Collective cancer cell invasion requires RNA accumulation at the invasive front

Significance Specific RNAs are enriched at protrusive regions of migrating cells. This localization is important for cell migration on 2D surfaces. However, in vivo, tumor cells navigate complex 3D environments often in collective groups. Here, we investigated protrusion-enriched RNAs during collective 3D invasion. We show that specific RNAs exhibit a striking accumulation at the front of invasive leader cells. We provide insights into the mechanism underlying RNA accumulation at the invasive front, and we further demonstrate that it is required for efficient 3D invasion of tumor cells. We additionally observe RNA enrichment at invasive sites of in vivo tumors, supporting the physiological relevance of this mechanism and suggesting a targeting opportunity for perturbing cancer cell invasion.


Plasmid constructs and lentivirus production
To express Citrine-CAAX, the coding sequence of mCitrine was fused to the Cterminal targeting sequence of K-Ras4B and inserted into the pCDH-CMV lentiviral vector (System Biosciences, cat #CD510B-1).
To express shRNAs targeting luciferase (sh-Control) or E-cadherin (sh-Ecad), the following sequences were cloned into the pLVTHM lentiviral vector (Addgene #12247): sh-Control: CGTACGCGGAATACTTCGA sh-Ecad: GGCCTGAAGTGACTCGTAA To re-express shRNA-resistant E-cadherin in sh-Ecad cells, site directed mutagenesis was used to introduce the following silent mutations (underlined residues; GG CCA GAG GTC ACA AGG AA) within the human E-cadherin coding sequence (corresponding to isoform NM_004360). As a control, instead of E-cadherin, the GFPnanobody sequence was used. Each coding sequence was cloned into the pENTR1A vector and subsequently transferred into the pINDUCER20 lentiviral vector (Addgene #44012) using the Gateway LR clonase II Enzyme mix (Thermo Fisher Scientific, cat# 11791-020).
Lentivirus was produced in HEK293T cells cultured in DMEM supplemented with 10% FBS and penicillin/streptomycin and grown at 37°C, 5% CO2. HEK293T cells were transfected with pCDH-CMV, pLVTHM or pINDUCER20 lentivectors, together with packaging plasmids pMD2.G and psPAX2 using PolyJet In Vitro DNA Transfection Reagent (SignaGen) for 48 h. Harvested virus was precipitated with Polyethylene Glycol at 4°C overnight. 3 Exogenous expression of E-cadherin was induced by the addition of doxycycline (Thermo Fischer Scientific). The concentration of doxycycline treatment varied as indicated, otherwise 0.1µg/ml was used.

Spheroid formation and invasion assay
To form spheroids, MDA-MB-231 cells were diluted (25,000 cells/ml) and plated as 25µl droplets on the lid of a 15cm culture dish. The lid was inverted onto a dish containing PBS, and aggregate spheroids were allowed to form for 72 h. Spheroids were collected via gentle scraping and spun down. On ice, spheroids were then resuspended in Matrigel (Corning, cat# 354234) and plated in an 8-well chamber slide (30µl/well, 1 plate spheroids/2wells). The bottom of each well was pre-coated with 10µl of Matrigel 30 min prior to plating. Regular media (600µl/well) was added 30 min after plating.
To induce invasion, 2 h after adding media, spheroids were washed once with PBS, and low-serum media (0.1% FBS, 600µl/well) was added. Spheroids were then incubated overnight. To illuminate spheroid borders, calcein dye (Thermo Fischer Scientific, cat# C3100MP, 5µM in PBS) was added, followed by a 20 min incubation at 37°C. Spheroids were fixed with 4% PFA in PBS for 10 min, washed once with PBS and stored in fresh PBS at 4°C until imaging.

Drug treatments
Spheroids were incubated in low-serum media (0.1% FBS) overnight, after which cytochalasin D (10µM) or nocodazole (2.5µM) were added directly to the media for 45 min.
A low concentration of nocodazole was chosen in order to not abolish invasion entirely, and enough protrusive strands were retained for analysis. Spheroids were fixed and FISH performed as described.

Proliferation assay
Cells were plated in a 24-well plate for approximately 8 h (6,000 cells/well), after which the media was replaced and Vivo-Morpholinos (20µM) were added directly to each well. Cells were counted every 24hrs for 5 days from triplicate wells.

Flow cytometry
For flow cytometry, cells were plated and induced with doxycycline (0.1 µg/ml) for 72 h. To detach cells, they were treated with EDTA (5mM) in PBS for 15 min at 37°C. Cells were then resuspended in flow cytometry buffer (PBS containing Ca 2+ , Mg 2+ , and 2% fetal bovine serum) and incubated for 30 min with an antibody diluted in the same buffer.
Incubation at 37°C for 1 h following EDTA treatment was included in one trial with no noticeable difference. The antibodies used were anti-E-cadherin (R&D Systems, cat# FAB18381P, 5µl) and anti-mouse IgG (R&D Systems, cat# IC0041P, 5µl) conjugated to phycoerythrin for flow cytometry detection. For analysis, the median fluorescence intensity of E-cadherin was divided by the median intensity of IgG for each sample.

Immunofluorescence and Western blot
For single-cell IF, cells were plated onto collagen IV-coated glass coverslips (10µg/ml) in low-serum media (0.1% FBS) for 1.5 h, then treated with cytochalasin D and nocodazole (same concentration/duration as spheroids). Cells were fixed with 4% PFA in PBS for 15 min, permeabilized with 0.2% Triton X-100 in PBS for 4 min, blocked with 5% fetal bovine serum in PBS for 1 h, and incubated with anti-a-tubulin (Sigma-Aldrich, cat# T6199, 1/500 dilution) for 1.5 h. Secondary antibodies were conjugated with Alexa 647 (Thermo Fisher Scientific), and phalloidin conjugated with Alexa 488 (Thermo Fisher Scientific, cat# A12379, 1/100 dilution) was added to stain F-actin. Nuclei were stained with DAPI, and Fluoromount-G Mounting Media (Thermo Fisher Scientific) was used to mount samples.
For Western blotting, cells were plated and induced with doxycycline for 72 h. To make lysate from spheroids, after hanging droplets were incubated for 72 h, spheroids were collected, spun down, resuspended in Matrigel and plated in a 12-well dish (60µl/well, 1 plate spheroids/well). Spheroids were cultured in either normal or low-serum media (0.1% FBS) overnight. To collect spheroids, they were washed once with PBS and a pre-chilled solution of cell harvesting buffer (Trevigen, cat#3448-020-01) in PBS was added to degrade the Matrigel, for 1.5 h on ice at 4°C. For each sample, two wells of spheroids were combined into an Eppendorf tube, spun down (1000xg, 1 min) and resuspended in lysis buffer. The following antibodies were used for detection: anti-Ecadherin mouse clone 36 (BD Transduction Laboratories, cat # 610182, 1/1000 dilution), anti-GAPDH rabbit monoclonal 14C10 (Cell Signaling Technology, cat# 2118, 1/2000 dilution).

Fluorescence in situ hybridization (FISH)
For single-cell FISH, HeLa-O3-v cells were plated in low-serum (0.1% FBS) media on fibronectin-coated glass coverslips (5µg/ml) for 4-5 h and then fixed with 4% PFA for 20 min. The View RNA ISH Cell Assay kit (Thermo Fischer Scientific) was used to perform FISH, according to the manufacturer's instructions. The following probe sets were used: human RAB13 #VA1-12225, human NET1 #VA1-20646 and human RHOA #VA6-14829.
CellMask stain (Thermo Fischer Scientific) was used to identify cell outlines, and DAPI was used to stain nuclei. ProLong gold antifade reagent (Thermo Fischer Scientific) was used to mount samples. Image analysis and quantification of RNA distributions was performed using the RDI calculator (62).
For FISH of spheroids, following overnight invasion, spheroids were washed once with PBS. To fix spheroids and dissolve the Matrigel, a pre-chilled solution of 4% PFA/cell harvesting buffer in PBS was then added for 2 h, on ice at 4°C. If the matrix was not fully dissolved after 2 h, additional harvesting buffer in PBS was added overnight. FISH was performed as described above, but with 5 min wash steps throughout.

Imaging and image analysis
For spheroids, z-stacks of magnified protrusive cell strands were obtained in addition to images of the spheroid body. RNA distributions were then analyzed using a two-step process. First, a custom MATLAB script identified individual RNA spots based on the ratio of local signal intensity peaks to the surrounding background. The script then determined the distance of each spot to two user-defined "edges" in the image. For leader cells, they were the front of the nucleus ("nuclear edge") and protrusive tip of the cytoplasm ("invasive edge"). We consider as "invasive edge" the tip of the most extended protrusion even though secondary, side protrusions might exist. Any RNA spots located within nuclei were omitted. RNA distances were then further analyzed using a custom RStudio script.
For each cell, the length between edges (L) was calculated: where n equals the total number of RNAs in a cell. The mean normalized distance (M) of each RNA type, RAB13 (or NET1) and RHOA, to the invasive edge was then determined: where n equals the total number of each respective RNA type in a cell. M is a number between 0 and 1, with values closer to 0 representing an RNA distribution that is biased toward the invasive edge. Across multiple cells, a probability density function (using the kernel density estimation) that described the mean normalized distance of each RNA type was plotted, and a Wilcoxon matched-pairs sign rank test statistically compared the distributions. The same analysis workflow was used to analyze RNA distributions in 7 follower cells, characterized as the cell immediately trailing a leader cell. In this case, the edges were defined as the front of the follower cell nucleus and back of the leader cell nucleus.
Analysis of RNA distributions in tissue sections was performed in a similar manner.
The imaging of tissue sections focused on xenograft edges, characterized by an interface between the xenograft and normal mouse tissue. Along these edges, strands of cells with distinct leader and follower cells were observed protruding into the surrounding tissue. We presumed these strands to have an invasive phenotype based on prior data (36).
Protrusions were randomly selected and imaged, without viewing the RNA channel prior to imaging. Leader and follower cells were manually identified, with leader cells being the frontmost cell in a protrusion and follower cells being a cell within the same strand that had a well-defined cytoplasm by human GAPDH signal. RNA distributions in leader and follower cells were quantified using the same process as spheroids. Within a cell, the distribution of a localized RNA was compared to GAPDH, which is randomly distributed throughout the cytoplasm. assigned an x-axis value of 0. Due to protrusions being different lengths, a mean intensity and 95% confidence interval were calculated and plotted for bins which contained data from greater than 50% of the protrusion sample size.

Statistical analysis
The normality of distributions was assessed using the D'Agostino & Pearson test or Shapiro-Wilk test (GraphPad Prism). Normally distributed datasets were analyzed using parametric statistical tests. Datasets deviating from a normal distribution were analyzed     The panels detecting specific human or mouse RNAs are the same as those in Figure 7C and 7D and are presented again here for comparison to the negative control. Scale bars: 20 µm