Testing plasmid stability ofEscherichia coliusing the Continuously Operated Shaken BIOreactor System

Adaptive laboratory evolution in a novel parallel shaken pH-auxostat

Adaptive laboratory evolution (ALE) is a widely used microbial strain development and optimization method. ALE experiments, to select for faster-growing strains, are commonly performed as serial batch cultivations in shake flasks, serum bottles, or microtiter plates or as continuous cultivations in bioreactors on a laboratory scale. To combine the advantages of higher throughput in parallel shaken cultures with continuous fermentations for conducting ALE experiments, a new Continuous parallel shaken pH-auxostat (CPA) was developed. The CPA consists of six autonomous parallel shaken cylindrical reactors, equipped with real-time pH control of the culture medium. The noninvasive pH measurement and control are realized by biocompatible pH sensor spots and a programmable pump module, to adjust the dilution rate of fresh medium for each reactor separately. Two different strains of the methylotrophic yeast Ogataea polymorpha were used as microbial model systems for parallel chemostat and pH-auxostat cultivations. During cultivation, the medium is acidified by the microbial activity of the yeast. For pH-auxostat cultivations, the growth-dependent acidification triggers the addition of fresh feed medium into the reactors, leading to a pH increase and thereby to the control of the pH to a predetermined set value. By controlling the pH to a predetermined set value, the dilution rate of the continuous cultivation is adjusted to values close to the washout point, in the range of the maximum specific growth rate of the yeast. The pH control was optimized by conducting a step-response experiment and obtaining tuned PI controller parameters by the Chien-Hrones-Reswick (CHR) PID tuning method. Two pH-auxostat cultivations were performed with two different O. polymorpha strains at high dilution rates for up to 18 days. As a result, up to 4.8-fold faster-growing strains were selected. The increased specific maximum growth rates of the selected strains were confirmed in subsequent batch cultivations.

Tips for efficiently maintaining pET expression plasmids

pET expression plasmids are widely used for producing recombinant proteins in Escherichia coli. Selection and maintenance of cells harboring a pET plasmid are possible using either a Tn3.1-type genetic fragment (which encodes a ß-lactamase and confers resistance to ß-lactam antibiotics) or a Tn903.1-type genetic fragment (which encodes an aminoglycoside-3'-phosphotransferase and confers resistance aminoglycoside antibiotics). Herein we have investigated how efficiently pET plasmids are maintained using these two fragments. The study reveals that pET plasmids are efficiently maintained with both Tn3.1 and Tn903.1 genetic fragments prior to the induction of recombinant protein production, and over short induction times (i.e., 2 h). However, over longer induction times (i.e., 20 h), the efficiency of plasmid maintenance depends on the host strain used, and the type of antibiotic selection cassette used. Based on our collective observations, we have 2 general tips for efficiently maintaining pET plasmids during recombinant production experiments. Tip #1: Use a strain with lowered levels of the T7 RNA polymerase, such as C41(DE3). pET plasmids will be efficiently maintained over long induction times with both the Tn3.1 and Tn903.1 genetic fragments, regardless of whether antibiotics are present during cultivation. Tip #2: If a strain with higher levels of T7 RNA polymerase strain is necessary, such as BL21(DE3)), keep induction times short or use a plasmid containing a Tn903.1-type fragment and select with kanamycin.

Chromosomal engineering of inducible isopropanol- butanol-ethanol production in Clostridium acetobutylicum

The use of environmentally damaging petrochemical feedstocks can be displaced by fermentation processes based on engineered microbial chassis that recycle biomass-derived carbon into chemicals and fuels. The stable retention of introduced genes, designed to extend product range and/or increase productivity, is essential. Accordingly, we have created multiply marked auxotrophic strains of Clostridium acetobutylicum that provide distinct loci (pyrE, argH, purD, pheA) at which heterologous genes can be rapidly integrated using allele-coupled exchange (ACE). For each locus, ACE-mediated insertion is conveniently selected on the basis of the restoration of prototrophy on minimal media. The Clostridioides difficile gene (tcdR) encoding an orthogonal sigma factor (TcdR) was integrated at the pyrE locus under the control of the lactose-inducible, bgaR::P_bgaL promoter to allow the simultaneous control of genes/operons inserted at other disparate loci (purD and pheA) that had been placed under the control of the P_tcdB promoter. In control experiments, dose-dependent expression of a catP reporter gene was observed with increasing lactose concentration. At the highest doses tested (10 mM) the level of expression was over 10-fold higher than if catP was placed directly under the control of bgaR::P_bgaL and over 2-fold greater than achieved using the strong P_fdx promoter of the Clostridium sporogenes ferredoxin gene. The utility of the system was demonstrated in the production of isopropanol by the C. acetobutylicum strain carrying an integrated copy of tcdR following the insertion of a synthetic acetone operon (ctfA/B, adc) at the purD locus and a gene (sadh) encoding a secondary dehydrogenase at pheA. Lactose induction (10 mM) resulted in the production of 4.4 g/L isopropanol and 19.8 g/L Isopropanol-Butanol-Ethanol mixture.

