Optimization of protoplast isolation, transformation and its application in sugarcane(Saccharum spontaneum L.)

Qiongli Wng*, Gungrun Yu1, Zhiyong Chen Jinlei Hn Yufng HuKi Wngb,*

aKey Laboratory of Genetics, Breeding and Multiple Utilization of Crops, Ministry of Education/Fujian Provincial Key Laboratory of Haixia Applied Plant Systems Biology, Center for Genomics and Biotechnology, Fujian Agriculture and Forestry University, Fuzhou 350002, Fujian,China

bNational Engineering Research Center of Sugarcane, Fujian Agriculture and Forestry University, Fuzhou 350002, Fujian, China

Keywords:Saccharum spontaneum L.Protoplast isolation PEG Transformation efficiency Transient expression

ABSTRACT Sugarcane is a prominent source of sugar and ethanol production.Genetic analysis for trait improvement of sugarcane is greatly hindered by its complex genome,long breeding cycle,and recalcitrance to genetic transformation.The protoplast-based transient transformation system is a versatile and convenient tool for in vivo functional gene analysis;however,quick and effective transformation systems are still lacking for sugarcane.Here,we developed an efficient protoplast-based transformation system by optimizing conditions of protoplasts isolation and PEG-mediated transformation in S.spontaneum.The yield of viable protoplasts was approximately 1.26×107 per gram of leaf material,and the transformation efficiency of 80.19%could be achieved under the optimized condition.Furthermore,using this approach,the nuclear localization of an ABI5-like bZIPs transcription factor was validated, and the promoter activity of several putative DNase I hypersensitive sites(DHSs)was assessed.The results indicated this system can be conveniently applied to protein subcellular localization and promoter activity assays. A highly efficient S. spontaneum mesophyll cell protoplast isolation and transient transformation method was developed,and it shall be suitable for in vivo functional gene analysis in sugarcane.

1.Introduction

Sugarcane(Saccharum spp.)is a major crop grown for sugar and biofuel in tropical and subtropical regions around the world,providing 80% of the world's sugar and 40% of its ethanol production [1]. Modern sugarcane cultivars are polyploid interspecific hybrids with high sugar content derived from Saccharum officinarum and hardiness, disease resistance, and ratooning derived from Saccharum spontaneum L. [2]. Sugarcane's large genome size,aneuploidy of commercial cultivars,and polyploidy of interspecific hybrids have always imposed a challenge for sugarcane research and improvement.The genome sequence of S. spontaneum is recently released[1]. Along with the accumulation of genomic and transcriptomic resources,the availability of sugarcane genome sequences has provided an opportunity to conduct functional genomic studies in sugarcane. Although stable genetic transformation systems have been established in sugarcane using a number of methods including particle bombardment (biolistic) and Agrobacterium-mediated transformation, the low transformation efficiency (TE), recalcitrance to genetic transformation, difficulty in vitro regeneration, and the long lifespan make it inefficient and time-consuming for the functional characterization of genes[3-6].

Transient expression assays using plant protoplasts is a versatile and convenient technique for functional gene analysis, such as subcellular localization, protein interaction,gene functions, promoter activities, hybridization, and highthroughput analysis of gene expression[7,8].To date,optimization and establishment of protoplast-based transformation systems had been reported in diverse plant species,including rice [9,10], wheat [10], perennial ryegrass [7], pineapple [8],soybean [11], and Ma bamboo [12]. Although protoplast isolation and transformation have also been reported in sugarcane [13,14], the reported method yielded very low TE and poor reproducibility.

Here,we developed an efficient protoplast-based transformation system by optimizing conditions of protoplasts isolation and PEG-mediated transformation in S. spontaneum. The yield of viable protoplasts could get 1.26 × 107g?1, and the TE could get above 80%.In the present study,the subcellular localization of an ABA INSENSITIVE 5 (ABI5)-like base LEUCINE-ZIPPERs (bZIPs)transcription factor, which has been widely reported in Arabidopsis and rice [15,16], was confirmed by our system.Furthermore, the promoter function of several putative DNase I hypersensitive sites (DHSs) was also explored through the protoplast-based transient transformation assay. The results showed that the current method is a highly efficient S.spontaneum mesophyll cell protoplast isolation and transformation system,and it may be a versatile and convenient technique for functional gene analysis in sugarcane.

2.Materials and methods

2.1. Plant materials and growth conditions

Leaves of Saccharum spontaneum L. (SES208) were used for protoplasts isolation in this study. The stems of half-year-old sugarcane were collected to generate seedlings. The sugarcane seedlings were grown in a growth chamber with the temperature at 28/22°C(day/night)and 16/8 h light-dark cycle.

2.2. Plasmid preparation and constructs

The 35S: GFP and 35S enhancer: GFP plasmids were kindly provided by Prof.Wenli Zhang[17].The vector of 35S:YFP was kindly provided by Prof.Rensen Zeng.

The cDNA of tested genes was isolated from the cDNA library of S.spontaneum by PCR using the gene-specific primers,and the gene sequences were obtained by searching the S. spontaneum genomic database(http://www.life.illinois.edu/ming/downloads/Spontaneum_genome/).To generate 35S:SsABI5-YFP,full-length of the SsABI5-coding regions without the termination codon was inserted in-frame at NcoI/BamHI sites of 35S:YFP vector using the seamless cloning method. To generate constructs for validating the function of promoter, the genomic sequences of DHSs sequences were cloned into the vector 35S enhancer-GFP at HindIII/BamHI sites. The gene-specific primers used in this study are listed in Tables S1 and S2.

2.3. Protoplast isolation

Two youngest leaves of the 2-month-old sugarcane seedlings were collected,and only the tender basal sections were used as the starting material for protoplast isolation. Selected leaf sections (～1 g) were cut into 0.5 mm strips using a sharp razor,and all strips were immediately transferred into the 10 mL enzyme solution (10 mmol L?1MES KOH, pH 5.7, 2% (w/v)Cellulase‘Onozuka'R-10,0.5%(w/v)MacerozymeR-10,10 mmol L-?1CaCl2, 20 mmol L?1KCl, 0.1% BSA (w/v) and 0.6 mol L?1Mannitol). After vacuum treatment under dark for 30 min, the enzymatic digestion was carried out in the dark for 5 h with gentle shaking(25 r min?1)at room temperature.Then,10 mL W5 buffer(154 mmol L?1NaCl,125 mmol L?1CaCl2,5 mmol L?1KCl,2 mmol L?1MES KOH,pH 5.7)was added to stop the digestion,and the digestion mixture was filtered through a stainless 100 μmmesh sieve. The filtrate was centrifuged at 100 ×g (Heraeus Multifuge X3R, Termo Fisher Scientific) for 5 min. The pelleted protoplasts were re-suspended with W5 buffer and treated at 4°C for 30 min, and the purified protoplasts were harvested by centrifuging at 100×g for 5 min again.Harvested protoplasts were resuspended in 1 mL MMG solution (0.6 mol L?1Mannitol,15 mmol L?1MgCl2,4 mmol L?1MES KOH,pH 5.7)for subsequent PEG-mediated transfection. Each treatment included four repeats.

Cellulase‘Onozuka'R-10(181005-02)and MacerozymeR-10(171208-02) were obtained from Yakult Honsha, Japan. MES(4432-31-9), PEG-4000 (25322-68-3), D-Mannitol (Sigma, 69-65-8), Bovine Serum Albumin (BSA, 9048-46-8), CaCl2·2H2O(10035-04-8),KCl(7447-40-7),NaCl(7647-14-5),and MgCl2·6H2-O(7791-18-6) were obtained from Sigma-Aldrich, USA.

2.4. Protoplasts counting

Protoplasts numbers were counted using a hemocytometer(XB.K.25, QiuJing, Shanghai, China). Eight microliters of protoplast solution was added on the surface of the hemocytometer, and then the cover slide was laid carefully to avoid bubbles generation.The number of intact protoplasts in the four corners of the grid was counted under the microscope. The protoplast density was calculated: Protoplasts number mL?1= the average number of the intact cells in the four corners of the grid×104.

2.5. Protoplasts viability assessment

Fluoresce in diacetate(FDA)(596-09-8,Fushen)staining was used to determine the protoplast viability. FDA dissolved in acetones was added into protoplast solution at a final concentration of 0.05%. After incubation for 2 min, the viability of protoplasts was determined with fluorescence microscopy under ultraviolet light.The viable cells were stained green, whereas the dead cells and cell debris were not stainable. The viable protoplasts ratio was calculated: Viable protoplasts (%) = (green stained protoplasts determined under ultraviolet)/(total protoplasts observed under the bright field).

2.6. Protoplast transfection

For protoplast transfection, 100 μL of freshly isolated protoplasts in MMG solution and 30 μg of plasmids were mixed,followed by adding an equal volume of 40% (w/v) PEG solution(PEG 4000, 0.2 mol L?1Mannitol and 0.1 mol L?1CaCl2) and then the contents were mixed by gently inverting the tubes. After 15 min incubation in the dark, the protoplasts were gently washed by adding 1 mL W5 buffer and harvested by centrifuging at 100×g and then resuspended in 1 mL W5 buffer. The transfected protoplasts were incubated in the dark at 25 °C for 12-16 h. GFP or YFP signals were observed and recorded with a confocal microscope. Each treatment included four repeats.

2.7. Subcellular localization

For subcellular localization, 30 μg vector of 35S: SsABI5-eYFP and nucleus marker plasmid mCherry-ARF191V were cotransformed into protoplasts. After 12 h incubation, the YFP signal was observed using a confocal microscope. Each treatment included four repeats.

2.8. Microscopy

Counting of total and the viable protoplasts was performed with the ordinary optical microscope (Leica) and fluorescent microscope (OLYMPUS Model BX53, Tokyo, Japan) respectively. To observe the GFP and YFP fluorescence in the transformed protoplasts using a confocal microscope (Leica TCS SP8), the excitation wavelengths and emission filters sets were as follows: GFP, 488 nm (Ex)/BP505 to 530 nm (Em); YFP,514 nm (Ex)/BP525 to 550 nm (Em); Image analysis was conducted with the Leica LAS AF software.

3.Results

3.1. A high-efficiency method of protoplasts isolation from seedlings of S.spontaneum

To optimize an efficient transient expression system in sugarcane for functional gene analysis, selecting the proper source of leaf material is critical for obtaining a high yield of protoplasts. Young leaves of S. spontaneum were used for the evaluation of protoplast isolation. The outermost 1-2 leaf blades and sheathes were discarded,and the youngest leaves without visible dewlap and innermost were retained. To obtain the most suitable leave segments for protoplast generation, we divided the selected young leaves into three regions (i, ii and iii) to compare the digestive efficiency (Fig.1A). The yield of protoplasts from section i was significantly higher than those of section ii and iii (Fig. 1B, C). Thus, the basal segment of young leaves is the best source material for protoplast isolation in S. spontaneum. Moreover, we also compared the protoplast yields of the basal segments from S. spontaneum seedlings at different growth stages from 1 month to 6 months (Fig. 1D-F). The results showed that protoplasts isolated from one month-old plants were significantly fewer than those from other plants (Fig. 1F). Considering the time consumed for material preparation, the twomonth-old seedlings might be the most suitable source material.

To obtain a high yield of live protoplasts in sugarcane,the effects of different concentrations of Macerozyme R10 and Cellulase R10(Table 1)on the protoplast quantity and quality were tested with the basal segment of young leaves in twomonth-old seedlings as source material. After digestion for 4 h, the numbers of intact protoplasts and live protoplasts were counted(Fig.2A). Among the series of enzyme solution,while there were more total protoplasts (12.73 × 106protoplasts g?1fresh weight) generated in the combination of 2.5%Cellulase R-10 and 0.5% Macerozyme R-10, the combinations of 2% Cellulase R10 with 0.5% Macerozyme R10 led to the highest yield of viable protoplasts (9.69 × 106protoplasts g?1fresh weight)(Fig.2B).

Next, we analyzed the effects of enzymatic digestion time(1-24 h) on protoplast isolation efficiency with optimal concentrations of digestion enzymes.In the beginning,the yields of the protoplasts increased significantly and peaked to 16.67 × 106protoplasts g?1fresh weight at 5 h. Then, it tended to decrease with digestion time. For the viable protoplasts, its yield also peaked at 5 h with 12.61 × 106protoplasts g?1fresh weight(Fig. 2C). Overall, 2% Cellulase R10 and 0.5% Macerozyme R10 is the optimal combination of enzymes solution, and 5 h is the suitable digestion time for protoplast isolation in S.spontaneum.

3.2.Effects of PEG concentration on protoplast TE and viability

PEG is widely used to promote DNA uptake in protoplast transformation. With the optimal enzymes concentration and digestion time, we further detected the effects of different concentrations of PEG (10% to 80%) on protoplast TE. After 15 min PEG incubation and 12 h cultivation, 40% and 50% PEG mediated the highest TE up to 65% (Fig. 3B). Besides, we also evaluated the effects of PEG concentration on protoplast viability.The numbers of both total intact protoplasts and viable protoplasts tended to decrease as the PEG concentration increased(Fig.3C),which might be caused by the PEG induced high permeability.Therefore, the optimal PEG concentration for sugarcane protoplast transformation is 40%,at which the highest TE and proper viable protoplasts could be achieved.

3.3. Effects of PEG incubation time on protoplast TE and viability

To optimize the PEG incubation time,we evaluated the influence of different PEG incubation times on TE and protoplast viability with the optimal PEG concentration.The results showed that the TE could get 66.7% after incubation for 15 min.With incubation time prolonging,the TE did not change obviously within 35 min and reduced to 44.8% when incubated for 45 min (Fig. 4A).Besides,we found that the numbers of both total protoplasts and viable protoplasts were negatively correlated with the PEG incubation time. Therefore, 15 min was selected as the optimal PEG incubation time.

Fig.1-Effects of leaf region and age of Saccharum spontaneum L.seedlings on protoplasts isolation.(A)Different leave segments of seedlings for preparation of protoplasts.The youngest and innermost leaves of 1-month-old seedlings were used for protoplasts isolation.Leaves were divided into three regions(I,basal region;ii,middle region;iii,upper region).Scale bar,5 cm.(B)The protoplasts isolated from different regions under bright field of microscope.Scale bars,50 μm.(C)Protoplasts counting of(B).(D)Representative 1-month-old(1 m),2-month-old(2 m),3-month-old(3 m)and 6-month-old(6 m)seedlings.Scale bar,5 cm.(E)The basal regions of young leaves from(D)for protoplasts isolation.Scale bar,1 cm.(F)The total number of protoplasts isolated from(E).Values represent means±SE(n=4).The different letters indicate statistically significant differences at P<0.05.

3.4. Effects of DNA amount on protoplast TE

?

Fig.2-Effect of different combinations of enzymes and digestion times on Saccharum spontaneum L.protoplast isolation.(A)Micrograph of viable protoplasts stained with FDA under fluorescence field(left),total protoplasts under bright field(middle),and merged(right).Red arrows indicate viable cells,and black arrows indicate inactive cells.Scale bars,50 μm.(B)Yield of protoplast isolation from different enzyme combinations.(C)Yield of protoplast isolation from different enzyme digestion time.Blue bars and yellow bars present total intact cells and activate cells,respectively.At least 50 protoplasts were counted in one scope,and values represent means±SE(n=4).The different letters indicate significant differences at P<0.05.

The amount of plasmids is critical for protoplast TE[18].Thus,we test the effect of the plasmid amount on protoplast TE.Using the optimized conditions (40% PEG, incubated for 15 min), we examined the effects of different amounts of 35S: GFP plasmids on TE of S.spontaneum protoplasts.The results showed that the TE increased up to 80%with the increasing amount of plasmids from 5 to 30 μg, but did not increase when the amount of plasmids increased from 30 to 50 μg(Fig.4C),indicating that 30 μg plasmids might be the threshold of highest TE and was considered to be the optimal amount of plasmids in the present system.

3.5. Protein localization assay using a protoplast-based transient expression system

In order to test the reliability of the optimized S. spontaneum system in protein subcellular localization and functional gene assays,we verified the subcellular localization of SsABI5.ABI5-like bZIPs are nucleus localization transcription factors.Those have been widely reported in Arabidopsis, rice, and other plants [15,16]. In this study, the SsABI5 (Sspon.03G0002290-2B)gene with 65.2% similarity to OsABI5 was cloned from S.spontaneum.Protein sequence prediction by LOCALIZER(http://localizer.csiro.au) showed that SsABI5 contains a predicted NLS peptide. Therefore, we examined the localization of SsABI5 in protoplast using 35S: SsABI5-YFP transformation(Fig. 5A). The YFP signal could be observed in both cytoplasm and nucleus when transfected with 35S:YFP vector and only in the nucleus of protoplast transfected with 35S: SsABI5-YFP,indicated by RFP signal of nucleus location marker mCherry-ARF191V(Fig.5B).

Fig.3- Effect of PEG concentration on protoplast transfection.(A)Micrographs of protoplasts expressing 35S: GFP under GFP field(left),bright field(middle)and merged(right).Red arrows indicate the fluorescent protoplasts that are successfully transfected.Black arrows indicate protoplasts that are not transfected.Scale bars,50 μm.(B)The transformation efficiency(TE)of protoplasts cultivated with various concentrations of PEG.TE was calculated after 12 h cultivation.(C)The effects of PEG concentration on the number of protoplasts. The number of total protoplasts and viable protoplasts was counted after 12 h cultivation.Values represent means±SE (n = 4).The different letters indicate significant differences at P<0.05.

3.6. Validation of promoter activity using protoplast-based transformation system

Understanding the core promoter function is significant for investigating the gene regulatory mechanisms. In this study,we randomly selected several DNase I hypersensitive sites(DHSs) from our DNase-seq data (unpublished observations)of S. spontaneum leaves to validate their potential promoter activities through our protoplast transformation system(Table S2). The DHSs sequences were inserted into a 35S enhancer vector to validate their promoter function, as previously reported [17]. While the positive control, 35S: GFP,induced strong GFP signal,no GFP signal could be observed in the negative control, 35S enhance-GFP (Fig. 6B). Among the 6 DHSs examined, DHS1 (Chr3C:75712351_75712543) and DHS2 (Chr1C:95132890_95133036) induced the expression of GFP signal. These results suggested that the developed protoplast transient transfection system could also be used for functional validation of promoter activity in sugarcane.

Fig.4-Effects of PEG incubation time and plasmid amount on protoplast transfection.(A)The transformation efficiency(TE)of protoplasts with different PEG incubation time.(B)Effect of PEG incubation time on the numbers of total protoplasts and viable protoplasts.(C)Effect of plasmid amount on protoplast transfection.The protoplasts TE was evaluated after incubation in 40%PEG solution for 15 min.Values represent means± SE(n =4). The different letters indicate significant differences at P <0.05.

4.Discussion

Generating high-quality protoplasts depends on the proper leaf materials and the composition of the enzyme solution,which is essential for subsequent transformation and functional gene analyses[19,20].The middle leave region of young seedlings is always the priority selection for protoplasts isolation in monocotyledonous plants, such as wheat and rice [9,10,20]. Our results showed that the best choice for protoplast isolation is the base region of inner young leaves,which are soft and easy for protoplast isolation (Fig. 1C).Furthermore, the protoplast isolation is not limited in young seedling, and even the 2-month-old plants or the older ones can get more protoplasts(Fig.1F).The digestive enzyme mix is critical to ensure a high yield and integrity of protoplasts. In the previous protoplast isolation protocol of other plants,the optimal Cellulase R10 concentration is between 1.5% and 1%[8,10,11,21].However,in the present study,it is found that 2%of Cellulase R10 was optimal for obtaining the most viable protoplasts in sugarcane. This might be caused by the higher cellulose content in the cell wall of the S. spontaneum leaves.The optimal digestion time also varies depending on the plant species. For example, the optimal digestion time is 3 h for pineapple and Brachypodium[8,21],and 4 h for wheat[10].Our study showed that the optimal digestion time was 5 h for sugarcane,and the maximum yields of viable protoplasts was 12.61 × 106protoplasts g?1fresh weight (Fig. 2C), which is higher than those reported in wheat and soybean (～7 × 106protoplasts g?1fresh weight)[10,11].

Fig.5-Subcellular localization of SsABI5 in Saccharum spontaneum L.protoplasts.(A)Schematic diagram of SsABI5 construction for transient expression.35S,35S promoter;E,enhancer;T,terminate.(B)Subcellular localization of free YFP and SsABI5-YFP in protoplasts. mCherry-ARF191V is a nucleus location marker with RFP signal.Scale bars,10 μm.

