Garrigós‑Martínez et al. Microb Cell Fact https://doi.org/10.1186/s12934-019-1240-8 (2019) 18:187 Microbial Cell Factories RESEARCH Open Access Specific growth rate governs AOX1 gene expression, affecting the production kinetics of Pichia pastoris (Komagataella phaffii) ­PAOX1‑driven recombinant producer strains with different target gene dosage Javier Garrigós‑Martínez, Miguel Angel Nieto‑Taype, Arnau Gasset‑Franch, José Luis Montesinos‑Seguí, Xavier Garcia‑Ortega and Francisco Valero* Abstract Background: The ­PAOX1-based expression system is the most widely used for producing recombinant proteins in the methylotrophic yeast Pichia pastoris (Komagataella phaffii). Despite relevant recent advances in regulation of the methanol utilization (MUT) pathway have been made, the role of specific growth rate (µ) in AOX1 regulation remains unknown, and therefore, its impact on protein production kinetics is still unclear. Results: The influence of heterologous gene dosage, and both, operational mode and strategy, on culture physiolog‑ ical state was studied by cultivating the two ­PAOX1-driven Candida rugosa lipase 1 (Crl1) producer clones. Specifically, a clone integrating a single expression cassette of CRL1 was compared with one containing three cassettes over broad dilution rate and µ ranges in both chemostat and fed-batch cultivations. Chemostat cultivations allowed to establish the impact of µ on the MUT-related MIT1 pool which leads to a bell-shaped relationship between µ and P ­ AOX1-driven gene expression, influencing directly Crl1 production kinetics. Also, chemostat and fed-batch cultivations exposed the favorable effects of increasing the CRL1 gene dosage (up to 2.4 fold in qp) on Crl1 production with no significant detrimental effects on physiological capabilities. Conclusions: PAOX1-driven gene expression and Crl1 production kinetics in P. pastoris were successfully correlated with µ. In fact, µ governs MUT-related MIT1 amount that triggers ­PAOX1-driven gene expression—heterologous genes included—, thus directly influencing the production kinetics of recombinant protein. Keywords: AOX1 promoter, Heterologous gene dosage, Transcription analysis, Pichia pastoris, Gene expression/ regulation, MIT1, Specific growth rate influence, Operational mode Background for metabolite production. Mattanovich et al. have sum- In the last two decades, Komagataella phaffii, which marized the main uses of P. pastoris for metabolite pro- was formerly known as Pichia pastoris, has emerged as duction [7]. The potential of P. pastoris for hosting the a promising host for recombinant protein production production of recombinant proteins is increased by its (RPP) [1–6]. Also, it has lately been increasingly used ability to grow at high cell densities (ca. 100 g L−1 dry cell weight) on defined media, the availability of strong protein expression systems, the possibility to secrete the *Correspondence: francisco.valero@uab.cat target proteins to the extracellular medium, its enabling Department of Chemical, Biological and Environmental Engineering, eukaryotic post-translational modifications [8, 9] and a School of Engineering, Universitat Autònoma de Barcelona, 08193 Bellaterra (Cerdanyola del Vallès), Spain reference genome sequence [10]. © The Author(s) 2019. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat​iveco​mmons​.org/licen​ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided 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 Creative Commons Public Domain Dedication waiver (http://creativecommons.org/ publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 2 of 15 The alcohol oxidase 1 promoter (­PAOX1) expression Furthermore, specific growth rate (µ) has been con- system has been widely used for recombinant protein firmed as a key parameter that affects the specific protein production on P. pastoris. In terms of regulation, P ­ AOX1 production rate (qp). To date, many attempts to correlate is strongly inducible by methanol and repressible by both parameters have been successfully made. Thus, a both glucose and glycerol. Its tight regulation allows positive relationship between them was observed when bioprocess decoupling into a first phase of biomass gen- producing different proteins under the ­PGAP [29–31] and eration and a second phase of where heterologous gene ­PAOX1 control [32]. As the P. pastoris endogenous genes expression is induced by the addition of methanol. Prop- controlled by these promoters play crucial roles in gly- erly designing the induction phase is crucial to obtain colysis and methanol metabolism, respectively, the pro- acceptable amounts of recombinant protein [2, 6, 11, 12]. tein production driven by these expression systems are ­PAOX1 typically allows large amounts of proteins to be coupled to cell growth. By contrast, other authors point obtained [3, 13–15]; however, the need to use methanol out the presence of a maximum in the qp–µ curve. Thus, leads to some drawbacks related to plant safety, high oxy- Prielhofer et al. [33], observed a bell-shaped relationship gen consumption and also high heat production [16, 17]. between qp and µ when expressing i-bodies under the In the literature, recent relevant advances in ­PAOX1 regu- control of an improved glucose-repressible P ­ GTH1 pro- lation can be found [3]. Thus, promoter sequence analysis moter. These results led them to devise an optimized bio- has allowed several binding sites for transcription fac- process strategy based on a stepwise decrease in µ during tors (TFs) to be identified. Most such TF were previously their fed-batch experiments. Canales et al. [34] studied known and have been related to stress response, glucose the effect of glycerol:methanol mixtures in the chemostat repression and oxygen consumption [18]. Three of them feeding stream and the specific growth rate on Rol pro- (Mig1, Mig2 and Nrg1) have emerged as strong repres- duction under P ­ AOX1 promoter. They found µ to be much sors of genes involved in methanol uptake [19], whereas more influential on qp than was the methanol fraction in three others (Mxr1, Mit1 and Prm1) have proved crucial the feeding. triggers of MUT genes expression [20–22]. The increas- In this work, the integrated effect of µ and gene dos- ing information gathered about MUT gene expression age on AOX1 gene regulation and production kinetics has allowed some researchers