We are searching data for your request:
Upon completion, a link will appear to access the found materials.
Source: BiochemFFA_8_1.pdf. The entire textbook is available for free from the authors at http://biochem.science.oregonstate.edu/content/biochemistry-free-and-easy
To separate compounds from cellular environments, one must first break open (lyse) the cells. There are several ways of accomplishing this.
- Osmotic shock and enzymes: One way to lyse cells is by lowering the ionic strength of the medium the cells are in. This can cause cells to swell and burst. Mild surfactants may be used to disrupt membranes. Most bacteria, yeast, and plant tissues are resistant to osmotic shocks, because of the presence of cell walls, and stronger disruption techniques are usually required. Enzymes may be useful in helping to degrade the cell walls. Lysozyme, for example, is very useful for breaking down bacterial walls. Other enzymes commonly employed include cellulase (plants), proteases, mannases, and others.
- Mechanical disruption: Mechanical agitation may be employed in the form of beads that are shaken with a mixture of cells. In this method, cells are bombarded with tiny, glass beads that break the cells open. Sonication (20-50 kHz sound waves) provides an alternative type of agitation that can be effective. The method is noisy, however, and generates heat that can be problematic for heat-sensitive compounds.
- Pressure disruption: Another means of disrupting cells involves using a “cell bomb”. In this method, cells are placed under very high pressure (up to 25,000 psi) and then the pressure is rapidly released. The rapid pressure change causes dissolved gases in cells to be released as bubbles which, in turn, break open cells.
- Cryopulverization: Cryopulverization is often employed for samples having a tough extracellular matrix, such as connective tissue, seed, and cartilage. In this technique, tissues are frozen using liquid nitrogen and then impact pulverization (typically, grinding, using a mortar and pestle or a powerful electric grinder) is performed. The powder so obtained is then suspended in the appropriate buffer.
Whatever method is employed to create a lysate, crude fractions obtained from it must be further processed via fractionation.
Laboratory Methods in Cell Biology
Brandon L. Miller , . William G. Lowrie , in Methods in Cell Biology , 2012
5.2 Step 2—Lyse Red Blood Cells Using Lysis Buffer
|Overview||Lyse red blood cells by adding whole blood to tubes containing lysis buffer.|
|2.1||Prepare lysis tubes by adding 42.5 mL water and 5 mL lysis buffer stock solution to each 50 mL tube.|
|Tip One tube can lyse up to 2.5 mL blood. For more blood, prepare additional tubes.|
|2.2||Add up to 2.5 mL blood to each of the tubes prepared in the previous step.|
|2.3||Cap the tubes and mix by inverting each tube several times.|
|2.4||Incubate at room temperature for 5 min.|
|2.5||Centrifuge the tubes at 350g for 5 min. Discard the supernatant.|
|Tip||Step 2 can be repeated, if necessary, to further lyse any remaining red blood cells. However, the recovery of nucleated cells will also be affected. It is generally not required, nor recommended, to perform more than two lysis steps.|
See Fig. 3 for the flowchart of Step 2.
FIGURE 3 . Flowchart of Step 2.
Gene expression profiling has traditionally been performed on rather large samples with plenty of material. However, tissues contain many cell types that respond differently to stimuli and environmental changes, which complicate interpretation. Many studies are confounded by the intrinsic heterogeneity of biological samples. With single-cell analysis this complexity is eliminated and the true response of each cell type can be studied (1, 2). Recent single-cell profiling studies have shown large variability in transcript levels among individual cells in seemingly homogeneous cell populations and have revealed previously unknown subpopulations (3, 4). Analysis of individual cells clearly opens up for new possibilities to study biological processes such as cell transitions, signaling, differentiation, and proliferation (5, 6).
