Microbioreactor-assisted cultivation workflows for time-efficient phenotyping of protein producing Aspergillus niger in batch and fed-batch mode

In recent years, many fungal genomes have become publicly available. In combination with novel gene editing tools, this allows for accelerated strain construction, making filamentous fungi even more interesting for the production of valuable products. However, besides their extraordinary production and secretion capacities, fungi most often exhibit challenging morphologies, which need to be screened for the best oper-ational window. Thereby, combining genetic diversity with various environmental parameters results in a large parameter space, creating a strong demand for time-efficient phenotyping technologies. Microbioreactor systems, which have been well established for bacterial organisms, enable an increased cultivation throughput via parallelization and miniaturization, as well as enhanced process insight via noninvasive online monitoring. Nevertheless, only few reports about microtiter plate cultivation for filamentous fungi in general and even less with online monitoring exist in literature. Moreover, screening under batch conditions in microscale, when a fed-batch process is performed in large-scale might even lead to the wrong identification of optimized parameters. Therefore, in this study a novel workflow for Aspergillus niger was developed, allowing for up to 48 parallel microbioreactor cultivations in batch as well as fed-batch mode. This workflow was validated against lab-scale bioreactor cultivations to proof scalability. With the optimized cultivation protocol, three different micro-scale fed-batch strategies were tested to identify the best protein production conditions for intracellular model product GFP. Subsequently, the best feeding strategy was again validated in a lab-scale bioreactor.


| INTRODUCTION
Starting with the commercial production of citric acid in 1919 where the first large-scale process utilizing filamentous fungi was established mostly without knowledge of basic molecular and bioprocess principles, 1,2 filamentous fungi have been established as one of the major workhorses in industrial biotechnology. Nowadays a billion dollar industry has been established, utilizing these cell factories for the manufacture of value-added products. 3 The natural portfolio of products is very diverse and consists of organic acids, secondary metabolites, antibiotics, proteins and enzymes. [3][4][5][6] A general advantage of filamentous fungi is their extraordinary secretion and production capacity. Especially Trichoderma and Aspergillus strains are known for their capability to secrete high amounts of enzymes, making them interesting platforms for industrial production. For example, up to 30 g L −1 of amylases and cellulases can be produced with A. niger and T. reesei, respectively. 7,8 More than 250 species are known in the genus Aspergillus 9 with industrially relevant examples being A. niger, A. oryzae, A. awamori, A.
sojae, and A. terreus. Generally, Aspergillus can withstand and endure harsh environmental conditions. With a wide range of tolerated temperatures (10-50 C), pH values (2)(3)(4)(5)(6)(7)(8)(9)(10)(11) and water activities, these organisms are very robust. [10][11][12] As reviewed by Ward et al., 13 members of the genus Aspergillus are capable of metabolizing various complex substrates, such as hemicellulose and cellulose, as well as pectin and xylan. In combination with their extraordinary production and secretion capacity, the utilization of renewable substrates allows for the efficient and potentially cheap manufacture of value-added products. Consequently, Aspergillus has become one of the key players for industrial biotechnology. Commercial applications are in different industries, such as food, pharma, feed, paper, pulp, and biofuels. 14 Since 2010 more than ten genomes of various Aspergillus species have been made publicly available. 15 In combination with novel highthroughput gene editing technologies being adapted for filamentous microorganisms, 16 an increase in the product portfolio and subsequently the industrial manufacture of novel products is to be expected. This enables the utilization of Aspergillus as a multi-purpose cell factory in the future.
In contrast to bacterial organisms, filamentous fungi follow a distinctive life cycle. Beginning with spore germination under beneficial environmental conditions, spore swelling is followed by germ tube formation.
These grow larger and form hyphae, compartmentalized by septa. At a certain size, hyphae start branching, leading to a complex interconnected hyphal network, the so-called mycelium. 14 Sporulation may again be initiated upon shifting to unfavorable environmental conditions. This change in morphology becomes especially apparent in submerged cultivation, creating challenging conditions for reproducible bioprocesses.
As thoroughly discussed by Veiter et al., 17 there is often a complex interplay between process conditions, predominant morphology and productivity during biotechnological production with filamentous fungi.
While pelletous growth has clear advantages with respect to culture broth viscosity and thereby also energy input needed for mixing and oxygen transfer, microfilamentous growth is beneficial for homogeneous nutrient supply. Although there is indication that pelletous morphology may be preferable for the production of organic acids while mycelial growth seems to support enzyme production, 18 there are no generally applicable rules established regarding preferable morphological state for manufacture of different products. Therefore, many screening experiments must be conducted making cultivation with an increased throughput highly desirable while integrated online monitoring can help to obtain better process insight and understanding.
Although microbioreactor (MBR) systems have proved to be valuable tools for such applications, 19 only few reports on their successful application for filamentously growing fungi can be found in literature, among these being for example Hortsch et al., 20 Huth et al., 21 or Jansen et al.. 22 In the study presented here, MBR-assisted workflows for parallelized microscale cultivation were developed for A. niger producing green fluorescent protein (GFP) as a model for heterologous protein production. Special emphasis was put on achieving reproducible process patterns under batch, as well as strictly controlled fed-batch conditions. As literature reports about improved enzyme production in A.
niger by tailoring processes toward mycelia rather than pellets, 18 maintaining microfilamentous morphology during microscale cultivation runs was targeted as the desired morphological state in this study.