to develop methanol-free of Candida rugosa lipase 1 (Crl1) driven by P ­ AOX1 in P. expression systems based on MUT machinery [19, 23, pastoris was studied for designing a rational approach 24]. Such systems do not need methanol to trigger MUT to optimize the operating conditions. For this purpose, a genes because their TF genes have been derepressed by single-copy clone (SCC) and a multi-copy clone (MCC) genetic engineering. were both cultivated under chemostat conditions to Some researchers have focused on the relationship establish the relationship between µ, CRL1 relative tran- between heterologous gene dosage and protein pro- script levels (RTL) and qp. This correlation has allowed duction rate. As previously reported, in ­ PAOX1-driven determining the operational strategy that maximizes Crl1 expression systems, gene dosage and protein production production. Additionally, transcriptional analyses of two are usually positively correlated, albeit with a relatively key genes involved in methanol metabolism—AOX1 and small number of copies (2 or 3) only [25–28]. However, MIT1—were used in order to establish whether this path- producer clones integrating high gene of interest (GOI) way might be limited under specific conditions. Finally, expression cassettes are often subject to folding and Fed-batch cultivations were used to confirm the qp–µ secretion restrictions that result in oxidative stress in the profile pattern observed with chemostat cultivations endoplasmic reticulum, thereby having a direct impact to validate this experimental platform for the standard on protein production. Also, producer clones contain- industrial operation mode used in P. pastoris cell factory. ing large numbers of copies have been found to possess a limited transcription efficiency [25, 28]. According to Results and discussion Cámara et al. [28], the main limitation in strains with Effect of increasing CRL1 gene dosage on culture a large number of GOI copies occurs at the transcrip- physiological state tional level rather than in folding or secretion processes. Increasing the dosage of heterologous genes is known Interestingly, both ­PAOX1-driven Rhizopus oryzae lipase to affect homeostasis in P. pastoris cultivations through (ROL) gene and MUT genes (AOX1 included) have been restrictions in protein processing [35, 36]. Also, found to be downregulated in clones with a large num- ­PAOX1-driven expression systems have been found to ber of GOI copies, a limitation that results in decreased exhibit attenuated MUT gene expression [28], thereby Rol production and methanol accumulation in chemostat affecting the methanol uptake rate (qs) of producer cultivations. strains and potentially reducing their ability to grow [27, 36, 37]. Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 3 of 15 Fig. 1 Pichia pastoris physiological response to an increase in CRL1 gene dosage in chemostat cultivations. a Specific methanol consumption rate (qs), overall biomass-to-substrate yield (YX/S*). b Specific oxygen uptake rate (qO2), specific carbon dioxide production rate (qCO2) and respiratory quotient (RQ). Error bars represent the standard deviation of two biological replicates Figure 1a shows the variation of the specific substrate clones. As a result, intrinsic substrate-to-biomass yield uptake rate (qs) and overall biomass-to-substrate yield (YS/X), and their respective maintenance coefficients (ms), (YX/S*) over a wide range of dilution rates (D) (0.020– were very similar (Table 1). Interestingly, both clones 0.095 h−1) in chemostat cultivations of SCC and MCC. had mean YS/X values around 2.2 ­gMetOH g−1 X . This value No methanol accumulation was observed under any is similar to the yield for the wild-type strain [38] and a conditions, but no D values above 0.095 h−1 were used slightly lower than reported for an important number of in order to avoid washout. In addition, the carbon and recombinant protein producer strains, which YS/X ranges electron balances were checked and deviations prior to 2–3 gMetOH g−1 X . However, for the recombinant produc- data reconciliation found to be less than 5%. With both tion of other target proteins YS/X can reach higher values clones, qs increased linearly across the D range, and qs [6]. For instance, YS/X reached in the production of Rol values at equivalent D values were rather similar for both under the same expression system was twofold higher Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 4 of 15 Table 1 Intrinsic yields (Yi/X) and maintenance coefficients MCC in fed-batch cultivation was observed at the high- (mi/X) from biomass growth obtained from chemostat est µ tested (0.08 h−1). Thus, maximum µ tested for MCC cultivations was decreased to 0.065 h−1 in order to maintain carbon- Single-copy clone Multi-copy clone limiting conditions. Although significant differences in qCO2 and qO2 were observed, RQ was quite similar in both YS/X ­(gMeOH g−1 X ) 2.16 ± 0.08 2.21 ± 0.05 operational modes irrespective of µ. ms ­(gMeOH g−1 −1 X ­h ) 0.014 ± 0.005 0.007 ± 0.004 YCO2 /X (molCO2 C-mol−1 X ) 0.95 ± 0.07 1.00 ± 0.05 Relationship between Mit1 limitation and a decreased mCO2 (molCO2 C-mol−1 −1 X h ) 0.012 ± 0.005 0.007 ± 0.003 AOX1 relative expression (RE) YO2 /X (molO2 C-mol−1 X ) 1.92 ± 0.11 1.99 ± 0.07 The induction of strains with multiple copies of a mO2 (molO2 C-mol−1 X −1 ­h ) 0.018 ± 0.007 0.010 ± 0.005 ­ AOX1-driven heterologous gene with methanol has been P ± Indicates standard error (SE) from regression analysis reported to result in transcriptional limitation of MUT genes [28]. Therefore, transcriptional analysis of AOX1, CRL1 and the methanol-induced transcription factor 1 than those obtained in the present work [13]. About the (MIT1) genes were performed in order to examine their value of ms reached for both clones, no significant sta- impact on the first step of methanol metabolism (see tistical differences were observed. Results fell within the Fig. 3a, b). range of values reported in the literature, from 0.007 to These analyses were only carried out in chemostat cul- 0.042 ­gMetOH g−1 −1 X ­h . tivations, which were performed in steady-state condi- Other factors affecting the physiological state of the cell tions. As can be seen from Fig. 3a, and consistent with factory are related to respiration parameters, such as the previous results [37], AOX1 gene expression was