Reverse transcription quantitative real-time PCR (RT-qPCR) is the golden standard for gene expression profiling (7, 8). Through the implementation of the guidelines “minimum information for publication of RT-qPCR experiments” (MIQE) the technique has become robust and reliable (9). Usually samples composed of hundreds of thousands of cells are analyzed. These samples are lysed with strong chaotropic agents that release and protect nucleic acids, which are then purified using protocols that remove contaminants and substances that might interfere with downstream RT-qPCR (10, 11). Common to these methods is that they include one or more washing steps that lead to losses. As we write, the catalog of known RNA types is growing, resulting in increased appreciation for the numerous biological functions carried out by RNA (12). A typical single-cell contains rather few transcripts of most genes. Recent RNA sequencing data suggest there are some 22,000 mRNAs in a mouse embryonic stem cell and some 505,000 mRNAs in a mouse embryonic fibroblast. The top thousand transcripts are present in 1.4 molecules per stem cell and 77 molecules per fibroblast (13, 14). Clearly, when analyzing single-cells any loss during extraction caused by washing can introduce serious uncertainty and even total loss of some transcripts. Hence, classical purification protocols based on washing are not suitable for single-cell analysis (15, 16). A protocol based on a lysis medium that disrupts the cell membrane, makes RNA accessible for RT and maintains RNA integrity without inhibiting the downstream enzymatic reactions offers great advantages in quantitative single-cell gene expression profiling.
In this work we study lysis buffers that are suitable for small samples (1 cells) and do not require washing. We test several lysis agents in use today, comparing lysis yield, reproducibility, and RNA stability. The effect on sample handling after cell lysis is another important parameter to consider, since the time from cell collection to storage can vary from minutes to hours. We also compare the sensitivity of direct cell lysis to traditional column based RNA extraction protocols, and we test how many cells can be analyzed without downstream inhibition. To assess yields and validate reproducibilities we use RNA and DNA spikes (17).
Reference cells as spike-in controls identify contamination in single-cell RNA-seq
To systematically assess the effect of drug treatments on primary pancreatic islets with quantitative cell-type-specific resolution, we exposed intact human and mouse islets to 10 μM artemether, 100 μM GABA, 1 μM FoxO inhibitor AS1842856 (FoxOi), and control DMSO for 72 h ex vivo (Fig. 1a). After that period, we dissociated islets, filtered for single cells, and performed single-cell transcriptome analysis on the 10X Chromium platform. Following alignment and initial quality control, we obtained a total of 142,165 transcriptome profiles.
Spike-ins enable assessment of contamination in scRNA-seq experiments. a Islet treatment and sample preparation scheme. b Expression (log TPM) of insulin and glucagon in all cells of DMSO-treated islets from human donors I and II. Horizontal lines indicate the shifts of the non-alpha and non-beta cells from non-expressing levels. Red dotted line: baseline expression for non-islet cells as reported in the Human Cell Atlas (HCA). c Expression of insulin and glucagon in external datasets. TPM and RPKM values reported in each dataset were log transformed. d Alignment of reads to both the human and mouse genomes to identify spike-in reference cells, here mouse spike-ins in a human sample (DMSO, donor II). e Contamination of mouse spike-ins in human samples. Contamination is quantified as the percentage of reads aligned to the human genome in mouse spike-ins. f Correlation of the contamination profile within the mouse spike-in cells in human samples. g Average contamination in mouse spike-ins (y-axis) versus average expression in the sample (human pancreatic cells, x-axis) shown for sample DMSO, human donor II. Spearman correlation R = 0.888. h Average expression (log TPM) of genes with the largest difference in mouse and human spike-ins, and external references
Analyzing the resulting human transcriptomes, we observed striking sample-specific differences in glucagon (GCG) and insulin (INS) expression across all cell types between replicates (Fig. 1b). In addition, we observed surprisingly high levels of expression for both GCG (3.0% and 6.1% of all reads in replicates I and II, respectively) and INS (6.5% and 4.5%) in all cells, and no INS-negative cells were observed (Fig. 1b). Similar patterns are present in most published droplet-based scRNA-seq datasets of pancreatic islets (Fig. 1c). In contrast, hormone-negative cell populations were detected in some scRNA-seq studies based on sorting cells to individual wells, although also there the majority of islet cells have detectable levels of both INS and GCG. The high study dependence and low fraction of monohormonal cells indicates potential technical biases in the scRNA-seq results.
