0
views
0
recommends
+1 Recommend
0 collections
    0
    shares
      • Record: found
      • Abstract: found
      • Article: found
      Is Open Access

      A call for caution in analysing mammalian co-transfection experiments and implications of resource competition in data misinterpretation

      Read this article at

      Bookmark
          There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.

          Abstract

          Transient transfections are routinely used in basic and synthetic biology studies to unravel pathway regulation and to probe and characterise circuit designs. As each experiment has a component of intrinsic variability, reporter gene expression is usually normalized with co-delivered genes that act as transfection controls. Recent reports in mammalian cells highlight how resource competition for gene expression leads to biases in data interpretation, with a direct impact on co-transfection experiments. Here we define the connection between resource competition and transient transfection experiments and discuss possible alternatives. Our aim is to raise awareness within the community and stimulate discussion to include such considerations in future experimental designs, for the development of better transfection controls.

          Related collections

          Most cited references 38

          • Record: found
          • Abstract: found
          • Article: not found

          Mammalian cell transfection: the present and the future

          Transfection is a powerful analytical tool enabling study of the function of genes and gene products in cells. The transfection methods are broadly classified into three groups; biological, chemical, and physical. These methods have advanced to make it possible to deliver nucleic acids to specific subcellular regions of cells by use of a precisely controlled laser-microcope system. The combination of point-directed transfection and mRNA transfection is a new way of studying the function of genes and gene products. However, each method has its own advantages and disadvantages so the optimum method depends on experimental design and objective.
            Bookmark
            • Record: found
            • Abstract: found
            • Article: not found

            Emergent bistability by a growth-modulating positive feedback circuit.

            Synthetic gene circuits are often engineered by considering the host cell as an invariable 'chassis'. Circuit activation, however, may modulate host physiology, which in turn can substantially impact circuit behavior. We illustrate this point by a simple circuit consisting of mutant T7 RNA polymerase (T7 RNAP*) that activates its own expression in the bacterium Escherichia coli. Although activation by the T7 RNAP* is noncooperative, the circuit caused bistable gene expression. This counterintuitive observation can be explained by growth retardation caused by circuit activation, which resulted in nonlinear dilution of T7 RNAP* in individual bacteria. Predictions made by models accounting for such effects were verified by further experimental measurements. Our results reveal a new mechanism of generating bistability and underscore the need to account for host physiology modulation when engineering gene circuits.
              Bookmark
              • Record: found
              • Abstract: found
              • Article: found
              Is Open Access

              Promoters maintain their relative activity levels under different growth conditions