in aver- specific ­O2 uptake rate (qO2), the specific C ­ O2 produc- age twofold higher in SCC than it was in MCC, which- tion rate (qCO2), the corresponding intrinsic yields (Yi/X) ever the dilution rate. It indicates that the resources and maintenance coefficients (mi/X), and the respiratory needed to trigger transcription of ­ PAOX1-driven genes quotient (RQ). As can be seen from Fig. 1b and Table 1, may be shared among them—heterologous gene cassettes these factors exhibited identical trends and similar values and the endogenous AOX1 gene included. across the D range regardless of the gene dosage. Similar This phenomenon was further studied by analyzing tran- YO2 /X were obtained for the human serum albumin [39], scriptional levels of MIT1, a crucial TF for P ­ AOX1 induction antibody production [40] and human chymotrypsinogen [21], in both clones. As can be seen from Fig. 3b, MIT1 rel- B [41] between 0.08 and 0.09 molO2 g−1 X . However highest ative transcript levels (RTL) were not significantly different values around 0.15 molO2 g−1 X were obtained for Rhizopus comparing both clones across the D range tested, except oryzae lipase [13]. The higher values of YO2 /X are directly for two dilution rates (0.02 h−1 and 0.08 h−1). Therefore, related with the lower YX/S* obtained. The results suggest as expected, increasing heterologous gene dosage has not that the values of the different yields are a function of the led to a proportional increase in MIT1 transcription rate. target protein expressed. One should therefore hypothesize that the Mit1 pool is a The fact that the major macrokinetic parameters limited resource, all the genes whose expression depends related to the physiological status were very similar for on ­PAOX1 promoter would compete with one another for both clones suggests that increasing the CRL1 cassette this TF—and hence for being transcribed. Consequently, dosage from one to three in P. pastoris genome had no the AOX1 gene was less strongly expressed in MCC than it effect on the physiological performance of the yeast in was in SCC owing to competition with three CRL1 copies chemostat cultivations. On the other hand, a marked for the equivalent Mit1 resources. This hypothesis of Mit1 influence of the operational mode on such parameters limitation is reinforced by the work of Cámara et al. [42] was observed in FB cultivations of both clones. As can be where the overexpression of Mit1 is enough to reverse the seen in Fig. 2a, b, the evolution of qs, qCO2 and qO2 across transcriptional limitation derived from increasing heterol- the range of µ tested were different for both clones. These ogous gene dosage. Moreover, deregulating the expression rates followed a nearly linear trend for SCC, similar to of some MUT-related TFs increased protein production the behaviour observed on chemostat cultivations. In driven by the ­PAOX1 expression system even in absence of contrast, a saturated curve trend was obtained for MCC, methanol [19, 24, 42, 43]. thus indicating that the higher MCC metabolic burden Although these insights were previously obtained from caused by an increase heterologous gene dosage affects a heterologous gene dosage comparison, no similar stud- the clones capabilities for both methanol and ­O2 con- ies had examined a potential correlation of MUT-related sumption, as well as C ­ O2 production in this operational genes RTL with the µ. A positive proportional relationship mode. Consequently, accumulation of methanol on between µ, MIT1 RTL and P ­ AOX1-driven transcription rate Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 5 of 15 Fig. 2 Pichia pastoris physiological response to an increase in CRL1 gene dosage in fed-batch (FB) cultivations. a Specific methanol consumption rate (qs), overall biomass-to-substrate yield (YX/S*). b Specific oxygen uptake rate (qO2), specific carbon dioxide production rate (qCO2) and respiratory quotient (RQ) should be expected since the more methanol was fed to Overall, the previous results show that the MIT1 the culture, the greater was the amount of AOX1 enzyme RTL is governed by µ, affecting the transcription rate of needed to consume it. However, as can be seen in Fig. 3a, ­PAOX1-driven genes. One could hypothesize that a simi- b, the correlation of the MUT-related genes RTL with D lar phenomenon could also take place for to other MUT- was bell-shaped for both clones, which suggests a close related TFs such as Mxr1 and Prm1, since the regulation of relationship between ­PAOX1-driven genes expression and their expression in presence of methanol must be coupled MIT1. Further research at transcriptional level would be [21]. Although AOX1 RTLs were low in MCC relative to needed to elucidate why P. pastoris decreases its methanol SCC, AOX1 was expressed strongly enough to produce the consumption resources when it approaches its maximum minimum amount needed to catabolize all methanol fed to specific growth rate, µmax. chemostat cultivations. However, as noted in the previous Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 6 of 15 Fig. 3 Influence of the dilution rate on relative gene transcription levels (RTLs) in chemostat cultivations. a Genes AOX1 and CRL1. b Gene MIT1. MTH1 was used as housekeeping gene for the analysis. Error bars represent the standard deviation of two biological replicates section, growing FB cultivations of MCC at the maximum Influence of operational mode on production‑related µ level reached by SCC (0.08 h−1) led to methanol accumu- parameters lation during early stages of the feeding phase. The primary aim of this work was to elucidate the Crl1 production kinetics for both clones in chemostat and FB cultivations. Furthermore, transcriptional analysis Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 7 of 15 provided valuable information, which also should be The improvement of product-related parameters related with both the growth and the recombinant pro- could be considered slightly lower than expected, since tein production. Since Crl1 production in this cell fac- MCC harbor three CRL1 expression cassette copies. tory is governed by the ­PAOX1 promoter, it was expected However, as widely reported, increasing the heterolo- to be coupled to growth because the sole carbon source gous GOI dosage in the genome need not lead to a pro- used was methanol [32]. However, qp was not linearly portional increase