| Chemicals, strain, and media
All chemicals were of analytical grade and obtained from either Sigma-Aldrich (Steinheim/Germany) or Roth (Karlsruhe/Germany).
Aspergillus niger anip7-gfp2 (formerly denoted as A. niger ARAn701) was kindly provided by Prof. Spieß and Mathias Papenfuß at the Technische Universität Braunschweig (Germany). The strain is deviated from the protease deficient A. niger AB1.13 23  Details on strain construction are provided by Driouch et al. 24 Yeast extract peptone dextrose medium (YEPD) consisted of 10 g yeast extract, 20 g peptone and 20 g glucose per liter water and was set to pH 3.8 with sulfuric acid.

| Strain maintenance
For spore generation, 100 μl of an existing spore suspension was plated on PDA plates and incubated at 37 C for 5-7 days in an incubation chamber. After full sporulation, the spores were collected through addition of 1 ml cryoprotectant solution containing 20% (w v −1 ) glycerol and 0.9% (w v −1 ) NaCl and scrapping with a spatula.
The suspension was filtered through a self-made cotton filter to remove loose mycelium. Subsequently, the spore concentration was determined utilizing a Neubauer chamber and adjusted to 10 8 spores ml −1 with cryoprotectant solution. Aliquoted stocks were stored at −80 C until further use.

| Microbioreactor cultivation
All microscale cultivations were conducted in either a BioLector I (batch and robotic dosing) or BioLector Pro (microfluidic feeding) MBR (m2p-labs, Baesweiler/Germany). Detailed description of the devices can be found elsewhere. 28 FlowerPlates are specialized microtiter plates for BioLector applications that provide high mass transfer similar to bioreactors as typically needed during aerobic cultivation. 31 For robotic dosing experiments, the MBR was integrated into a customized Freedom Evo 200 liquid handling platform (Tecan, Männedorf/Switzerland). The robot was equipped with polytetrafluoroethylene-coated steel needles that were used to repeatedly dose 10 μl substrate pulses (80 g L −1 maltose) into the culture wells with continuous shaking of the cultivation MTP while initial culture volume was reduced to 800 μl.

Microfluidic feeding experiments were conducted in microfluidic
FlowerPlates (MTP-MF32C-BOH1) sealed with special sealing foils (F-GPRSMF32-1) at 1400 rpm shaking speed, 35% headspace oxygen concentration and 800 μl initial culture volume. 400 g L −1 maltose was filtrated (0.2 μm cellulose acetate) to prevent potential blocking of microfluidic channels by any solid particles and served as feeding solution.
General conditions for all experiments were 37 C, ≥85% relative humidity and inoculation of cultures either to 10 5 spores ml −1 or with 10% (v v −1 ) preculture suspension as stated in the individual experiments. Biomass and GFP were non-invasively monitored via scattered light at 620 nm wavelength, hereinafter referred to as "backscatter" and fluorescence (488/520 nm), respectively. In case of volume changes due to robotic or microfluidic feeding, measured values were normalized to the initial culture volume. Dissolved oxygen (DO) was measured via immobilized sensor spots on the bottom of the plate. It has to be noted that absolute values from intensity measurements (backscatter and GFP) often differ from device to device due to technical reasons and are therefore not comparable across different machines. To compensate for that issue, individual experimental runs have been executed with a reference batch process, which was used for data normalization and enabled cross comparison of cultivation data from different runs. Therefore, figures from Sections 3.3 to 3.5 show individual reference batches instead of referring to a single, general reference.