Inclusion of internal standards is widely appreciated for quality assessment, normalization, calibration, and quantitation of analytical data . Such internal standards should be spiked in early during the sample processing workflow and have highly similar, yet clearly discriminable properties to the analytical samples. For single-cell transcriptomics, ideal internal standards are therefore well-characterized homogenous cell populations. We chose to use two cell lines, mouse 32D and human Jurkat cells, as internal standards, both of which we characterized by RNA-seq. Methanol-fixed cells (
5% of all cells) were spiked into all samples shortly before droplet formation, and thus prior to processing, sequencing, and bioinformatics analyses as part of the entire dataset. We found that cross-species spike-ins (here mouse 32D cells) provide particularly clean reference points when aligned to a combined human/mouse reference genome, for two reasons. First, they were easily separable from other cells (here: human islet and spike-in) by the ratio of human/mouse reads per cell (Fig. 1d and Additional file 1: Fig. S1). Second, counting the number of cross-aligned reads (here: human reads found in mouse cells) provided a straightforward way of evaluating sample-specific biases.
Measuring contamination as the percentage of reads that aligned to the human reference in mouse spike-in cells (Fig. 1e), we observed a high, sample-specific, level of contamination in both samples (medians 8.1% and 17.4% in samples from human donors I and II, respectively), with a maximum of 19.9% in sample I. However, the contamination profile was highly similar within the cells of each sample (Fig. 1f). This human contamination in mouse spike-ins was highly correlated with the average expression in human cells (Fig. 1g), suggesting that the contamination was indeed derived from islet RNA. To identify affected genes, we next compared mouse and human spike-ins in human islet samples to mouse and human spike-ins sequenced in isolation as external reference, all aligned to the human genome (Fig. 1h). Both human and mouse spike-ins from islet samples showed strong expression of major islet marker genes such as INS, GCG, TTR, SST, PPY, IAPP, and REG1A. Given that expression of these genes was virtually absent from reference cells, it can be attributed to contamination.
Likely the contamination was found in the medium or buffer the cells are resuspended in, in the form of cell-free RNA originating from dying cells that were enclosed in droplets during processing. Possibly cell lysis may occur during the single-cell transcriptomics workflow during incubation of the cell suspension with the master mix providing reagents for the reverse transcription step prior to droplet generation. Importantly, both empirical and theoretical considerations excluded the alternative hypothesis that contamination occurs between samples during sequencing through index switching . Empirically, we observed low overlap in barcodes between experiments run on the same lane (Additional file 1: Fig. S2). Theoretically, all contaminated cells (barcodes) in a given sample would require a corresponding cell (barcode) in another sample that was processed on the same lane and contains all contaminating genes. This is highly unlikely given that only a small fraction of all barcodes is labeled as cells.
Spike-in reference cells enable accurate computational correction of cell signatures
We model contamination as a transcriptional signature that is added to the transcriptional signature of each cell. While the signature of contamination is highly similar across cells within each sample, the extent to which this profile is added is specific to each cell (Fig. 1e). Decontamination thus requires the estimation of two variables. First, the extent of total contamination of each cell, quantified as the fraction of contaminating reads (fc). Second, the extent to which each gene contributes to the contamination in each sample, i.e., the transcriptional signature of contamination (Sc). Estimation of these factors is difficult in classical single-cell RNA-seq experiments where only the final, contaminated profile of each cell is observed. However, both variables can be estimated from spike-in cells, where contaminated expression profiles can be compared to clean reference profiles, which were obtained by sequencing spike-ins separately (in the absence of potentially contaminating cells). Of particular advantage are cross-species spike-ins, where cross-species aligned reads enable straightforward quantification of contamination signature and fraction in those spike-ins (Fig. 1g). For example, reads aligned to the human genome in a mouse cell are likely to arise from contamination.