in protein production rates [25–28]. related to D or µ in either operational mode. Rather, In this case, the low CRL1 transcription efficiency— both chemostat and FB cultivations exhibited a bell- CRL1 RTL was just 2.2-fold higher on MCC—is the shaped trend in both clones, production being optimal responsible of the lower than expected product-related at D = 0.08 h−1 in chemostat cultivations (Fig. 4a) and parameters values. µ = 0.045 h−1 in FB cultivations (Fig. 4b). In FB cultiva- The increase in protein production derived from tions, MCC gave a more pronounced bell-shape curve increasing the CRL1 gene dosage in FB cultivations was than did SCC. Therefore, MCC would require a more similar to that in chemostat cultivations. In fed-batch precise control of µ because a slight deviation from the cultures, however, the ratios between clones increased optimal set-point would result in a marked decrease in with increasing µ (Table 3). Thus, at low µ levels, prod- qp. Consequently, the optimum differences in µ should uct-related parameters such as titer, qp, Qp (volumet- be considered in designing bioprocesses for recombinant ric productivity) and product yields were about 1.8 fold protein production. higher in MCC than they were in SCC. At intermediate The overall product-to-biomass and product-to-sub- µ levels, the previous parameters were roughly 2.3 fold strate yield (YP/X and YP/S, respectively) exhibited a lin- higher in MCC. At the highest µ level, however, the com- ear decreasing trend in chemostat cultivations of both parison was not reliable since the culture conditions were clones (Fig. 5a). Although SCC behaved identically in FB not equivalent. Specifically, the µ set-point used with cultivations, the MCC exhibited a maximum value at an MCC had to be adapted to avoid an eventual methanol intermediate µ level (0.045 h−1; Fig. 5b). These yields are accumulation. important inasmuch as they are closely related to product Overall, increasing the CRL1 gene dosage resulted in titer (Fig. 6), which is a parameter susceptible to be opti- increased protein production in both chemostat and FB mized in industry due to its influences on downstream cultivations. processing costs. Therefore, MCC would be the strain of choice for optimum Crl1-related yields and titers when Transcriptional efficiency differences between AOX1 cultivated at intermediate µ values. Irrespective of gene and CRL1 genes, and their impact on Crl1 production dosage, qp and product-related yields were more than Regarding the balance between the transcription levels of twice greater in FB cultivations than they were in chemo- the AOX1 and the CRL1 genes, being both P ­ AOX1-driven, stat cultivations. an unexpected ratio between AOX1 RTL and CRL1 RTL was found in chemostat cultivations. As can be seen from Increasing CRL1 gene dosage boosts protein production Fig. 8, AOX1 RTL was considerably greater than CRL1 Although a transcriptional limitation had been proved RTL in SCC, the difference ranging from 11 times at low in at least two MUT-related genes, it was necessary to D values to 6 times at the highest one. Even in MCC, quantify to what extent the GOI transcription rate was which harbor three CRL1 expression cassettes—versus affected by an increase in gene dosage, and hence how it only one of AOX1—, AOX1 RTL exceeded clearly the influenced qp. CRL1 RTL. The amount of mRNA a given gene contains In chemostat cultivations, CRL1 RTL was on average is known to depend on the balance between transcrip- 2.2 fold higher in MCC than in SCC across the D range tion rate and mRNA decay [44]. Therefore, since both (Fig. 7), which was also reflected on qp raises between coding sequences were flanked by the same promoter 2.1 and 2.4 (Table 2). This correlation between CRL1 ­(PAOX1) and transcription terminator (AOX1) here, one RTL and qp ratios suggests the absence of a bottleneck should expect the transcription rate to be similar. Hence, in further protein processing-secretion steps, and hence the differences in mRNA between CRL1 and AOX1 might in overall Crl1 production rate. This hypothesis was sup- be related with mRNA stability and hence with mRNA ported by conducting a transcriptional analysis of UPR- degradation. related genes such as KAR2 and HAC1, the expression As shown in Fig. 7, CRL1 RTL was closely cor- levels of which were rather constant across the D range related with qp as a result of the absence of stacks in (results not shown). Likewise, raises of other product- folding, trafficking and secretion processes. The AOX1 related parameters were also around 2 and 2.8-fold higher RTL/CRL1 RTL ratio was thus a crucial parameter. As when comparing equivalent D conditions (Table 2). noted earlier, the pool of MUT-related TFs that can Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 8 of 15 Fig. 4 Comparison of SCC and MCC Crl1 production kinetics and its relationship with CRL1 relative transcription levels. The specific Crl1 production rate (qp) was calculated for chemostat (a) and fed-batch cultivations (b). CRL1 transcriptional analyses were also done on chemostat cultivations (a). MTH1 was used as housekeeping gene for CRL1 RTL calculations. Error bars represent the standard deviation of two biological replicates be shared by all the ­PAOX1-driven genes is expected to expression system, increasing the CRL1 gene dosage to be limited. As a result, increasing the number of CRL1 three expression cassettes (MCC) reduced the AOX1- cassettes in the genome should gradually increase the to-CRL1 RTL ratio from 10 to 7 in SCC to 2 in MCC; as ratio up to a point where AOX1 expression would not a result, qp was increased by a factor of 1.7–3 without be enough to consume all the methanol fed in the cul- appreciably affecting the ability to metabolize methanol ture. This hypothesis was confirmed in those cases fed to the culture. Therefore, the SCC expression sys- where methanol accumulation was substantial. In this tem could be considered inefficient for producing Crl1 Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 9 of 15 Fig. 5 Comparison of SCC and MCC strain Crl1-related yields. Overall product-to-biomass yield (YP/X) and product-to-substrate yield (YP/S). a Chemostat cultivations. b Fed-batch cultivations. In a, error bars represent the standard deviation of two biological replicates because the cell factory P. pastoris expresses