              Introduction Quantitative characterization of gene expression is a fundamental yet complex challenge. One of the major challenges stems from the dynamic nature of gene expression, whereby every gene can change its expression value across conditions. Although genome-wide analyses of expression are among the most commonly used methods in modern biology (Edgar, 2002), most studies produce lists of upregulated and downregulated genes, with limited focus on the numerical change in values. Here, we identify quantitative relationships between expression profiles under different conditions and observe a unifying behavior that simplifies our quantitative understanding of gene regulation. Traditionally, gene expression research has focused on isolated genes and has generally shown that the transcriptional response is highly and specifically regulated. For example, upon exposure to lactose, bacteria respond by transcribing lactose-assimilating genes (Jacob and Monod, 1961). More recently, microarray and sequencing technologies have challenged this paradigm by enabling a genome-wide view of expression, and establishing that the responses to different conditions involve changes in expression of thousands of genes (Pedersen et al, 1978; DeRisi, 1997; Spellman et al, 1998; Gasch et al, 2000, 2001; O'Rourke and Herskowitz, 2002; Boer et al, 2003; Saldanha et al, 2004; Tu et al, 2005; Lai et al, 2005; Shalem et al, 2008; Chechik et al, 2008; Brauer et al, 2008; Yassour et al, 2009; Costenoble et al, 2011; Tirosh et al, 2011). Such massive expression changes between conditions raise several fundamental questions. Primarily, it is unclear why the expression of so many genes changes even between conditions whose phenotypic differences appear to be minor. As one example, it is unclear why growing yeast on either glucose or its epimer galactose leads to detectable expression changes in over half of the yeast genome (Gasch et al, 2000; Chechik et al, 2008) even though only a few enzymatic reactions separate the two substrates. Initial attempts to bridge the gap between specific regulation and the wide spread changes observed in the data suggested that specific responses actually encompass more genes than initially appreciated (Spellman et al, 1998; Gasch et al, 2000, 2001; Tu et al, 2005). More recently, it was shown that many changes in expression are correlated with growth rate (Pedersen et al, 1978; Regenberg et al, 2006; Castrillo et al, 2007; Brauer et al, 2008; Fazio and Jewett, 2008; Zaslaver et al, 2009; Klumpp et al, 2009; Levy and Barkai, 2009), as proposed decades ago by the Copenhagen school (Maaloe, 1969; Ingraham and Ole Maaløe, 1983; Neidhardt, 1999), suggesting that they may result from global factors affecting many genes. Although several new works have attempted to incorporate global factors into gene expression models by analyzing synthetically constructed constitutive promoters (Klumpp et al, 2009; Scott et al, 2010; Gerosa et al, 2013), to date there is still no methodology to tease apart and decouple global and specific regulation. Therefore, it remains unknown what fraction of the gene expression changes observed upon a change in the growth condition can be explained by changes in global cellular parameters and which genes are specifically regulated. The ability to differentiate the two is critical for understanding the gene expression regulation. Moreover, except for isolated works (Van de Peppel et al, 2003; Bakel and Holstege, 2008; Islam et al, 2011), the importance of global changes in expression remains under-appreciated and is typically overlooked when analyzing high-throughput expression data. Such practice may lead to major misinterpretation of expression changes, as recently shown in Lovén et al (2012). A second question invoked by the large magnitude of expression changes across conditions regards the combinatorics and complexity of gene expression programs. Considering all genes in all possible conditions, there is practically an infinite range of possible expression patterns that cells can reach. For example, yeast has ∼6000 potential transcriptional degrees of freedom as each of its ∼6000 promoters can potentially attain a different rate of transcription under each growth condition. It is intriguing to consider the degree to which this complexity is realized in cells. Indeed, genome-wide surveys of expression suggest that not all expression patterns are possible, as functionally related genes tend to be co-regulated (Pedersen et al, 1978; DeRisi, 1997; Gasch et al, 2000; Gasch et al, 2001; O'Rourke and Herskowitz, 2002; Boer et al, 2003; Saldanha et al, 2004; Lai et al, 2005; Chechik et al, 2008; Shalem et al, 2008; Brauer et al, 2008; Yassour et al, 2009; Costenoble et al, 2011; Tirosh et al, 2011). However, since most of these studies typically focus on the directionality of co-regulation (up or down) and not on the numerical change in values, to date the quantitative aspects of expression changes across conditions and their compliance with both classic (Maaloe, 1969; Ingraham and Ole Maaløe, 1983) and recent (Klumpp et al, 2009; Zaslaver et al, 2009; Scott et al, 2010) models of expression remain largely unaddressed. We aim to extend the existing qualitative description of co-regulated gene modules (Ihmels et al, 2002; Gasch and Eisen, 2002; Segal et al, 2003) and identify quantitative relationships between expression profiles under