Based on the above rationale, we devised a computational decontamination procedure based on spike-in cells to correct expression data of all cells (Fig. 2a). First, we estimated the Sc in each sample by comparing contaminated mouse spike-ins to clean references. Next, we predicted the fraction of contamination for all cells. To do so, we relied on mouse spike-ins, where fc was calculated as the fraction of cross-species aligned reads. Based on these ground truth fc values in mouse spike-ins, we fitted linear models that learn to predict fc based on the expression of the most contaminating genes. Next, these models were applied to predict fc in all human cells. Finally, Sc scaled by fc was subtracted from the expression profile of all cells.
Accurate computational correction of cellular signatures based on spike-in cells. a Mouse spike-ins are used to estimate the fraction of contamination fc and the signature of contamination Sc. b Predicted versus measured fc. c, d Pearson correlation of raw (black) and corrected (red) values of the spike-ins in the human and mouse samples to the external reference spike-in transcriptome. e Flow diagram of data correction of DMSO samples from donors I and II. Left to right: 1—Scatter plot of raw INS and GCG from human donors I and II. 2 and 3—Density plots of donors separately of raw and corrected data. 4—Scatter plot of corrected INS and GCG from human donors I and II. ***p < 0.0001, **p < 0.001, *p < 0.005
To evaluate our correction procedure, we first assessed the accuracy of our fc predictions using 3-fold cross validation within mouse spike-ins. This assessment showed high prediction performance (Fig. 2b, R = 0.72). Next, we correlated the corrected values of mouse and human spike-ins to clean external references. In all cases, corrected values correlated more strongly with external references than did the raw data. This evaluation was done in cross-validated mouse as well as human data unseen by the predictor (Fig. 2c). Similar high performance was observed when decontaminating mouse cells based on human spike-ins (Fig. 2d).
After decontamination, hormone reads were only present in endocrine cells and not in all cell populations (Additional file 1: Fig. S3a). Also, our decontamination method showed better performance when compared to other correction methods  (Additional file 1: Fig. S3b, Additional file 2: Table S1).
Altogether, these analyses demonstrate that cellular spike-in controls not only are powerful tools to detect sample contamination with cell-free RNA, but also enable highly accurate correction of contaminated data (Fig. 2e). Importantly, spike-ins require marginal sequencing resources as they only constitute a small fraction of all cells analyzed.
Machine learning enables marker-free assignment of cell types in pancreatic islets
Based on our contamination-corrected expression data, we next sought to identify cell types to assess cell-type-specific effects of drug treatment. To identify islet cell subpopulations, we initially performed principal component analysis followed by t-distributed stochastic neighborhood embedding (t-SNE) on corrected gene expression data (Additional file 1: Fig. S4). We identified clusters of non-endocrine acinar, ductal, endothelial, and immune cells. In addition, we identified same-species spike-ins by their correlation to reference transcriptomes. Endocrine cell types were not well separated in this initial analysis. We therefore repeated clustering and t-SNE analysis separately for endocrine cells (Fig. 3a) and observed improved resolution but still no clear separation (Fig. 3b).