higher lev- yeast. However, potential bottlenecks arising from an els of AOX1 than the essentially needed, which is detri- increased protein production should be considered. mental to Crl1 production. Further increasing the number of CRL1 cassettes is Conclusions therefore the way of identifying the optimum AOX1 In this work, the influence of the heterologous gene dos- RTL/CRL1 RTL ratio for maximal Crl1 production with- age was used to expose the high importance of µ on the out detracting from the physiological capabilities of the transcription of MUT genes, production kinetics and culture physiological status. Increasing the CRL1 gene Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 10 of 15 for MCC. Specifically, a saturation pattern was observed in all physiology-related macrokinetic parameters across the µ range studied. This result departs from the expected linear trends, which were indeed observed in chemostat cultivations. Also, the influence of the operational mode on the physiological status was significantly higher in MCC than in SCC. According to the results presented, µ seemingly deter- mines the expression of MIT1, which have a key role in triggering transcription of ­PAOX1-driven genes, thus influ- encing the amount of protein of interest that is produced at the end of the process. Also, increasing the number of CRL1 expression cassettes integrated in the genome from one to three in P. pastoris effectively boosted production Fig. 6 Crl1 production time evolution expressed as total activity units without significantly altering the physiological status of in fed-batch cultivations at different µ: (filled circle, open circle), 0.028 the yeast. Furthermore, since increasing the CRL1 dosage ­h−1; (filled square, open square), 0.047 ­h−1; (filled triangle), 0.063 ­h−1; (open triangle), 0.084 ­h−1 strongly reduced AOX1 expression, one could hypoth- esize an eventual limitation of the TF MIT1 pool, which is supported by our results. Finally, the strong correlation Fig. 7 Effect of dilution rate on the CRL1 relative transcription level and specific production rate ratios between MCC and SCC. Error bars represent the standard deviation of qp and RTL ratios dosage was found not to affect the clone capabilities in between CRL1 RTL and specific CRL1 production rate in terms of methanol and oxygen uptake rate (qs and qO2, both clones suggests the absence of bottlenecks in pro- respectively), nor the carbon dioxide production rate tein processing processes for this particular expression (qCO2), at least at the macrokinetic level in chemostat system in chemostat cultivations. cultivations. On the other hand, a significant effect on As shown in the present work, the operational these parameters was observed in fed-batch cultivations mode used considerably influences product-related Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 11 of 15 Table 2 Comparison of Crl1 production-related parameters for chemostat cultivations Single copy clone Multi copy clone D ­(h−1) 0.019 0.035 0.050 0.064 0.079 0.095 0.020 0.036 0.050 0.066 0.079 0.097 Product titer (AU ­mL−1) 124 101 118 107 93 68 321 267 274 243 212 155 Product titer ratio – – – – – – 2.59 2.64 2.32 2.27 2.28 2.28 qP (AU g−1 −1 X ­h ) 132 183 306 346 352 312 297 449 646 735 751 656 qP ratio – – – – – – 2.25 2.45 2.11 2.12 2.13 2.10 QP (kAU L­ −1 ­h−1) 2.43 3.51 5.91 7.59 7.31 6.51 6.46 9.81 13.72 16.02 16.71 15.15 QP ratio – – – – – – 2.66 2.79 2.32 2.11 2.29 2.33 YP/S (kAU g−1 S ) 2.50 1.96 2.45 2.17 1.92 1.41 6.31 5.52 5.41 4.79 4.05 3.01 YP/S ratio – – – – – – 2.52 2.8 2.21 2.21 2.11 2.13 YP/X (kAU g−1 X ) 6.76 5.25 6.09 4.88 4.48 3.27 15.77 13.08 12.92 11.18 9.55 6.74 YP/X ratio – – – – – – 2.33 2.49 2.12 2.29 2.13 2.06 The ratios between the MCC and SCC were calculated by dividing the MCC parameter values to SCC ones at similar dilution rate (D) Table 3 Comparison of Crl1 production-related parameters for fed-batch cultivations Single-copy clone Multi-copy clone µ ­(h−1) 0.028 0.047 0.084 0.027 0.046 0.063 Product titer (AU ­mL−1) 769 660 261 1386 1542 1033 Product titer ratio – – – 1.80 2.32 3.95 qP (AU g−1 −1 X ­h ) 364 460 326 628 1322 958 qP ratio – – – 1.73 2.25 2.93 QP (kAU L­ −1 ­h−1) 7.87 10.89 5.53 14.66 21.81 18.01 QP ratio – – – 1.86 2.00 3.26 YP/S (kAU g−1 S ) 5.37 4.67 1.86 9.74 12.19 7.25 YP/S ratio – – – 1.81 2.74 3.90 YP/X (kAU g−1 X ) 13.16 9.76 3.91 23.84 28.88 15.32 YP/X ratio – – – 1.81 2.56 3.91 The ratios between the MCC and SCC were calculated by dividing the MCC parameter values to SCC ones at similar specific growth rate (µ) parameters such as YP/X and qp, which were twice system as well as the target protein. Therefore, similar greater in the fed-batch cultivations than they were in experiments should be conducted in each case, not only the chemostat cultivations at identical specific growth to maximize production rates, but also to identify the rates for Crl1 production in this expression system. most suitable conditions for testing other strains with Interestingly, by means of analysing both ­PAOX1-driven industrial potential. expression, we prove that the transcription efficiency is an important factor to study in any case. In our case, Materials and methods despite being flanked by same promoter and termi- Plasmid construction and strain generation nator, the expression of AOX1 and CRL1 genes were Recombinant strains of P. pastoris expressing CRL1 gene highly different. Therefore, when using endogenous under the regulation of ­PAOX1 were constructed by using expression systems to produce a protein—like ­PAOX1, the pPICZαA plasmid (Invitrogen, Carlsbad, CA, US) the most important factor to analyse is not only the assembled with the codon-optimized synthetic open gene dosage but the expression difference between the reading frame (ORF) encoding the CRL1 gene sequence endogenous and the heterologous gene. (GeneScript, Piscataway, NJ, USA). Then P. pastoris X-33 The outcome of these experiments expects to provide cells (Invitrogen, Carlsbad, CA, US) were transformed a wealth of knowledge for designing a rational approach with chimeric vector under the conditions described to optimizing the operating conditions. Although the elsewhere [42]. Therefore, two clones having one and production patterns are expected to be similar for dif- three CRL1 expression cassette copies, respectively, were ferent proteins