Single Cell Track and Trace: live cell labelling and temporal transcriptomics via nanobiopsy

Single-cell RNA sequencing has revolutionised our understanding of cellular heterogeneity, but whether using isolated cells or more recent spatial transcriptomics approaches, these methods require isolation and lysis of the cell under investigation. This provides a snapshot of the cell transcriptome from which dynamic trajectories, such as those that trigger cell state transitions, can only be inferred. Here, we present cellular nanobiopsy: a platform that enables simultaneous labelling and sampling from a single cell without killing it. The technique is based on scanning ion conductance microscopy (SICM) and uses a double-barrel nanopipette to inject a fluorescent dye and to extract femtolitre-volumes of cytosol. We used the nanobiopsy to longitudinally profile the transcriptome of single glioblastoma (GBM) brain tumour cells in vitro over 72hrs with and without standard treatment. Our results suggest that treatment either induces or selects for more transcriptionally stable cells. We envision the nanobiopsy will transform standard single-cell transcriptomics from a static analysis into a dynamic and temporal assay.

, membrane touch (2) and vibrational noise (3) subphases whose detection guarantees a successful membrane penetration and nanoinjection. The bright field (BF) and epifluorescence (FITC, TxRed) micrographs of a HeLa (c) and M059K (d) cell acquired before and after the nanoinjection phase show the cell emitting a fluorescent signal following injection of a green (ATTO 488) and red (ATTO 565) fluorophore, respectively. (e) Boxplots showing the maximum indentation Δz in the case of nanoinjections performed into HeLa (n=10) and M059K cells (n=256), and (f) the total injection duration Δt (n=256).
Following the nanoinjection step, our platform technology enables the extraction of intracellular RNA. 132 Figure 3a shows an example of the ion current iorg and the potential Vorg applied to the electrode in the 133 organic barrel during the nanobiopsy phase where the electric potential is switched to Vorg = -500 mV for 10 134 s. Upon switching the potential, the ion current in the organic barrel increases from iorg = 0.35 nA to iorg = -135 2.83 nA and it decreases slowly until the end of the 10 s when the potential is returned to Vorg = 300 mV.

136
This behaviour is driven by the velocity of the displacement of the liquid-liquid interface between the 137 solution in the barrel (organic phase) and the cytoplasm (aqueous phase) which is fast at the beginning and       M059KGFP cell before and after nanobiopsy and nanoinjection and before (day-1) and following standard treatment (longitudinal). (C) Optical and fluorescence micrographs of an individual M059KGFP cell that is nanobiopsied and nanoinjected on day 1, survive treatment and divide, whose progeny is nanobiospied and nanoinjected a second time following treatment. (D) Illustration of the nanobiopsy count of untreated (blue) and treated (red) cells that were nanobiopsied on day 1 and died (Day 1 -died), nanobiopsied on day 1, survived, didn't divide and nanobiopsied again on day 4 (Day 1 Day 4) and cells that are nanobiopsied on day 1, divided, and whose progeny was nanobiopsied again on day 4 (Day 1 Day 4 -divided). ratio of 1:350 (M059KGFP : M059KWT). This ratio facilitated the identification of individual M059KGFP is crucial to ensure that the same cell is sampled. Therefore, the total (Δxtot) and maximum (Δxmax) cellular 173 migration over the planned 72-hour time course was quantified and showed not to differ between repeats (t-174 test, p>0.5) and to be substantially less (Δxmax < 800 µm, Δxmax < 600 µm) than the field of view (1300 x

179 180
We are therefore confident that the same cell was longitudinally sampled. Next, we tested whether the same     via STARsolo (v2.7.10b) with "-outFilterMatchNmin" set to 60 to ensure high stringency alignments.
>150 expressed genes, <30% mRNA bases, and <10% ribosomal bases. Genes were filtered separately in 480 each dataset for those that have >3 counts in each of 2 or more cells. Non-protein coding genes were also 481 removed. Gene counts were normalised and scaled via Seurat (v4.3.0) using default settings with the addition 482 of "scale.max = 5" and regressing out number of genes expressed and % mRNA bases, for the nano biopsies, 483 and number of genes expressed for the whole cell lysates. The top 600 highly variable genes were identified 484 in each using the Seurat "disp" method and used to perform PCA. UMAPs were then generated from the first