Novel insights into construct toxicity, strain optimization, and primary sequence design for producing recombinant silk fibroin and elastin-like peptide in E. coli

Spider silk proteins (spidroins) are a remarkable class of biomaterials that exhibit a unique combination of high-value attributes and can be processed into numerous morphologies for targeted applications in diverse fields. Recombinant production of spidroins represents the most promising route towards establishing the industrial production of the material, however, recombinant spider silk production suffers from fundamental difficulties that includes low titers, plasmid instability, and translational inefficiencies. In this work, we sought to gain a deeper understanding of upstream bottlenecks that exist in the field through the production of a panel of systematically varied spidroin sequences in multiple E. coli strains. A restriction on basal expression and specific genetic mutations related to stress responses were identified as primary factors that facilitated higher titers of the recombinant silk constructs. Using these findings, a novel strain of E. coli was created that produces recombinant silk constructs at levels 4-33 times higher than standard BL21(DE3). However, these findings did not extend to a similar recombinant protein, an elastin-like peptide. It was found that the recombinant silk proteins, but not the elastin-like peptide, exert toxicity on the E. coli host system, possibly through their high degree of intrinsic disorder. Along with strain engineering, a bioprocess design that utilizes longer culturing times and attenuated induction was found to raise recombinant silk titers by seven-fold and mitigate toxicity. Targeted alteration to the primary sequence of the recombinant silk constructs was also found to mitigate toxicity. These findings identify multiple points of focus for future work seeking to further optimize the recombinant production of silk proteins and is the first work to identify the intrinsic disorder and subsequent toxicity of certain spidroin constructs as a primary factor related to the difficulties of production.

Engineering Escherichia coli for l-homoserine production

l-homoserine, a nonprotein amino acid, is used to synthesize many active substances in the industry. Here, to develop a robust l-homoserine-producing strain, Escherichia coli W3110 was used as a chassis to be engineered. Based on a previous construct with blocked competing routes for l-homoserine synthesis, five genes were overexpressed by promoter replacement strategy to increase the l-homoserine production, including enhancement of precursors for l-homoserine synthesis (ppc, thrA, and asd), reinforcement of the NADPH supply (pntAB) and efflux transporters (rhtA) to improve the l-homoserine production. However, the plasmid losing was to blame for the wildly fluctuating fermentation performance of engineered strains, ranging between 2.1 and 6.2 g/L. Then, a hok/sok toxin/antitoxin system was introduced into the free plasmid expression cassette to maintain the genetic stability of the episomal plasmid; consequently, the plasmid-losing rate sharply decreased, resulting in the engineered strain SHL17, which exhibited excellent stability in l-homoserine production, with 6.3 g/L in shake flasks and 44.4 g/L in a 5-L fermenter without antibiotic addition. This work verified the effective use of the hok/sok toxin/antitoxin system combined with promoter engineering to improve the genetic stability of E. coli episomal plasmids without antibiotics.