different conditions, thus reducing the space of possible expression patterns. Here, we accurately measured the activities of ∼900 S. cerevisiae and ∼1800 E. coli promoters in 10 and 9 environmental conditions, respectively, using libraries of fluorescent reporters (Zaslaver et al, 2006, 2009; Zeevi et al, 2011). Notably, we found that most promoters (60–90%, depending on the pair of conditions compared) change their expression between conditions by a constant scaling factor that depends only on the conditions and not on the promoters' identity. Thus, although there is a major change in values for nearly all promoters, the relative activity levels are preserved. Accounting for global effects allows precise quantification of more limited specific regulation—promoters deviating from global scaling. These can be organized into a handful of functionally related groups, such that within each group, promoters also preserve their relative activity levels across conditions in which they are activated. Hence, we can accurately describe 97% of the variability of the apparently complex promoter activity profiles across conditions using only several scaling factors. Finally, we present a parameter-free model that encompasses growth rate and specific gene expression and accounts for ∼90% of the observed variability in the global scaling factors. Our results provide a mean to decouple global and specific changes in activity between conditions, and a first quantitative characterization of the global response. They suggest that most changes in expression across conditions result from global effects and propose that proportional scaling is a major determinant of genome-wide expression profiles. Results Obtaining accurate measurements of promoter activity across different growth conditions To obtain accurate measurements of promoter activity in yeast, we employed an experimental system based on the genomic fusion of promoters to fluorescent reporters (Zaslaver et al, 2006; Zeevi et al, 2011). We selected 867 native yeast promoters of genes that represent a wide variety of cellular functions, processes, and compartments (Supplementary Table S1). Although these genes cover only ∼1/6 of the S. cerevisiae genome, they cover the various Gene Ontology (GO) categories (Ashburner et al, 2000), promoter types (divergent/unique) (Saccharomyces Genome Database, available at: http://www.yeastgenome.org/), promoter architectures (OPN/DPN) (Tirosh and Barkai, 2008), and transcription regulation strategies (TFIID/SAGA-dominated) (Huisinga and Pugh, 2004). In addition, their combined expression represents ∼60% of the protein mass expressed in rich media (Wang et al, 2012) and thus accounts for much of the cellular activity under standard growth conditions (Supplementary Table S1, Supplementary material 1.1). We genomically integrated each promoter upstream of a yellow fluorescent protein (YFP) and used a robotically automated plate fluorometer to track the amount of reporter expression over time, in living cells, and across various growth conditions. Altogether, 859/867 (99%) promoters were successfully constructed. Simultaneous measurements of optical density (OD), indicative of population mass (Bremer and Dennis, 1987), enabled us to extract the doubling time of the culture and calculate the YFP production rate per OD unit per second (Methods, Zeevi et al, 2011), hereafter referred to as the promoter activity (Figure 1). These strains represent the largest library of native promoter-reporter fusions in eukaryotes to date. The use of fluorescent reporters for measuring expression is a well-established approach (Bronstein et al, 1994; Kalir et al, 2001; Zaslaver et al, 2004; Newman et al, 2006; Ligr et al, 2006; Murphy et al, 2007; Cox et al, 2007; Gertz et al, 2009; Zeevi et al, 2011; Raveh-Sadka et al, 2012; Sharon et al, 2012), with several pronounced advantages. Unlike most current high-throughput techniques, which require cell lysis, fluorescence enables to perform live non-invasive imaging of the same cells over time with high temporal resolution. Accordingly, it does not require elaborate analysis and normalization techniques as required by other high-throughput measurement systems, such as microarrays or sequencing (Churchill, 2002; Marshall, 2004; Frantz, 2005; Bammler et al, 2005; Tang et al, 2007; Balázsi and Oltvai, 2007; Oshlack and Wakefield, 2009). This is especially important for our current study question as such normalizations may obliterate shared global effects (Lovén et al, 2012). It was also shown that fluorescent reporters provide highly precise and reproducible values (Bronstein et al, 1994; Kalir et al, 2001; Zaslaver et al, 2004; Ligr et al, 2006; Cox et al, 2007; Murphy et al, 2007; Gertz et al, 2009; Zeevi et al, 2011; Raveh-Sadka et al, 2012; Sharon et al, 2012). Here, we used replicate biological measurements to validate that our system provides highly sensitive and precise measurements for our set of promoters (CV ranging from 0.05 to 0.36 for promoters with high to very low activity, see Materials and methods and Supplementary Figure S2); more reproducible than those obtained by microarrays, sequencing, or mass spectrometry (Supplementary Figure S3). Together, these features of the system are critical for the ability to detect and quantify the global and specific changes in expression reported here. To make sure that our synthetic fluorescence measurements are