of interest to be expressed, the outcome selected and used for this study. usually depends on the particular expression regulation Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 12 of 15 Fig. 8 Effect of dilution rate on the AOX1–CRL1 relative transcription level ratio between MCC and SCC Error bars represent the standard deviation of RTL ratios Gene dosage determination Synthesis of cDNA and determination of transcriptional cDNA was synthetized with the iScript™ cDNA Synthesis The number of CRL1 expression cassettes integrated levels into the genome was determined by droplet digital PCR (ddPCR) according to Cámara et al. [45]. The actin gene kit (Bio-Rad, Hercules, CA, USA), following the manu- (ACT1) was selected as a housekeeping gene for the anal- facturer’s instructions. For qPCR, a set of primers were ysis. The specific primers used are presented in Addi- designed for specific target cDNA. The set of selected tional file 1. genes comprised CRL1 (heterologous gene); AOX1, the alcohol oxidase 1 native gene; and MIT1, which codi- fies a key transcription factor of the methanol-induced Total RNA extraction transcription. Furthermore, KAR2 and HAC1, two genes Chemostat samples for RNA isolation were collected involved in the unfolded protein response (UPR), were ture broth samples were resuspended in 1 mL of TRIzol™ according to Landes et al. [46]. Pellets from 1 mL cul- For qPCR, reactions were done with SYBR™ Select also analyzed. reagent (Waltham, Massachusetts, USA) and lysed with Master Mix (Thermo Scientific™ Waltham, MA, US). glass beads (425–600 µm, Sigma-Aldrich, St. Louis, MO, Additionally, and as suggested by the manufacturer, to USA) for mechanical disruption. Cell lysis was attempted ­ pMotion® 5075 robot (Eppendorf, Germany). assure the maximum accuracy the reaction mixes were by alternating cycles of 30-s of vortexing and freezing. All made by E further steps were performed according to the manufac- 12 K Flex Real-Timer from Thermo Scientific™ (Waltham, The amplification program was run on a QuantStudio turer’s instructions. RNA integrity was checked by agarose electrophoresis MA, US), following the manufacturer’s instructions. The and the RNA concentration determined from Nanodrop annealing extension temperature was set at 57.4 °C. Rela- tific™ (Waltham, MA, US). measurements on an instrument from Thermo Scien- tive transcript levels (RTLs) were determined by using MTH1 as a housekeeping gene as it shows basal expres- sion across the conditions tested. Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 13 of 15 Chemostat cultivation Lipolytic activity Chemostat cultivation of the two clones were run in An enzymatic p-nitrophenyl butyrate (pNPB) based duplicate in a 2 L Biostat B plus Bioreactor (Sartorius Ste- assay was selected to determine Crl1 activity by using dim, Goettingen, Germany) according to García Ortega a procedure described elsewhere [49] albeit with slight et al. [30]. The medium composition was the same except modifications. Thus, 20 μL volumes of the samples were that glucose was replaced with methanol as sole carbon mixed with 980 μL of reaction buffer, which contained source at a final concentration of 50 g ­L−1 on feeding. 1 mM pNPB, 50 mM phosphate buffer (pH 7) and 4% A wide range of dilution rates was covered. Specifically, (v/v) acetone. The absorbance at 348 nm was monitored it was tested the following dilution rates: 0.020 h−1, at 30 °C by using a Specord 200 Plus spectrophotometer 0.035 h−1, 0.050 h−1, 0.065 h−1, 0.080 h−1, 0.095 h−1. from Analytic Jena (Jena, Germany). One activity unit Under each set of conditions, samples were obtained was defined as the amount of enzyme needed to release after five residence times. To ensure that a steady state 1 mmol of p-nitrophenol per minute under assay condi- was reached, samples were analyzed from the third resi- tions. RSD was less than 1%. dence time, to check the stability of the parameters of interest. Process parameters determination, consistency checking and data reconciliation Fed‑batch cultivation Mass balance and stoichiometric equations Both clones were also cultivated in the fed-batch mode, All equations derived from the mass balances used to using a 5 L Biostat B Bioreactor (Sartorius Stedim, Goe- calculate yields and rates in the chemostat [30] and fed- ttingen, Germany) at different specific growth rates batch experiments [13] can be found elsewhere. The from 0.030 to 0.08 h−1 for SCC and 0.030 to 0.065 h−1 mean elemental biomass composition C ­ H1.78 ­O0.62 ­N0.18 for MCC. An exponential pre-programming feeding ­S0.006 with an ash content of 9% was determined as pre- rate was performed to maintain the specific growth rate viously reported [47]. The carbon and electron balances constant at the selected set-point. All cultivations were were checked and less than 5% of deviation observed grown under carbon-limiting conditions. The procedure prior to reconciliation. is described in detail elsewhere [29]—by exception, the glucose/glycerol pair was replaced with a methanol con- ­ −1 in the feed. centration of 400 g L Consistency checking and data reconciliation Measurement consistency was checked by using the Biomass determination as dry cell weight (DCW) standard test with carbon and electron balances as con- Biomass concentrations were measured as DCW values straints. Both online and offline measurements allowed as described elsewhere [47]. The relative standard devia- five key specific rates in the black-box process model to tion (RSD) of the measurements was about 3%. be calculated, namely: biomass generation (μ), glucose uptake rate (qs), product generation rate (qp), oxygen uptake rate ( qO2 ) and carbon dioxide production rate Quantification of the carbon source and byproducts ( qCO2 ). The method used for this purpose is described The concentration of the different carbon sources used in detail elsewhere [13]. in the batch (glycerol), and chemostat and fed-batch cultivations (methanol), and the potential fermentation Supplementary information byproducts, were all determined by HPLC. The column Supplementary information accompanies this paper at https​://doi. and program used are described elsewhere [48]. RSD was org/10.1186/s1293​4-019-1240-8. invariably less than 1%. Additional file 1. List of primers used for qPCR analyses. Off‑gas analyses A BlueInOne Cell gas analyzer (BlueSens, Herten, Ger- Acknowledgements many) was used with both chemostat and fed-batch cul- Not applicable. tivations. The ­CO2 and ­O2 mol fractions were recorded online with provision for off-gas pressure and humid- Authors’ contributions JG-M and XG-O constructed the recombinant strain, and JG-M and AG-F ity. The data thus obtained were used to calculate the performed the chemostat and fed-batch cultivations, respectively. JG-M and oxygen uptake rate (OUR), carbon dioxide evolution MAN-T analyzed the expression of the selected genes. XG-O, JLM-S and FV rate (CER), specific rates ( qO2 and qCO2 ) and respiratory conceived and supervised the study. All authors prepared the final manu‑ script. All authors read and approved the final manuscript. quotient (RQ). RSD was less than 5% in all cases. Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 14 of 15 Funding 15. Barrigón JM, Valero F, Montesinos JL. A macrokinetic model-based This work was funded by MINECO and FEDER under Project CTQ2016-74959-R. comparative meta-analysis of recombinant protein production by Pichia The authors’ group is member of 2017-SGR-1462 and the Reference Network pastoris under AOX1 promoter. Biotechnol Bioeng. 2015;112:1132–45. in Biotechnology (XRB) of the Generalitat de Catalunya. M.A. Nieto-Taype 16. Heyland J, Fu J, Blank LM, Schmid A. Quantitative physiology of Pichia acknowledges funding by the National Council of Science, Technology and pastoris during glucose-limited high-cell density fed-batch cultivation for Technological Innovation (CONCYTEC, FONDECYT), and J. Garrigós-Martínez a recombinant protein production. Biotechnol Bioeng. 2010;107:357–68. PIF scholarship from the Universitat Autònoma de Barcelona. 17. Çalık P, Ata Ö, Güneş H, Massahi A, Boy E, Keskin A, Örtürk S, Zerze GH, Özdamar TH. Recombinant protein production in Pichia pastoris under Availability of data and materials glyceraldehyde-3-phosphate dehydrogenase promoter: from carbon All data generated or analysed during this study are included in this published source metabolism to bioreactor operation parameters. Biochem Eng J. article and its additional files. 2015;95:20–36. 18. Hartner FS, Ruth C, Langenegger D, Johnson SN, Hyka P, Lin-Cereghino Ethics approval and consent to participate GP, Lin-Cereghino J, Kovar K, Cregg JM, Glieder A. Promoter library Not applicable. designed for fine-tuned gene expression in Pichia pastoris. Nucleic Acids Res. 2008;36:76. Consent for publication 19. Wang J, Wang X, Shi L, Qi F, Zhang P, Zhang Y, Zhou X, Song Z, Caic M. Not applicable. Methanol-independent protein expression by AOX1 promoter with trans-acting elements engineering and glucose-glycerol-shift induction Competing interests in Pichia pastoris. Sci Rep. 2017;7:41850. The authors declare that they have no competing interests. 20. Lin-Cereghino GP, Godfrey L, De La Cruz BJ, Johnson S, Khuongsathiene S, Tolstorukov I, Yan M, Lin-Cereghino J, Veenhuis M, Subramani S, Cregg JM. Received: 23 July 2019 Accepted: 23 October 2019 Mxr1p, a key regulator of the methanol utilization pathway and peroxiso‑ mal genes in Pichia pastoris. Mol Cell Biol. 2006;26:883–97. 21. Wang X, Wang Q, Wang J, Bai P, Shi L, Shen W, Zhou M, Zhou X, Zhang Y, Cai M. Mit1 transcription factor mediates methanol signaling and regulates the alcohol oxidase 1 (AOX1) promoter in Pichia pastoris. J Biol References Chem. 2016;291:6245–61. 1. Cereghino GPL, Cereghino JL, Ilgen C, Cregg JM. Production of recombi‑ 22. Zhan C, Yang Y, Zhang Z, Li X, Liu X, Bai Z. Transcription factor Mxr1 nant proteins in fermenter cultures of the yeast Pichia pastoris. Curr Opin promotes the expression of Aox1 by repressing glycerol transporter 1 in Biotechnol. 2002;13:329–32. Pichia pastoris. FEMS Yeast Res. 2017;17:4. 2. Ahmad M, Hirz M, Pichler H, Schwab H. Protein expression in Pichia 23. Shen W, Xue Y, Liu Y, Kong C, Wang X, Huang M, Cai M, Zhou X, Zhang Y, pastoris: recent achievements and perspectives for heterologous protein Zhou M. A novel methanol-free Pichia pastoris system for recombinant production. Appl Microbiol Biotechnol. 2014;98:5301–17. protein expression. Microb Cell Fact. 2016;15:178. 3. Yang Z, Zhang Z. Engineering strategies for enhanced production of 24. Vogl T, Sturmberger L, Fauland PC, Hyden P, Fischer JE, Schmid C, Thal‑ protein and bio-products in Pichia pastoris: a review. Biotechnol Adv. linger GG, Geier M, Glieder A. Methanol independent induction in Pichia 2018;36:182–95. pastoris by simple derepressed overexpression of single transcription 4. Juturu V, Wu JC. Heterologous protein expression in Pichia pastoris: latest factors. Biotechnol Bioeng. 2018;115:1037–50. research progress and applications. ChemBioChem. 2018;19:7–21. 25. Sha C, Yu XW, Lin NX, Zhang M, Xu Y. Enhancement of lipase r27RCL pro‑ 5. Vogl T, Hartner FS, Glieder A. New opportunities by synthetic biology for duction in Pichia pastoris by regulating gene dosage and co-expression biopharmaceutical production in Pichia pastoris. Curr Opin Biotechnol. with chaperone protein disulfide isomerase. Enzyme Microb Technol. 2013;24:1094–101. 2013;53:438–43. 6. Looser V, Bruhlmann B, Bumbak F, Stenger C, Costa M, Camattari A, Foti‑ 26. Shu M, Shen W, Yang S, Wang X, Wang F, Wang Y, Ma L. High-level expres‑ adis D, Kovar K. Cultivation strategies to enhance productivity of Pichia sion and characterization of a novel serine protease in Pichia pastoris by pastoris: a review. Biotechnol Adv. 2014;33:1177–93. multi-copy integration. Enzyme Microb Technol. 2016;92:56–66. 7. Peña DA, Gasser B, Zanghellini J, Steiger MG, Mattanovich D. Metabolic 27. Yu X-W, Sun W-H, Wang Y-Z, Xu Y. Identification of novel factors enhanc‑ engineering of Pichia pastoris. Metab Eng. 2018;50:2–15. ing recombinant protein production in multi-copy Komagataella phaffii 8. Cereghino JL, Cregg JM. Heterologous protein expression in the methylo‑ based on transcriptomic analysis of overexpression effects. Sci Rep. trophic yeast Pichia pastoris. FEMS Microbiol Rev. 2000;24:45–66. 2017;7:16249. 9. Daly R, Hearn MTW, Hearn MTW. Expression of heterologous proteins 28. Cámara E, Landes N, Albiol J, Gasser B, Mattanovich D, Ferrer P. Increased in Pichia pastoris: a useful experimental tool in protein engineering and dosage of AOX1 promoter-regulated expression cassettes leads to tran‑ production. J Mol Recognit. 