Construction of constitutive expression of Eimeria tenella eukaryotic initiation factor U6L5H2 on the surface of Lactobacillus plantarum and evaluation of its immunoprotective efficiency against chicken coccidiosis

Lactobacillus strains exhibit preferable properties that make them attractive candidates for vaccine delivery systems because of their ability to regulate intestinal mucosal immunity in the body. To date, live Lactobacillus delivery vaccines reported for the defense against Eimeria tenella have been inducer-dependent systems whose applications are significantly limited due to their unattainable induction conditions in vivo. Here, a constitutive expression of Lactobacillus plantarum NC8 surface display system was constructed. Then, this system was used to prepare a live oral vaccine to constitutively express the E. tenella U6L5H2 (EtU6) protein on the NC8 surface and to evaluate its protective efficacy against E. tenella challenge in chickens. The results showed that the heterologous protein (EGFP or EtU6) was successfully expressed on the surface of L. plantarum NC8 without any inducer. The immunoprotection of EtU6 with constitutive expression in L. plantarum NC8 system (NC8/Pc-EtU6) was significantly stronger than that of EtU6 with induced expression of L. plantarum NC8 system (NC8/Pi-EtU6) (ACI: 168.28 vs. 152.74) as evidenced by increased body weight, decreased oocyst output and lesion scores. Furthermore, the constitutive system NC8/Pc-EtU6 produced higher levels of specific cecal SIgA, serum IgG, transcription of cytokines IFN-γ and IL-2, and lymphocyte proliferation than the induced system NC8/Pi-EtU6. These results indicate that, compared to the inducible system, the constitutive surface display system of L. plantarum has the advantages of continuously expressing antigens in vivo and stimulating the host immune system. It could be an ideal platform for vaccine expression. The live vector vaccine for coccidiosis constructed by this constitutive system greatly improves the application potential in chicken production and provides a novel platform for the prevention of coccidiosis in chickens.

Efficient strategies to enhance plasmid stability for fermentation of recombinant Escherichia coli harboring tyrosine phenol lyase

OBJECTIVE

To solve the bottleneck of plasmid instability during microbial fermentation of L-DOPA with recombinant Escherichia coli expressing heterologous tyrosine phenol lyase.

RESULTS

The tyrosine phenol lyase from Fusobacterium nucleatum was constitutively expressed in E. coli and a fed-batch fermentation process with temperature down-shift cultivation was performed. Efficient strategies including replacing the original ampicillin resistance gene, as well as inserting cer site that is active for resolving plasmid multimers were applied. As a result, the plasmid stability was increased. The co-use of cer site on plasmid and kanamycin in culture medium resulted in proportion of plasmid containing cells maintained at 100% after fermentation for 35 h. The specific activity of tyrosine phenol lyase reached 1493 U/g dcw, while the volumetric activity increased from 2943 to 14,408 U/L for L-DOPA biosynthesis.

CONCLUSIONS

The established strategies for plasmid stability is not only promoted the applicability of the recombinant cells for L-DOPA production, but also provides important guidance for industrial fermentation with improved microbial productivity.

Repetitive Fed-Batch: A Promising Process Mode for Biomanufacturing With E. coli

Recombinant protein production with Escherichia coli is usually carried out in fed-batch mode in industry. As set-up and cleaning of equipment are time- and cost-intensive, it would be economically and environmentally favorable to reduce the number of these procedures. Switching from fed-batch to continuous biomanufacturing with microbials is not yet applied as these cultivations still suffer from time-dependent variations in productivity. Repetitive fed-batch process technology facilitates critical equipment usage, reduces the environmental fingerprint and potentially increases the overall space-time yield. Surprisingly, studies on repetitive fed-batch processes for recombinant protein production can be found for yeasts only. Knowledge on repetitive fed-batch cultivation technology for recombinant protein production in E. coli is not available until now. In this study, a mixed feed approach, enabling repetitive fed-batch technology for recombinant protein production in E. coli, was developed. Effects of the cultivation mode on the space-time yield for a single-cycle fed-batch, a two-cycle repetitive fed-batch, a three-cycle repetitive fed batch and a chemostat cultivation were investigated. For that purpose, we used two different E. coli strains, expressing a model protein in the cytoplasm or in the periplasm, respectively. Our results demonstrate that a repetitive fed-batch for E. coli leads to a higher space-time yield compared to a single-cycle fed-batch and can potentially outperform continuous biomanufacturing. For the first time, we were able to show that repetitive fed-batch technology is highly suitable for recombinant protein production in E. coli using our mixed feeding approach, as it potentially (i) improves product throughput by using critical equipment to its full capacity and (ii) allows implementation of a more economic process by reducing cleaning and set-up times.