Cell type assignment. a t-SNE plot of endocrine human cells. b Expression of marker genes on t-SNE plot. c Cell type probability predicted for each cell not used in training in human samples. Cell type reassignment based on predictions was only done for those cells not previously assignable, here labeled “endocrine”
We then employed a machine learning-based approach, which was trained on unambiguously assignable cells, to predict cell type of difficult to assign cells. First, we assigned clear representatives of alpha, beta, gamma, and delta cells solely based on high expression of insulin, glucagon, somatostatin, or pancreatic polypeptide. In addition, a cluster expressing high levels of REG1A was assigned as “acinar like” in human samples (Fig. 3b). We next trained a classifier to predict cell type from the transcriptome based on these clear representatives of each cell type, which were first subsampled to reduce class imbalance. Importantly, cell type defining hormones were removed from the transcriptomes. The trained classifier was then applied to all cells not used in training, demonstrating the high prediction performance (Fig. 3c and Additional file 1: Fig. S5a). Computationally predicted class was then used to assign cell types to only those cells that were not assignable based on marker genes (labeled “endocrine” in Fig. 3c and Additional file 1: Fig. S5a). Since we are interested in studying cell type identities, a heterodoublet between two different cell types could have an impact on our analysis. Given that there was no clear cutoff for doublets removal based on the classical cutoffs of nGene or nUMI, we performed a cell-type-specific filtering based on cell type predictions and nGenes (Additional file 1: Fig. S6 and Additional file 3: Table S2).
Following assignment of single cells to cell types, we first analyzed the ratios of different cell types compared to the total population (Additional file 1: Fig. S5b-c). Overall, we observed strong donor-to-donor differences, which dominated over the effects of drug treatment with FoxOi, artemether, and GABA. To evaluate cell-type-specific drug effects, we calculated relative cell-type-specific gene expression changes by comparing compound treatment to donor-matched DMSO controls (Additional file 4: Table S3).
Pharmacological inhibition of FoxO induces islet cell dedifferentiation in vitro
We first analyzed the effects of the FoxOi on islet cells, to evaluate whether the compound can model beta cell dedifferentiation in vitro. In line with genetic models of FoxO loss [33, 56, 57], the FoxOi caused a reduction of insulin expression in mouse beta cells that we also observed in human beta cells (Fig. 4a, Additional file 5: Table S4).
FoxOi induces dedifferentiation in alpha and beta cells of human and mouse islets. a Insulin expression in beta cells from human (INS) and mouse (Ins1 and Ins2) islets treated with 1 μM FoxOi for 72 h (*p < 10 −10 , **p < 10 −45 ). b GSEA with the set of upregulated genes in triple Foxo knockout mice (FoxO1 −/− , FoxO3a −/− , and FoxO4 −/− )  in beta cells from human and mouse islets treated with 1 μM FoxOi for 72 h. c Gene expression changes in alpha (top) and beta (bottom) cells from human and mouse islets treated with 1 μM FoxOi for 72 h compared to DMSO treatment. d Correlation of alpha and beta cell gene expression from human and mouse islets treated with 1 μM FoxOi for 72 h to an alpha or beta cell gene signature set
To analyze whether transcriptome-wide changes reflected a true dedifferentiation event, we used gene set enrichment analysis (GSEA) to compare expression changes in beta cells treated with FoxOi to known beta cell dedifferentiation signatures . Genes with increased expression in a triple FoxO KO mouse model were upregulated upon FoxOi treatment in mouse and human islets (Fig. 4b). Important beta cell transcription factors such as NEUROD1, ISL1, NKX6-1, NKX2-2, FOXA2, MAFA, PDX1, and FOXO1 were downregulated (Fig. 4c). Consistent with the genetic models of FOXO loss, we identified LY96 and immature beta cell markers CRYBA2 and C2CD4A to be upregulated in beta cells after FoxOi treatment. GSEA revealed that the main pathways associated with the downregulated genes in FoxOi-treated beta cells were “regulation of gene expression in beta cells,” “pancreatic secretion,” and “type II diabetes” (Additional file 6: Table S5).
Interestingly, we also observed a loss of alpha cell identity in response to FoxOi treatment. Glucagon, as well as the master transcription factor ARX, and other alpha cell-specific genes such as GC, NEUROD1, TTR, PAX6, PCSK2, MAFB, and TM4SF4 were downregulated (Fig. 4c).