PEG promotes the uptake of DNA into protoplasts during transformation, but the too high concentration of PEG could reduce TE due to its toxicity to protoplasts[18].The 40%PEG is widely applied to PEG-mediated protoplasts transformation,while 20%-40% PEG is also used in some well-established protocols [21,22]. Our data indicated that using both 40% and 50% PEG could achieve a high TE, but viable protoplasts tended to decrease with the increasing PEG concentration,and 50%PEG resulted in lower viable protoplasts than 40%PEG(Fig. 3). Therefore, in consistent with the previous PEG-mediated protoplast transformation methods, the optimal PEG concentration for sugarcane protoplast transformation is also 40%, at which the highest TE and proper viable protoplasts could be achieved.

Another important factor influencing TE is the quantity of plasmid and the optimal dosage of plasmid for protoplast transformation. Different amounts of plasmids, such as 2 μg for tobacco, 10 μg for soybean, and 20 μg for Brachypodium distachyon, have been reported to be the optimal amounts of DNA in their established protocols respectively[11,21,23].Our assay demonstrated an increase TE could be obtained with an increase in plasmid amount in S. spontaneum, but it got a plateau at 30 μg(Fig.4C).This plateau tendency has also been reported previously[18].Thus,30 μg was considered to be the optimal amount of plasmids for the present sugarcane protoplast transformation.

Given the difficulty and time consuming of the stable transformation in sugarcane, the transient expression in the protoplast system is an ideal tool for functional gene identification in its homologous system. In the current study, the nuclear localization of SsABI5, a predicted ABI5-like bZIPs transcription factor with NLS peptide, was confirmed by our protoplast-based transformation system(Fig.5).bZIPs transcription factors have been widely reported to localize in the nucleus in Arabidopsis and rice [15,16].Moreover, the protoplast-based transient transformation assay has been used to validate the promoter and/or enhancer functions of maize DNase I hypersensitivity sites (DHSs),which are typically associated with cis-regulatory elements such as enhancers and promoters [17,19]. Here, the potential promoter function of DHSs could also be explored through our system (Fig. 6). These results indicated that the developed system could be efficiently applied to protein subcellular localization and promoter activity assays. Besides, protoplast transfection systems are also commonly used for many other molecular biology analyses, including in vivo protein-protein and protein-DNA interactions, protein tracking, and signal transduction, etc. [8,12]. Therefore, this protoplast isolation and transformation system can be easily used for protein subcellular localization and promoter function validation and shall be a versatile and convenient homologous system for in vivo functional gene analysis in sugarcane.

Fig.6- Promoter function validation of DNase I hypersensitive sites(DHSs)in Saccharum spontaneum L. protoplasts.(A)Schematic diagram of constructions for promoter function test of DHSs.(B)Functional test of DHSs in protoplasts. Protoplasts transfected with 35S:GFP was used as a positive control,in which GFP expression was driven by 35S promoter.Protoplasts transfected with 35S enhancer:GFP was used as a negative control,in which GFP expression could not be driven due to the lacking of core promoter. DHS1 and DHS2 represent the DHS sequence.Scale bars,50 μm.

5.Conclusions

A highly efficient S. spontaneum mesophyll cell protoplast isolation and transformation system was developed.With the basal segment of young leaves in two-month-old seedlings as source material, digestion in enzyme solution with 2%Cellulase R10 and 0.5%Macerozyme R10 for 5 h is the optimal condition for S. spontaneum protoplast isolation, in which 12.61 × 106protoplasts g?1fresh weight could be readily achieved. For the PEG-mediated protoplast transformation,the TE could be up to 80.19% with 30 μg plasmids incubating with the 40%PEG for 15 min.Moreover,the developed method could be a convenient technique for protein subcellular localization, promoter function validation, and many other molecular biology studies in sugarcane.

Declaration of competing interest

The authors declare that they have no competing interests.

Acknowledgments

This research was funded by the National Natural Science Foundation of China(3190020451 and 31771862).We thank Prof.Rensen Zeng of Fujian Agriculture and Forestry University for providing the vector for protein subcellular localization and Prof.Wenli Zhang from Nanjing Agriculture University for providing the vector for promoter function validation.

Author contributions

Qiongli Wang designed the experiment and wrote the manuscript. Zhiyong Chen and Yufang Hu performed vector construction. Guangrun Yu performed protoplast isolation and transformation experiments. Qiongli Wang and Jinlei Han analyzed the data.Kai Wang revised the manuscript.All authors read and approved the submission of this manuscript.

Appendix A. Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2020.05.006.