Optimizing the linker length for fusing an alcohol dehydrogenase with a cyclohexanone monooxygenase

The use of enzymes in organic synthesis is highly appealing due their remarkably high chemo-, regio- and enantioselectivity. Nevertheless, for biosynthetic routes to be industrially useful, the enzymes must fulfill several requirements. Particularly, in case of cofactor-dependent enzymes self-sufficient systems are highly valuable. This can be achieved by fusing enzymes with complementary cofactor dependency. Such bifunctional enzymes are also relatively easy to handle, may enhance stability, and promote product intermediate channeling. However, usually the characteristics of the linker, fusing the target enzymes, are not thoroughly evaluated. A poor linker design can lead to detrimental effects on expression levels, enzyme stability and/or enzyme performance. In this chapter, the effect of the length of a glycine-rich linker was explored for the case study of ɛ-caprolactone synthesis through an alcohol dehydrogenase-cyclohexanone monooxygenase fusion system. The procedure includes cloning of linker variants, expression analysis, determination of thermostability and effect on activity and conversion levels of 15 variants of different linker sizes. The protocols can also be used for the creation of other protein-protein fusions.

The Lazarus Escherichia coli Effect: Recovery of Productivity on Glycerol/Lactose Mixed Feed in Continuous Biomanufacturing

Continuous cultivation with Escherichia coli has several benefits compared to classical fed-batch cultivation. The economic benefits would be a stable process, which leads to time independent quality of the product, and hence ease the downstream process. However, continuous biomanufacturing with E. coli is known to exhibit a drop of productivity after about 4–5 days of cultivation depending on dilution rate. These cultivations are generally performed on glucose, being the favorite carbon source for E. coli and used in combination with isopropyl β-D-1 thiogalactopyranoside (IPTG) for induction. In recent works, harsh induction with IPTG was changed to softer induction using lactose for T7-based plasmids, with the result of reducing the metabolic stress and tunability of productivity. These mixed feed systems based on glucose and lactose result in high amounts of correctly folded protein. In this study we used different mixed feed systems with glucose/lactose and glycerol/lactose to investigate productivity of E. coli based chemostats. We tested different strains producing three model proteins, with the final aim of a stable long-time protein expression. While glucose fed chemostats showed the well-known drop in productivity after a certain process time, glycerol fed cultivations recovered productivity after about 150 h of induction, which corresponds to around 30 generation times. We want to further highlight that the cellular response upon galactose utilization in E. coli BL21(DE3), might be causing fluctuating productivity, as galactose is referred to be a weak inducer. This “Lazarus” phenomenon has not been described in literature before and may enable a stabilization of continuous cultivation with E. coli using different carbon sources.

Exploring small-scale chemostats to scale up microbial processes: 3-hydroxypropionic acid production in S. cerevisiae

BACKGROUND

The physiological characterization of microorganisms provides valuable information for bioprocess development. Chemostat cultivations are a powerful tool for this purpose, as they allow defined changes to one single parameter at a time, which is most commonly the growth rate. The subsequent establishment of a steady state then permits constant variables enabling the acquisition of reproducible data sets for comparing microbial performance under different conditions. We performed physiological characterizations of a 3-hydroxypropionic acid (3-HP) producing Saccharomyces cerevisiae strain in a miniaturized and parallelized chemostat cultivation system. The physiological conditions under investigation were various growth rates controlled by different nutrient limitations (C, N, P). Based on the cultivation parameters obtained subsequent fed-batch cultivations were designed.

RESULTS

We report technical advancements of a small-scale chemostat cultivation system and its applicability for reliable strain screening under different physiological conditions, i.e. varying dilution rates and different substrate limitations (C, N, P). Exploring the performance of an engineered 3-HP producing S. cerevisiae strain under carbon-limiting conditions revealed the highest 3-HP yields per substrate and biomass of 16.6 %C-mol and 0.43 g gCDW^-1, respectively, at the lowest set dilution rate of 0.04 h^-1. 3-HP production was further optimized by applying N- and P-limiting conditions, which resulted in a further increase in 3-HP yields revealing values of 21.1 %C-mol and 0.50 g gCDW^-1 under phosphate-limiting conditions. The corresponding parameters favoring an increased 3-HP production, i.e. dilution rate as well as C- and P-limiting conditions, were transferred from the small-scale chemostat cultivation system to 1-L bench-top fermenters operating in fed-batch conditions, revealing 3-HP yields of 15.9 %C-mol and 0.45 g gCDW^-1 under C-limiting, as well as 25.6 %C-mol and 0.50 g gCDW^-1 under phosphate-limiting conditions.