In order to validate the loss of alpha cell and beta cell identity in an unbiased approach, we correlated gene expression signatures of FoxOi-treated cells to gene sets comprised of non-hormone alpha and beta cell markers (Fig. 4d). Thus, we obtained a value of correlation to an alpha or beta cell signature gene set for each single cell. In both mouse and human islets, FoxOi induced a loss of both alpha and beta cell identity and a shift in both cell populations to less committed gene expression signatures.
A subset of alpha cells in islets treated with artemether upregulate insulin and beta cell markers
We next analyzed the effects of artemether on alpha cells based on the scRNA-seq data. In DMSO-treated controls, 1–6% of mouse and human alpha cells express detectable levels of insulin (4–6% Ins1, 0.7–3.3% Ins2, 1.3–4.3% INS). Following 10 μM artemether treatment for 72 h, we observed a consistent increase in the population of insulin-expressing alpha cells by approximately 3-fold in human islets and 2-fold in mouse islets (Fig. 5a, Additional file 1: Fig. S7a and Additional file 7: Table S6). This effect was also observed in a third human human donor treated with artemether for 72 h, but not for 36 h (Additional file 1: Fig. S8a). This increase was consistently detected in cells predicted to be alpha cells with > 50% probability for both human and mouse samples, ensuring that these cells should be considered as alpha cells rather than any other endocrine cell type (Additional file 1: Fig.S7b-c). This effect was observed independently of the insulin/glucagon ratio in islets from the different donors (Additional file 1: Fig. S7d). Importantly, treatment with FoxOi failed to increase the fraction of alpha cells that express insulin, suggesting an artemether-specific effect not linked to general dedifferentiation (Fig. 5a and Additional file 1: Fig. S7A).
Artemether upregulates insulin in a subset of mouse and human alpha cells while effects in beta cells are species dependent. a Inverse cumulative distribution of insulin expression in assigned alpha cells from human (INS) and mouse (Ins1 and Ins2) islets treated with 1 μM FoxOi or 10 μM artemether or DMSO for 72 h. Plotted is the fraction of alpha cells that express insulin to a higher level as indicated on the x-axis. b Correlation of gene expression signatures in alpha cells without (Ins − ) and with (Ins + ) detectable insulin expression alpha or beta cell gene signature sets. c Gene expression changes of Ins + alpha cells relative to Ins − alpha cells. d Inverse cumulative distribution of insulin expression in beta cells from human (INS) and mouse (Ins1 and Ins2) islets treated with 1 μM FoxOi or 10 μM artemether or DMSO for 72 h. e Gene expression changes in human and mouse beta cells treated with 10 μM artemether or 1 μM FoxOi compared to DMSO
We further characterized insulin-positive alpha cells by comparing their transcriptomes to all other alpha cells (Additional file 8: Table S7). In order to assess whether the increase in insulin expression was a reflection of loss of alpha cell identity and a possible induction of beta cell identity, we tested the correlation to gene signatures specific to alpha cells or beta cells, excluding cell type defining hormones (as above). Alpha cells that expressed insulin lost alpha cell identity and gained relevant aspects of beta cell identity compared to insulin-negative alpha cells (Fig. 5b). In human, the beta cell-specific genes IAPP, DLK1, ABCC8, PDX1, MAFA, NKX6-1, and NKX2-2 were increased and the alpha cell-specific genes GCG, ARX, and TTR were decreased, in line with a more general loss of alpha cell identity (Fig. 5c). The main upregulated pathways were “insulin secretion” and “insulin signaling.” In mouse, the beta cell-specific genes Ins1, Igf1r, Pdx1, Nkx6-1, Nkx2-2, Iapp, Foxo1, Abcc8, and Slc2a2 were all upregulated, corresponding to an increase in “regulation of gene expression in beta cells” and “beta cell development”-specific gene sets (Additional file 9: Table S8) [58, 59].
We also observed this induction of insulin/glucagon double positive cells likely arising from alpha cells in artemether-treated islets from a fourth donor with Drop-seq as an alternative technology to capture single-cell RNA expression (Additional file 1: Fig. S9) .