representative of the true promoter activity (of the native gene in its native genomic location), we performed several analyses to gauge the integrity and accuracy of the system. We compared the promoter activities with quantitative real-time PCR measurements of 18 selected strains under two growth conditions, and confirmed that YFP levels are an accurate proxy for the corresponding mRNA levels (R=0.99 and R=0.98, Supplementary Figure S4A and B). In addition, we compared our promoter activity values with three microarray studies (Holstege et al, 1998; Shalem et al, 2008; Lipson et al, 2009), three RNA-seq studies (Nagalakshmi et al, 2008; Lipson et al, 2009; Yassour et al, 2009), protein abundance obtained by immuno-tagged proteins (Ghaemmaghami et al, 2003), fluorescently tagged proteins (Stewart-Ornstein et al, 2012), mass spectrometry (De Godoy et al, 2008), and a curated data set of protein abundances integrated from five different data sets (Wang et al, 2012). In all of these comparisons, our promoter activity data correlated well with mRNA and protein abundance (R=0.72–0.81 and R=0.57–0.74 respectively, similar to the correlations between these data sets; Supplementary Figure S5), suggesting that promoter activity as measured by our system is a major determinant of these properties (Supplementary material 1.2). For further discussion of the experimental system, see Supplementary material 1.3. Most promoters' expression changes between conditions by a constant factor dependent only on condition and not on the promoter's identity To compare promoter activities across conditions, we measured our library under 10 different environmental growth conditions, known to affect the expression of genes in the library (Materials and methods, Supplementary material 1.1, Supplementary Table S2). In line with observations using other methods, most promoters changed their activity levels between every pair of conditions, indicating that a major fraction of the genome responds to growth under different conditions (Supplementary Tables S3 and S4). Next, we plotted promoter activities in every pair of conditions (Figure 2; Supplementary Figure S6). Strikingly, we found that in each such pair, most promoters change their expression between conditions by a constant factor that depends only on the conditions and not on the promoter's identity. This result indicates that although most promoters change their activity between conditions, they preserve their relative values. To quantify this global effect, we robustly fitted a scale line to the promoter activities of each pair of conditions, with the slope of the line ranging from 0.19 to 1 (arbitrarily setting glucose to 1, Materials and methods, Supplementary Table S4). We quantified the extent to which promoters adhere to the scale line by three independent methods: (A) Analysis of variance: We found that the scale line captures 80–99% of the variance in the data (P 0.1, for which our measurements yield lower STD values; Supplementary Figure S2), we found that 58–88% are within ±30% of the global scale line (Materials and methods). These independent analyses confirm that for most promoters, their activity in condition B is equal to their activity in condition A multiplied by a single number (the slope of the scale line between conditions A and B). We term this number the global scaling factor between conditions A and B and return to analyze it below. In addition to the dominating global response, for each pair of conditions there remains a smaller subset of promoters that do not scale according to the global scaling factor (Figure 2, gray dots). We termed a promoter as condition specific if it is deviated from the scale line by more than three experimental standard deviations (Materials and methods), using glucose as a reference condition. We note that the relatively small sizes of the specifically responding groups does not appear to result from low representation of these promoters in the library, since our library was initially designed to represent different groups of genes (Supplementary Table S1; Supplementary material 1.1). Additionally, genes and environmental conditions were chosen together to include genes that are known to respond to the conditions (based on Gasch et al, 2000, 2001). Furthermore, we repeated the analysis, excluding known growth-related groups of genes, such as those involved in protein synthesis, and obtained very similar results (Supplementary Figure S10), indicating the robustness of our results with respect to input genes. We found that for each condition, its set of condition-specific promoters showed remarkable agreement with our understanding of yeast physiology and encompassed known co-regulated gene modules (Gasch and Eisen, 2002; Ihmels et al, 2002; Segal et al, 2003). For example, the specific response to galactose includes almost all the promoters whose corresponding genes belong to the galactose utilization pathway (8/9, P 30 biological replicates of the strain were measured and fitted to a normal distribution (Supplementary Figure S1), and the 95th percentile of the distribution was taken to be the detection level. For each strain in every condition, we took the final promoter activity levels to be the average of the strain across replicates. If this average was below the detection level, then we set the promoter activity to the detection level. Experimental variability The relative error was