CONCLUSIONS

Small-scale chemostat cultures are well suited for the physiological characterization of microorganisms, particularly for investigating the effect of changing cultivation parameters on microbial performance. In our study, optimal conditions for 3-HP production comprised (i) a low dilution rate of 0.04 h^-1 under carbon-limiting conditions and (ii) the use of phosphate-limiting conditions. Similar 3-HP yields were achieved in chemostat and fed-batch cultures under both C- and P-limiting conditions proving the growth rate as robust parameter for process transfer and thus the small-scale chemostat system as powerful tool for process optimization.

Engineering of multiple modular pathways for high-yield production of 5-aminolevulinic acid in Escherichia coli

5-aminolevulinic acid (ALA), an important precursor of tetrapyrroles, has various applications in medicine and agriculture fields. Several methods have been adopted to enhance ALA synthesis in our previous studies. In this study, systematic metabolic engineering strategies were implemented to further improve ALA production in Escherichia coli. Firstly, hemA and hemL with different strength of RBS from the artificially constructed mutation libraries were randomly assembled to balance metabolic flux. Then the expression of ALA dehydratase was rationally regulated by replacing promoter with fliCp to weaken ALA catabolism. Besides, the activity of glutamate-1-semialdehyde aminotransferase was increased through strengthening the native biosynthesis pathway of cofactor pyridoxal 5'-phosphate. Moreover, plasmid stability was improved by 21.4% by deleting recA and endA in the recombinant. Finally, stepwise improvements in ALA production were increased to 5.25 g/L with a pH two-stage strategy in a 3-L fermenter. This study proved the importance of metabolic balance in the pathway.

A minimum information standard for reproducing bench-scale bacterial cell growth and productivity

Reproducing, exchanging, comparing, and building on each other's work is foundational to technological advances. Advancing biotechnology calls for reliable reuse of engineered organisms. Reliable reuse of engineered organisms requires reproducible growth and productivity. Here, we identify the experimental factors that have the greatest effect on the growth and productivity of our engineered organisms in order to demonstrate reproducibility for biotechnology. We present a draft of a Minimum Information Standard for Engineered Organism Experiments (MIEO) based on this method. We evaluate the effect of 22 factors on Escherichia coli engineered to produce the small molecule lycopene, and 18 factors on E. coli engineered to produce red fluorescent protein. Container geometry and shaking have the greatest effect on product titer and yield. We reproduce our results under two different conditions of reproducibility: conditions of use (different fractional factorial experiments), and time (48 biological replicates performed on 12 different days over 4 months).

Elucidation of auxotrophic deficiencies of Bacillus pumilus DSM 18097 to develop a defined minimal medium

Background

Culture media containing complex compounds like yeast extract or peptone show numerous disadvantages. The chemical composition of the complex compounds is prone to significant variations from batch to batch and quality control is difficult. Therefore, the use of chemically defined media receives more and more attention in commercial fermentations. This concept results in better reproducibility, it simplifies downstream processing of secreted products and enable rapid scale-up. Culturing bacteria with unknown auxotrophies in chemically defined media is challenging and often not possible without an extensive trial-and-error approach. In this study, a respiration activity monitoring system for shake flasks and its recent version for microtiter plates were used to clarify unknown auxotrophic deficiencies in the model organism Bacillus pumilus DSM 18097.

Results

Bacillus pumilus DSM 18097 was unable to grow in a mineral medium without the addition of complex compounds. Therefore, a rich chemically defined minimal medium was tested containing basically all vitamins, amino acids and nucleobases, which are essential ingredients of complex components. The strain was successfully cultivated in this medium. By monitoring of the respiration activity, nutrients were supplemented to and omitted from the rich chemically defined medium in a rational way, thus enabling a systematic and fast determination of the auxotrophic deficiencies. Experiments have shown that the investigated strain requires amino acids, especially cysteine or histidine and the vitamin biotin for growth.

Conclusions

The introduced method allows an efficient and rapid identification of unknown auxotrophic deficiencies and can be used to develop a simple chemically defined tailor-made medium. B. pumilus DSM 18097 was chosen as a model organism to demonstrate the method. However, the method is generally suitable for a wide range of microorganisms. By combining a systematic combinatorial approach based on monitoring the respiration activity with cultivation in microtiter plates, high throughput experiments with high information content can be conducted. This approach facilitates media development, strain characterization and cultivation of fastidious microorganisms in chemically defined minimal media while simultaneously reducing the experimental effort.