The effect of artemether on beta cells in pancreatic islets is species-specific
We next analyzed the effects of artemether on beta cells. In mouse beta cells, artemether caused a strong decrease of insulin expression, in line with an earlier report that suggested the drug induces beta cell dedifferentiation  (Fig. 5d).
To confirm that these changes reflect a true dedifferentiation event, we compared the gene expression signatures of artemether-treated beta cells to the known transcriptomes of dedifferentiated beta cells using GSEA (Additional file 1: Fig. S10a). In mouse beta cells, artemether-induced gene expression changes correlated with those induced by FoxOi (Additional file 1: Fig. S10b, R = 0.518). With both compounds, we observed downregulation of genes in the insulin secretion, glucagon signaling, and FoxO signaling pathways such as Ucn3, Nkx6-1 Pcsk2, and FoxO1 (Fig. 5e).
In contrast to our observations for the mouse samples, beta cells isolated from human islets treated with artemether showed no reduction, and indeed a small increase, in insulin expression compared to DMSO-treated controls (Fig. 5d). Insulin expression was also not decreased in the third human donor at 36 nor 72 h (Additional file 1: Fig. S8b). This species specificity was in contrast to the effect of FoxOi, which caused insulin downregulation both in mouse and in human beta cells. In line with this difference between artemether and FoxOi treatment, the overall correlation of gene expression changes in human beta cells was found to be weaker between the two compounds (Additional file 1: Fig. S10b, R = 0.263). While FoxOi downregulated key beta cell genes including NKX2-2, PDX1, FOXA2, MAFA, and INSR, expression of these genes was mostly unaltered in artemether-treated human beta cells (Fig. 5e).
Finally, we compared drug effects across species by matching orthologous genes between mouse and human beta cells. We observed that the effects of the FoxOi were weakly correlated between species (Additional file 1: Fig. S10c, R = 0.296) whereas no correlation was observed for artemether effects in mouse and human beta cells (R = 0.129). While artemether downregulated key beta cell genes INS1/2,SLC2A2, ISL1, GCGR, UCN3, and SCG5 in mouse beta cells, the expression of these genes remained unchanged in human beta cells. In addition, many genes changed discordantly, for example artemether downregulated SPP1 in mouse beta cells, whereas it upregulated its expression in human beta cells, and vice versa for IGF1R. These data indicate that FoxOi effects are conserved in mouse and human beta cells, whereas artemether causes more species-dependent gene expression changes.
Correlating the transcription changes between individual samples further supports the finding that these drugs exert different effects in mouse and human islet cells (Fig. 6).
Artemether effects on beta cells are species dependent. a Spearman correlations of log fold changes of comparisons of 1 μM FoxOi, 10 μM artemether, or 100 μM GABA to DMSO for alpha and beta cells from mouse and human islets. b FoxOi induces dedifferentiation of alpha and beta cells in mouse and human, while artemether increases the fraction of INS + alpha cells. In beta cells, artemether effects are species dependent in mouse, the drug induces beta cell dedifferentiation, while in human there are no strong effects
Dual Targeting of Cell Wall Precursors by Teixobactin Leads to Cell Lysis
Teixobactin represents the first member of a newly discovered class of antibiotics that act through inhibition of cell wall synthesis. Teixobactin binds multiple bactoprenol-coupled cell wall precursors, inhibiting both peptidoglycan and teichoic acid synthesis. Here, we show that the impressive bactericidal activity of teixobactin is due to the synergistic inhibition of both targets, resulting in cell wall damage, delocalization of autolysins, and subsequent cell lysis. We also find that teixobactin does not bind mature peptidoglycan, further increasing its activity at high cell densities and against vancomycin-intermediate Staphylococcus aureus (VISA) isolates with thickened peptidoglycan layers. These findings add to the attractiveness of teixobactin as a potential therapeutic agent for the treatment of infection caused by antibiotic-resistant Gram-positive pathogens.
Copyright © 2016, American Society for Microbiology. All Rights Reserved.
Antibacterial activity of teixobactin is…
Antibacterial activity of teixobactin is dependent on autolysin. (A) Scanning electron microscopy of…
Teixobactin causes downregulation of atl…
Teixobactin causes downregulation of atl expression. (A) Bacteriolytic assay using supernatant from JE2…
Inhibition of WTA biosynthesis is…
Inhibition of WTA biosynthesis is responsible for teixobactin-mediated lysis. The lysis of SA113…
Mutation of saeS results in…
Mutation of saeS results in sensitization to vancomycin due to decreased WTA. (A)…
Teixobactin treatment causes Atl delocalization.…
Teixobactin treatment causes Atl delocalization. In vitro binding of amidase repeats (Cy3-R 1–3…
Genetic targeting of Card19 is linked to disrupted Ninj1 expression, impaired cell lysis, and increased susceptibility to Yersinia infection
Cell death plays a critical role in inflammatory responses. During pyroptosis, inflammatory caspases cleave Gasdermin D (GSDMD) to release an N-terminal fragment that generates plasma membrane pores that mediate cell lysis and IL-1 cytokine release. Terminal cell lysis and IL-1β release following caspase activation can be uncoupled in certain cell types or in response to particular stimuli, a state termed hyperactivation. However, the factors and mechanisms that regulate terminal cell lysis downstream of GSDMD cleavage remain poorly understood. In the course of studies to define regulation of pyroptosis during Yersinia infection, we identified a line of Card19-deficient mice (Card19 lxcn ) whose macrophages were protected from cell lysis and showed reduced apoptosis and pyroptosis, yet had wild-type levels of caspase activation, IL-1 secretion, and GSDMD cleavage. Unexpectedly, CARD19, a mitochondrial CARD-containing protein, was not directly responsible for this, as two independently-generated CRISPR/Cas9 Card19 knockout mice showed no defect in macrophage cell lysis, and expression of CARD19 in Card19 lxcn macrophages did not restore cell lysis. Card19 is located on chromosome 13, adjacent to Ninj1, which was recently reported to regulate cell lysis downstream of GSDMD activation. Intriguingly, RNA-seq and western blotting revealed that Card19 lxcn BMDMs are hypomorphic for NINJ1 expression, and reconstitution of Ninj1 in Card19 lxcn immortalized BMDMs restored cell lysis. Card19 lxcn mice exhibited significantly increased susceptibility to Yersinia infection, demonstrating that cell lysis itself plays a key role in protection against bacterial infection. Our findings identify genetic targeting of Card19 being responsible for off-target effects on the adjacent Ninj1 gene, thereby disrupting the ability of macrophages to undergo plasma membrane rupture downstream of gasdermin cleavage and impacting host survival and bacterial control during Yersinia infection.
Author Summary Programmed cell death is critical for regulating tissue homeostasis and host defense against infection. Pyroptosis is an inflammatory form of programmed cell death that couples cell lysis with release of inflammatory cytokines. Cell lysis is triggered by activation of particular intracellular pore forming proteins, but how regulation of cell lysis occurs is not well understood. Genetic targeting of Card19 on chromosome 13 resulted in decreased expression of the adjacent gene, Ninj1 which was recently found to regulate terminal lysis events in response to cell death-inducing stimuli. We found that macrophages from Card19-deficient mice were resistant to multiple forms of cell death in response to a variety of inflammatory stimuli, including canonical and non-canonical inflammasome activation, as well as triggers of cell-extrinsic apoptosis. Notably, Card19-deficient mice were more susceptible to Yersinia infection, indicating that cell lysis contributes to control of bacterial infections. Our data provide new insight into the impact of terminal cell lysis on control of bacterial infection and highlight the role of additional factors that regulate lytic cell death downstream of gasdermin cleavage.
Competing Interest Statement
The authors have declared no competing interest.