2005;18:119–38. scription attenuation of the methanol metabolism in Pichia pastoris. Sci 10. Sturmberger L, Chappell T, Geier M, Krainer F, Day KJ, Vide U, Trstenjak Rep. 2017;7:44302. S, Schiefer A, Richardson T, Soriaga L, Darnhofer B, Birner-Gruenberger 29. Garcia-Ortega X, Ferrer P, Montesinos JL, Valero F. Fed-batch operational R, Glick BS, Tolstorukov I, Cregg J, Madden K, Glieder A. Refined Pichia strategies for recombinant Fab production with Pichia pastoris using the pastoris reference genome sequence. J Biotechnol. 2016;235:121–31. constitutive GAP promoter. Biochem Eng J. 2013;79:172–81. 11. Vogl T, Glieder A. Regulation of Pichia pastoris promoters and its conse‑ 30. Garcia-Ortega X, Adelantado N, Ferrer P, Montesinos JL, Valero F. A step quences for protein production. N Biotechnol. 2013;30:385–404. forward to improve recombinant protein production in Pichia pastoris: 12. Barrigón JM, Montesinos JL, Valero F. Searching the best operational strat‑ from specific growth rate effect on protein secretion to carbon-starving egies for Rhizopus oryzae lipase production in Pichia pastoris Mut + phe‑ conditions as advanced strategy. Process Biochem. 2016;51:681–91. notype: methanol limited or methanol non-limited fed-batch cultures? 31. Rebnegger C, Graf AB, Valli M, Steiger MG, Gasser B, Maurer M, Mat‑ Biochem Eng J. 2013;75:47–54. tanovich D. In Pichia pastoris, growth rate regulates protein synthesis and 13. Ponte X, Montesinos-Seguí JL, Valero F. Bioprocess efficiency in Rhizopus secretion, mating and stress response. Biotechnol J. 2014;9:511–25. oryzae lipase production by Pichia pastoris under the control of P ­ AOX1 is 32. Jungo C, Rérat C, Marison IW, Von Stockar U. Quantitative characterization oxygen tension dependent. Process Biochem. 2016;51:1954–63. of the regulation of the synthesis of alcohol oxidase and of the expres‑ 14. Ponte X, Barrigón JM, Maurer M, Mattanovich D, Valero F, Montesinos- sion of recombinant avidin in a Pichia pastoris Mut + strain. Enzyme Seguí JL. Towards optimal substrate feeding for heterologous protein Microb Technol. 2006;39:936–44. production in Pichia pastoris (Komagataella spp) fed-batch processes 33. Prielhofer R, Reichinger M, Wagner N, Claes K, Kiziak C, Gasser B, under ­PAOX1 control: a modeling aided approach. J Chem Technol Bio‑ Mattanovich D. Superior protein titers in half the fermentation time: technol. 2018;93:3208–18. promoter and process engineering for the glucose-regulated GTH1 promoter of Pichia pastoris. Biotechnol Bioeng. 2018;115:2479–88. Garrigós‑Martínez et al. Microb Cell Fact (2019) 18:187 Page 15 of 15 34. Canales C, Altamirano C, Berrios J. Effect of dilution rate and metha‑ engineered methanol-independent Pichia pastoris strain. Process Bio‑ nol-glycerol mixed feeding on heterologous Rhizopus oryzae lipase chem. 2018;76:46–54. production with Pichia pastoris ­Mut+ phenotype in continuous culture. 44. Cheng J, Maier KC, Avsec Ž, Rus P, Gagneur J. Cis-regulatory elements Biotechnol Prog. 2015;31:707–14. explain most of the mRNA stability variation across genes in yeast. RNA. 35. Puxbaum V, Mattanovich D, Gasser B. Quo vadis? The challenges of 2017;23:1648–59. recombinant protein folding and secretion in Pichia pastoris. Appl Micro‑ 45. Cámara E, Albiol J, Ferrer P. Droplet digital PCR-aided screening and biol Biotechnol. 2015;99:2925–38. characterization of Pichia pastoris multiple gene copy strains. Biotechnol 36. Ben Azoun S, Ben Zakour M, Sghaier S, Kallel H. Expression of rabies virus Bioeng. 2016;113:1542–51. glycoprotein in the methylotrophic yeast Pichia pastoris. Biotechnol Appl 46. Landes N, Gasser B, Vorauer-Uhl K, Lhota G, Mattanovich D, Maurer M. Biochem. 2017;64:50–61. The vitamin-sensitive promoter ­PTHI11 enables pre-defined autonomous 37. Zhu T, Guo M, Zhuang Y, Chu J, Zhang S. Understanding the effect of for‑ induction of recombinant protein production in Pichia pastoris. Biotech‑ eign gene dosage on the physiology of Pichia pastoris by transcriptional nol Bioeng. 2016;113:2633–43. analysis of key genes. Appl Microbiol Biotechnol. 2011;89:1127–35. 47. Cos O, Serrano A, Montesinos JL, Ferrer P, Cregg JM, Valero F. Combined 38. Tomás-Gamisans M, Ødum ASR, Workman M, Ferrer P, Albiol J. Glycerol effect of the methanol utilization (Mut) phenotype and gene dosage on metabolism of Pichia pastoris (Komagataella spp.) characterised by recombinant protein production in Pichia pastoris fed-batch cultures. J 13C-based metabolic flux analyses. N Biotechnol. 2019;50:52–9. Biotechnol. 2005;116:321–35. 39. Liang J, Yuan J. Oxygen transfer model in recombinant Pichia pastoris and 48. Garcia-Ortega X, Valero F, Montesinos-Seguí JL. Physiological state its application in biomass estimation. Biotechnol Lett. 2006;29:27–35. as transferable operating criterion to improve recombinant protein 40. Potgieter TI, Kersey SD, Mallem MR, Nylen AC, d’Anjou M. Antibody production in Pichia pastoris through oxygen limitation. J Chem Technol expression kinetics in glycoengineered Pichia pastoris. Biotechnol Bioeng. Biotechnol. 2017;92:2573–82. 2010;106:918–27. 49. Chang S-W, Lee G-C, Shaw J-F. Codon optimization of Candida rugosa lip1 41. Curvers S, Brixius P, Klauser T, Thommes J, Weuster-Botz D, Takors R, Brun‑ gene for improving expression in Pichia pastoris and biochemical char‑ schier R, Kula MR. Human chymotrypsinogen B production with Pichia acterization of the purified recombinant LIP1 lipase. J Agric Food Chem. pastoris by integrated development of fermentation and downstream 2006;54:815–22. processing. Part 1. Fermentation. Biotechnol Prog. 2001;17:495–502. 42. Cámara E, Monforte S, Albiol J, Ferrer P. Deregulation of methanol metabolism reverts transcriptional limitations of recombinant Pichia pas- Publisher’s Note toris (Komagataella spp) with multiple expression cassettes under control Springer Nature remains neutral with regard to jurisdictional claims in pub‑ of the AOX1 promoter. Biotechnol Bioeng. 2019;116:1710–20. lished maps and institutional affiliations. 43. Shi L, Wang J, Wang X, Zhang Y, Song Z, Cai M, Zhou X. Transcriptome and metabolome analyses reveal global behaviour of a genetically Ready to submit your research ? Choose BMC and benefit from: • fast, convenient online submission • thorough peer review by experienced researchers in your field • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations • maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions