was prepared: 6.5 μL 10x PBS, 12.5 μL 0.1 M NaOH, 111 μL L-15,
and 20 μL hiFBS were combined with 50 μl 3 mg/mL collagen. The cell
suspension was spun down and resuspended in the collagen mixture for a
final concentration of 0.75 mg/mL collagen, and then added to the
channel of an Ibidi μ-Slide I (Ibidi, #80106). After 1 min incubation
at room temperature, the channel slide was inverted to help prevent
cell sedimentation and incubated at 37 °C for gel formation. After
20 min, the channel slide media reservoirs were filled with 2 mL total
L-15 media containing 10% hiFBS and imaged within 30 min to 1.5 h.
Cells were imaged at 37 °C on an inverted microscope (Nikon Ti Eclipse)
with a ×20 0.75 NA objective lens using sequential phase contrast and
epifluorescence illumination through a standard DAPI filter set. For
each sample, a 60 min time-lapse movie was acquired at 60 s intervals.
A z-stack was acquired over 200 μm with acquisitions every 3 μm. In
general experiments were performed over three different days, with two
60 min acquisitions taken each day. Images were captured on an iXon
EMCCD camera (Andor).
Cell tracking and quantification of cell migration characteristics
Cell tracks were extracted from the DNA channel of time-lapse
microscopy images using custom code^[285]104 with Python (v. 3.9.13).
Briefly, nuclei were first identified using a morphological mean filter
with a 50 pixel radial disk structural element and thresholding using
the Python package scikit-image (v. 0.19.2)^[286]115. For each nuclei
identified, the z-coordinate was calculated by taking a
weighted-intensity average along the z-axis. With cell coordinates in
hand, cell trajectories were determined by calculating all possible
cell-to-cell displacements between consecutive time points, and then
matching cells through minimization of the total displacement across
cells. For example, cells whose displacement changed very little
between two time points would most likely correspond to the same cell.
The cell density in the ECM was kept low enough that individual cell
tracks could be easily identified. Cell track information, including
position and time, we aggregated into a single table using pandas
Python package (v. 1.4.4)^[287]116.
Non-overlapping velocities and cell speeds were calculated using the
30 s (fibronectin-coated coverslips) and 60 s (collagen ECM) frame rate
of our image acquisition. To estimate average migratory persistence
from each cell trajectory, cell tracks were analyzed using a Bayesian
inference algorithm based on a persistence random walk^[288]63.
Specific parameters were chosen empirically to best capture persistence
changes in the tracks (inference grid size = 200, pMin = 10^−5,
persistence box kernel radius = 2, activity box kernel radius = 2).
Persistences were allowed to range from −1 to 1, and activities were
allowed to range from 0 to 0.5 μm/s.
Comparisons between knockdown and control cell lines were performed
using the two-sided Mann-Whitney U nonparametric test
(scipy.stats.mannwhitneyu(), using the SciPy Python package (v.
1.9.1)^[289]117.
Adhesion assay using individual sgRNA CRISPRi lines
Approximately 400,000 dHL-60 cells were resuspended in 250 μL RPMI
media containing 10% hiFBS and placed into wells of a tissue culture
treated polystyrene plate (Genesee Scientific, #25-107). Following
incubation at 37 °C for 45 min, wells were aspirated to remove
non-adherent cells. Control wells were also included for each cell line
where cells were not aspirated. One gentle wash was performed by adding
500 μL RPMI media containing 10% hiFBS slowly to the side of each well
and then aspirating the media. Cells were collected by dislodging all
remaining cells by adding 500 μL and pipetting repeatedly. This was
performed twice, resulting in cells resuspended in 1 mL PBS. Cells were
counted and the fraction of adhered cells was determined by dividing by
the total cell count found in the control wells.
Immunolabeling for flow cytometry
Live-cell immunofluorescence measurements of cell surface expression
for CD11b (integrin ɑ[M]), CD18 (integrin β[2]) and formyl peptide
receptor FPR1 were performed on a Sony SH800 Cell Sorter. All staining
and washes were done with cells suspended in phosphate-buffered saline
(PBS) containing 2% hiFBS and 0.1% sodium azide, chilled on ice. For
each sample, one million cells were first resuspended in 100 μL buffer
containing 5 μL Fc Receptor Blocking Solution (Biolegend, #422302) and
incubated for 15 minutes. Cells were then spun down and resuspended in
100 μL of buffer containing fluorescently conjugated antibodies for one
hour (5 μL of each antibody per sample). Following staining, the
samples were washed three times by resuspending in 300 μL of fresh
buffer. Following collection of flow cytometry data,.fcs files were
exported and processed using the Python package FlowCytometryTools (v.
0.5.1)^[290]118. Gating was performed on the forward and side scatter
to isolate the population of live cells (Supplementary Fig. [291]7).
Antibodies and dilution information
Fluorophore-conjugated anti-human primary antibodies: BB515 Mouse
Anti-Human CD11b (1:20; BD Biosciences, #B564517), CD18 Mouse
anti-Human, FITC (1:20; BD Biosciences, #B555923), fMLF Receptor Mouse
anti-Human, Alexa Fluor 647(1:20; BD Biosciences, #565623), and Alexa
Fluor® 647 Mouse Anti-Human CD52 (1:20; BD Biosciences, #563610).
Isotype controls: BB515 anti-IgG1, (1:20; BD Biosciences, #B564416),
FITC anti-IgG1 (1:20; BD Biosciences, #B550616), and Alexa Fluor 647
anti-IgG1 (1:20; BD Biosciences, #B557714).
Immunolabeling for Western blots
Whole-cell protein lysates were collected from dHL-60 cells for western
blot analysis. For each sample, 5 × 10^6 cells were collected, washed
twice in ice-cold PBS, and resuspended in 100 μL RIPA lysis buffer
(Cell Signaling, #9806) containing a protease and phosphatase inhibitor
cocktail (Cell Signaling, #5872). The suspension was incubated on ice
for 10 min and vortexed briefly prior to sonication with a bath type
sonicator (Diagenode, #B01020001). Sonication was performed on their
high power setting at 4°C with five cycles of 30 s on and 30 s off.
Following sonication the suspension was spun at 15,000 × g for 10 min
at 4°C. Finally each sample was diluted with 4× Laemmli SDS-PAGE sample
buffer and heated to 98°C for 5 min.
Samples were run on 7.5% polyacrylamide gels with a protein ladder
(Bio-rad, #1610317) and and transferred to nitrocellulose membranes
(Bio-rad, #1620233) by semi-dry transfer in buffer 10 mM CAPS pH 11,
10% methanol. Transferred protein was assayed using a reversible total
protein stain kit (Pierce, #24580) prior to blocking in Tris-buffered
saline with 0.1% Tween 20 detergent (TBST) with 0.2% fish skin gelatin
(FSG) for 30 min at room temperature. Protein loading was also assessed
by staining the residual protein on the gel using Coomassie stain
(0.006% Coomassie R250 with 10% acetic acid). Primary antibodies were
diluted in TBST with 0.2% FSG and incubated overnight at 4°C, which
were co-stained with an Alexa Fluor 790 Anti-GAPDH antibody for loading
control. Blots were washed with TBST for 15 min, with buffer exchanged
every 5 min, and then stained with an HRP conjugated secondary antibody
diluted in TBST with 0.2% FSG. Following incubation for 60 min at room
temperature, the blots were washed for 30 min in TBST, with buffer
exchanged every 5 min. The blots were imaged with a digital gel
documentation system (Azure c600), allowing for detection of the
secondary HRP antibody detected using a chemiluminescence peroxidase
substrate kit (Sigma, #CPS-1) and subsequent detection of the GAPDH
loading control detected using its laser based infrared detection
system. Three to four blots were performed for each sample and analyzed
using BioRad ImageLab (v. 1.6).
Antibodies and dilution information
Rabbit anti-human primary antibodies: Total S6K (1:1000; Cell
Signaling, #2708), p-Thr389 S6K pAb (1:1000; EMD Millipore, #07-018-I),
Total Akt (1:1000; Cell Signaling, #4691), p-Ser473 Akt (1:2000; Cell
Signaling, #4060), total mTOR (1:1000; Cell Signaling, #2983),
p-Ser2448 mTOR (1:1000; Cell Signaling, #5536), Alexa Fluor 790 to
GAPDH (1:1000; Abcam loading control, #ab184578). Secondary antibody:
HRP-linked anti-rabbit IgG (1:3000; Cell Signaling, #7074 S).
Immunolabeling for fluorescence microscopy during 3D migration
Immunofluorescence imaging was performed in cells migrating in 3D
collagen gels. We began by creating a thin layer of collagen on air
plasma-treated 25 mm glass coverslips (5 min at 200 mTorr treatment;
Harrick Plasma #PDC-001). This was achieved by placing a coverslip on
top of a 30 μL droplet of 0.75 mg/mL collagen mixture in a glass petri
dish, which caused the collagen mixture to spread across the coverslip.
The collagen was allowed to gel at 37 °C for 90 min. Coverslips were
lifted off of the dish by adding PBS and gentle nudging with clean
forceps. Coverslips were then flipped collagen-side up and allowed to
sit in a culture hood until visibly dry. Next, another 0.75 mg/mL
collagen solution was prepared and mixed with HL-60 cells to produce a
solution of 5000-10,000 cells/μL. 30 μL of the cell-collagen suspension
was pipetted onto the surface of a collagen-coated coverslip and
immediately placed into a 37 °C incubator for 20 min in a covered petri
dish.
Cell-laden gels were then fixed and immunostained, with all steps
performed at room temperature. Here, a warmed solution containing 4%
PFA, 5% sucrose, and PBS for 20 min. After fixation, coverslips were
washed twice for 5 min with PBS. Cells were permeabilized with 0.5%
Triton-X 100 in PBS for 10 min, washed twice for 5 min in PBS, and then
incubated for 30 min with PBS and 0.05% Tween-20. Samples were then
blocked using 20% goat serum in PBS with 0.05% Tween-20 for 30 min.
Next, cells were immunolabeled with primary antibodies in PBS, 5% goat
serum, and 0.05% Tween-20 for 1 hr. After incubation with primary
antibody, samples were washed three times for 5 min with PBS and 0.05%
Tween-20 and then stained with secondary antibodies, phalloidin, and
DAPI, diluted in 1× PBS, 5% goat serum, and 0.05% Tween-20. Following
fluorescent labeling, samples were washed three times for 5 min with
PBS and 0.05% Tween-20 and once for 5 min in PBS. Samples were then
stored at 4 °C in PBS or immediately imaged in PBS. Imaging was
performed using 3D instantaneous structured illumination microscopy
(iSIM) using a VisiTech iSIM mounted on a Nikon Ti Eclipse, with a ×100
1.35 NA silicone oil objective (Nikon). Images were captured using dual
CMOS cameras (Hamamatsu, ORCA-fusion Gen III).
Antibodies and dilution information
Rabbit anti-human CORO1A (1:100, Cell Signaling, #D6K5B), Mouse
anti-human FMNL1 (1:100, Santa Cruz Biotechnology, #sc-390466), Alexa
Fluor 488 Phalloidin (1:400, Invitrogen, A12379), Goat anti-Rabbit IgG
Alexa Fluor 594 (1:1000, Invitrogen, #A-11012), Rabbit anti-Mouse IgG
Alexa Fluor 594 (1:1000, Invitrogen, #ab150116). Note: Control
experiments using Alexa Fluor 488, 546, and 647 secondary antibodies
without primary antibody each resulted in background signals that
appeared as small, bright punctae within the cytoplasm of HL-60 cells
and were not used for imaging.
RNA-Seq
Total RNA was isolated 5 × 10^6 cells using the RNeasy Plus Mini kit
(Qiagen, #74134). PolyA enrichment, RNA-seq, sequencing (Illumina), and
data processing was performed by Novogene Corp. (Sacramento, CA).
Sequencing reads were aligned to Homo Sapiens GRCh38/hg38 genomic
research using Hisat2 (v2.0.5)^[292]119. Differential expression
analysis of two conditions/groups, six biological replicates per
condition, was performed using the DESeq2 package (v1.20.0)^[293]120.
The statistical significance of differential gene expression was
calculated in DESeq2 using a one-sided Wald test and adjusted for
multiple comparisons using the Benjamini and Hochberg’s approach. Genes
with an adjusted p value <0.05 were assigned as differentially
expressed. Enrichment analysis was performed using
clusterProfiler^[294]121 to identify Gene Ontology (GO) and KEGG
pathways with gene sets whose expression was significantly enriched by
differential expressed genes. Gene set p values estimate the
statistical significance of gene set enrichment, calculated using a
one-sided permutation test and adjusted for multiple comparisons using
the Benjamini–Hochberg procedure.
Reporting summary
Further information on research design is available in the [295]Nature
Portfolio Reporting Summary linked to this article.
Supplementary information
[296]Supplementary Tables and Figures^ (1.9MB, pdf)
[297]Peer Review File^ (3.7MB, pdf)
[298]41467_2023_41452_MOESM3_ESM.pdf^ (76.1KB, pdf)
Description of Additional Supplementary files
[299]Supplementary Dataset 1^ (798KB, csv)
[300]Supplementary Dataset 2^ (979.3KB, csv)
[301]Supplementary Dataset 3^ (3.1MB, csv)
[302]Supplementary Dataset 4^ (3.3MB, csv)
[303]Supplementary Dataset 5^ (987KB, csv)
[304]Supplementary Dataset 6^ (1,008.7KB, csv)
[305]Supplementary Dataset 7^ (907.5KB, csv)
[306]Supplementary Movie 1^ (12.1MB, avi)
[307]Supplementary Movie 2^ (24.8MB, avi)
[308]Reporting Summary^ (1.6MB, pdf)
Source data
[309]Source Data (.zip)^ (1.3MB, zip)
Acknowledgements