Electronic supplementary material

The online version of this article (10.1186/s12934-018-0956-1) contains supplementary material, which is available to authorized users.

Parallel substrate supply and pH stabilization for optimal screening of E. coli with the membrane-based fed-batch shake flask

Background

Screening in the fed-batch operation mode is essential for biological cultivations facing challenges as oxygen limitation, osmotic inhibition, catabolite repression, substrate inhibition or overflow metabolism. As a screening tool on shake flask level, the membrane-based fed-batch shake flask was developed. While a controlled supply of a substrate was realized with the in-built membrane tip, the possibilities for replenishing nutrients and stabilizing pH values was not yet exploited. High buffer concentrations were initially used, shifting the medium osmolality out of the biological optimum. As the growth rate is predefined by the glucose release kinetics from the reservoir, the resulting medium acidification can be compensated with a controlled continuous supply of an alkaline compound. The focus of this research is to establish a simultaneous multi-component release of glucose and an alkaline compound from the reservoir to enable cultivations within the optimal physiological range of Escherichia coli.

Results

In combination with the Respiratory Activity MOnitoring System, the membrane-based fed-batch shake flask enabled the detection of an ammonium limitation. The multi-component release of ammonium carbonate along with glucose from the reservoir resulted not only in the replenishment of the nitrogen source but also in the stabilization of the pH value in the culture medium. A biomass concentration up to 25 g/L was achieved, which is one of the highest values obtained so far to the best of the author’s knowledge with the utilization of a shake flask and a defined synthetic medium. Going a step further, the pH stabilization allowed the decrease of the required buffer amount to one-fourth establishing an optimal osmolality range for cultivation. As optimal physiological conditions were implemented with the multi-component release fed-batch cultivation, the supply of 0.2 g glucose in a 10 mL initial culture medium volume with 50 mM MOPS buffer resulted in a twofold higher biomass concentration than in a comparable batch cultivation.

Conclusions

The newly introduced multi-component release with the membrane-based fed-batch shake flask serves a threefold purpose of replenishing depleted substrates in the culture medium, stabilizing the pH throughout the entire cultivation time and minimizing the necessary amount of buffer to maintain an optimal osmolality range. In comparison to a batch cultivation, these settings enable to achieve higher biomass and product concentrations.

Electronic supplementary material

The online version of this article (10.1186/s12934-018-0917-8) contains supplementary material, which is available to authorized users.

Online measurement of the respiratory activity in shake flasks enables the identification of cultivation phases and patterns indicating recombinant protein production in various Escherichia coli host strains

Escherichia coli is commonly used for recombinant protein production with many available host strains. Screening experiments are often performed in batch mode using shake flasks and evaluating only the final product concentration. This conventional approach carries the risk of missing the best strain due to limited monitoring capabilities. Thus, this study focuses on investigating the general suitability of online respiration measurement for selecting expression hosts for heterologous protein production. The oxygen transfer rate (OTR) for different T7-RNA polymerase-dependent Escherichia coli expression strains was compared under inducing and noninducing conditions. As model enzymes, a lipase A from Bacillus subtilis (BSLA) and a 3-hydroxybutyryl-CoA dehydrogenase from Thermus thermophilus (HBD) were chosen. Four strains were compared during expression of both enzymes in autoinduction medium. Additionally, four strains were compared during expression of the BSLA with IPTG induction. It was found that the metabolic burden during recombinant protein production induces a phase of constant OTR, while undisturbed cell growth with no or little product formation is indicated by an exponential increase. This pattern is independent of the host strain, expressed enzyme, and induction method. Furthermore, the OTR gives information about carbon source consumption, biomass formation, and the transition from production to noninduced second growth phase, thereby ensuring a fair comparison of different strains. In conclusion, online monitoring of the respiration activity is suited to qualitatively identify, if a recombinant protein is produced by a strain or not. Furthermore, laborious offline sampling is avoided. Thus, the technique is easier and faster compared to conventional approaches. © 2018 American Institute of Chemical Engineers Biotechnol. Prog., 34:315-327, 2018.