estimated by the coefficient of variation (CV) of replicate measurements. Mean values across expression bins were used to estimate the CV of any promoter activity level, by linear interpolation (Supplementary material 2.6). The CV values ranged from 0.36 for very low promoter activity levels to 0.05 for high promoter activity levels (Supplementary Figure S2). Global scaling factors and error model For all conditions, the global scaling factor represents the best robust fit to the data (Supplementary material 2.7). For each pairwise comparison, we determined the variance explained by the global scale line. The promoter activity values, v(p), of each promoter p was projected to the global scale line. Denoting the difference between the vector and its projection by d(p), the variance explained by the clustering was calculated as 1−variance(d(p))/variance(v(p)). To obtain a P-value for the explained variance, for each comparison between conditions X and Y, we randomized the promoter activities of condition Y and quantified the variance explained by the original global scaling line, as described above. This was repeated 1000 times for each pairwise comparison. For each pairwise comparison, we then determined which promoters behave according to the global trend between these two conditions using two separate methods: (A) We analyzed all data points above detection and estimated their probability to behave according to the global trend. For each promoter, let (x, y) and (sx, sy) denote its activity level and standard deviation in conditions x and y, respectively. Denoting the scaling factor between the respective conditions by a, then promoters were defined as part of the global trend if 0.1 in both conditions. For such values, we found the average CV to remain constant at 0.05 (Supplementary Figure S2). For each pairwise comparison, we defined a promoter as part of the global trend if its value deviated by no more than 30% from its expected value according to the global scaling factor. These two methods complement each other as the first is relative yet it enables the analysis of the entire data set, taking into account our different level of confidence in low and high values. The second is restricted to only parts of the data, yet it enables to determine absolute values. Both yielded similar results with ∼60–90% of promoters (depending on the pair of conditions compared) changing between conditions according to the global scaling factor. Quantitative PCR analysis Eighteen representative strains belonging to cluster 1 (RPB10, TEF1, DPM1, SEC61, SHP1, CDC10, RPS3, GLY1, RPL3, RPL33A, RPL8B, RPS7A, RPS11B, RPL4B, RPL28) or cluster 6 (GAL1, GAL2, GAL7) were inoculated from frozen stocks into SCD (150 μl, 96-well plate) and grown at 30°C for 48 h, reaching complete saturation. Cells were then diluted 1:36 in fresh medium and pelleted at mid-exponential phase. RNA was extracted using the EPICENTER Yeast MasterPURE RNA extraction kit, and cDNA was created using random hexamers (sigma). Quantitative PCR was performed by RT–PCR (StepOnePlus, Applied Biosystems) using a ready-mix kit (KAPA, KK4605). For each strain, measurements were performed in two sets of triplicates, measuring both YFP and RFP mRNA. Reported values are of mean YFP/RFP from nine replicates derived from three independent experiments (Supplementary material 2.8). Functional annotation and enrichment analysis Sets of genes were assigned process, function, and cellular components according to the annotations from the GO (Ashburner et al, 2000). The significant representation of GO terms in the set was evaluated by Gorilla GO Term Finder (Eden et al, 2009) with a P-value threshold of 10−3. For TF analysis, we examined the distribution of known TF promoters (Badis et al, 2008; Zhu et al, 2009) across the different clusters. For enrichment analysis, promoters were classified as previously described according to their properties as: OPN/DPN (Tirosh and Barkai, 2008), SAGA-dominated/TFIID-dominated (Huisinga and Pugh, 2004), divergent/unique (this study, based on the Saccharomyces Genome Database, available at: http://www.yeastgenome.org/). P-values were computed according to the HG distribution and corrected for multiple hypothesis testing using FDR correction (Benjamini and Hochberg, 1995). Clustering promoter activities To partition the promoters into clusters that preserve proportionality, we used K-means clustering with the cosine metric (defined by (x,y)=1−cos , where x and y are vectors of promoter activity levels in a given condition and O is the origin). The clustering was repeated 100 times with different random starting points and the clustering that minimized the sum of distances from the centers was chosen. The number of clusters, K, was determined as the largest K for which the distance between any two centers is at least 0.05 (Supplementary Figure S8), thereby ensuring a minimal separation between any two clusters (Supplementary material 2.10). For generation of Supplementary Figure S10, this analysis was repeated excluding all ribosomal promoters. Variance explained by clustering For each promoter p, its vector of promoter activity levels across conditions, v(p), was projected to the center of the corresponding cluster. Denoting the difference between the vector and its projection by d(p), the variance explained by the clustering was calculated as 1−variance(d(p))/variance(v(p)). Predicting promoter activity levels We used the following scheme to predict promoter activity levels under growth condition Y from measurements of several other conditions x 1,…,x m . First, the number of clusters k for all promoters under the measured m conditions was determined using above criterion. Then, the promoters were clustered by the k-means algorithm using the cosine metric. Denote the centers of the clusters by c 1,…,c k . A small number of representative promoters were chosen as the training set, and their promoter activity levels under the new condition Y were used for the prediction task. For each cluster t, an extended center of size m+1 was calculated from the representative promoters that belong to cluster t. The activity level of a promoter under the new condition Y was predicted to be , where x is the vector of activity levels for that promoter. The representative promoters were chosen such that an equal number of promoters were chosen from each cluster, which are closest (by the cosine metric) to the centers c 1,…,c k of the relevant clusters (Supplementary material 2.12). E. coli Growth conditions All media for bacterial growth were based on a defined M9 minimal medium (Supplementary material 3.1). Specific growth conditions and the respective growth rates in each condition are listed in Supplementary Table S8. Robotic assay for genome-wide promoter activity data The library of reporter strains, each bearing a low-copy plasmid with one of E .coli promoters controlling fast-folding GFP (Supplementary Figure S14A; Cormack et al, 1996) was previously described (Zaslaver et al, 2006). This library includes 1820 reporter strains that represent ∼75% of E. coli promoters. Reporter strains were inoculated from frozen stocks into 96-well plates containing M9 minimal medium supplemented with 11 mM glucose, 0.05% casamino acids and 50 μg/ml kanamycin and grown overnight in a shaker at 37°C. All steps from this point were carried out using a programmable robotic system (Freedom Evo, Tecan Inc.). Overnight cultures were diluted 1:33 into M9 medium followed by a second 1:15 dilution into flat bottom microwell plate (nunc) containing one of the growth media (Supplementary Table S8) in a final culture volume of 150 μl. Bacteria were grown in an incubator with shaking (6 Hz) at 37°C for about 20 h. Every 8 min the plate was transferred by the robotic arm into a multiwall fluorometer (Infinite F200, Tecan) that reads the OD (600 nm) and GFP (excitation 480 (20), emission 515(10)). After 5 h of incubation, NaCl or casamino was added to the appropriate plates by automated pipetting. Computing promoter activity levels, detection level, experimental variability and error model Promoter activity was calculated by the rate of GFP production per OD unit, as described above for yeast (Supplementary material 2.4) for the 3-h window around mid-exponential growth (Supplementary Figure S14). For conditions in which a compound was added to the media, promoter activity was calculated for the window of time after its addition. Background fluorescence was measured using a promoter-less control strain in each plate. Promoter activities lower than 3 STDs above the mean background promoter activity were set to zero. In total, 969 promoters were active above background in at least one condition. Experimental variability was assessed as described above, using three replicate measurements in M9 glucose (Supplementary Figure S15) and error model was calculated as for S. cerevisiae. Identification of representative promoters for predictions was done iteratively. At each iteration, we calculated the best linear sum of the representative promoter, which predicted the experimental data, and added an additional representative promoter, which contribute the most to predict the experimental data. Supplementary Material Supplementary information Supplementary Figures S1–19 Data Set 1 Supplementary Tables S1–S10 Review Process File
                Bookmark

                Author and article information

                Contributors
                f.ceroni@imperial.ac.uk
                Journal
                Nat Commun
                Nat Commun
                Nature Communications
                Nature Publishing Group UK (London )
                2041-1723
                5 May 2021
                5 May 2021
                2021
                : 12
                Affiliations
                [1 ]GRID grid.7445.2, ISNI 0000 0001 2113 8111, Department of Chemical Engineering, , Imperial College London, South Kensington Campus, ; London, UK
                [2 ]GRID grid.7445.2, ISNI 0000 0001 2113 8111, Imperial College Centre for Synthetic Biology, South Kensington Campus, ; London, UK
                [3 ]GRID grid.25786.3e, ISNI 0000 0004 1764 2907, Synthetic and Systems Biology lab for Biomedicine, , Istituto Italiano di Tecnologia-IIT, Largo Barsanti e Matteucci, ; Naples (ITA), Italy
                Article
                22795
                10.1038/s41467-021-22795-9
                8099865
                33953169
                © The Author(s) 2021

                Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

                Categories
                Comment
                Custom metadata
                © The Author(s) 2021

                Uncategorized

                biomedical engineering, synthetic biology

                Comments

                Comment on this article