| Lab-scale cultivation
All stirred tank bioreactor cultivations in this study were carried out in

| Cell dry weight
1000 μl cell suspension was loaded onto a pre-dried (24 h at 80 C and subsequently cooled to room temperature in a desiccator) and pre-weighed centrifuge filter tube (SpinX, Costar, New York) with a 0.22 μm cellulose acetate membrane. The tubes were centrifuged for 3 min at 13,000 g. The cell-free supernatant was aliquoted and stored at −20 C for further analysis. The retentate was washed twice with 500 μl 0.9% (w v −1 ) NaCl and the flow-through was discarded. The washed pellet was then dried at 80 C for 24 h, cooled to room temperature in a desiccator and weighted on a precision scale. Cell dry weight (CDW) was calculated from the mass difference.

| Offline pH
Offline pH was measured electrochemically in crude cell suspension utilizing an S20 SevenEasy pH meter with a 6.0234.100 micro electrode (Metrohm, Filderstadt/Germany). All samples were measured at room temperature.
The photometer was set to an excitation wavelength of 488 nm and emission wavelength of 520 nm to detect GFP fluorescence in technical triplicates. A filling volume of 100 μl crude cell suspension was used for each measurement.

| Statistical analysis
The relative mean coefficient of variation of backscatter measurements was calculated as an indicator for reproducibility. The arithmetic mean x i of eight biological replicates and the corresponding standard deviation s i of each backscatter measurement was calculated for each measurement cycle i. The coefficient of variation c v,i was calculated according to Equation (1).
The sum of c v,i for all data points with an increase in scattered light measurement above the limit of detection (α = 0.01; i = 1) until the end of cultivation (n) was divided by the number of cycles (m) to calculate relative mean coefficient of variation (Equation (2)).  195% was observed until reaching a maximum at 30.6 h process time.
This timeframe goes far beyond reported maturation times for GFP 34 giving strong evidence for sustained expression of the target protein.
Strikingly, around two thirds of total target protein production took place after metabolization of the initially provided maltose. Although this finding was not investigated in further details at this point, it seems very likely that heterologous protein expression during "stationary phase" might be driven by usage of secondary carbon sources or storage compounds as already hypothesized for biomass increase seen during this phase (Figure 3a).  (Figure 4). Consequently, backscatter may be used as a valid online measure for A. niger biomass concentration.

| Biomass calibration for microbioreactor cultivation of A. niger
Here, it must be emphasized that calibration was not done by a dilution series of culture broth from one single point of the cultivation and thereby state of morphology (e.g. after harvest) but proved applicable along the complete culture process.

| Comparison of A. niger microbioreactor and lab-scale bioreactor cultivation
Having established a reliable setup for batch cultivation of A. niger in online-monitored microtiter plates (Section 3.1) enables for accelerated process development and phenotyping of such fungi. To use and apply the newly generated data sensibly, the transferability to conventional, stirred lab-scale bioreactors must be guaranteed.
Therefore, the cultivation conditions described at the end of

| Development of fed-batch processes for microscale cultivation of A. niger
The relevance of fed-batch cultivations already at a small scale has been discussed in literature such as Scheidle et al. 35 Of course, this also applies for filamentous fungi such as A. niger, since an increasing number of processes are performed in fed-batch mode, for example, to limit mass transfer and cooling capacity or for the sake  42 However, for the cultivation of A. niger it seems advantageous to utilize FlowerPlates to increase shear forces toward microfilamentous morphology and to provide the mass transfer needed to avoid an oxygen limitation. Therefore, two alternative strategies were tested to enable microscale fed-batch for A. niger: As the first one, robotic-assisted feeding was tested. Here, small pulses of a concentrated sugar stock are dosed into each cultivation well on either pre-defined time intervals or triggered by the respective DO signal. 43  Target protein production as monitored by GFP fluorescence slightly increased up to a value of 1 a.u. being coupled to growth until the depletion of primary carbon source maltose as indicated by the spike of the DO signal after approx. 15 h. However, product formation increased strongly during the subsequent phase. GFP fluorescence reached its maximum of approx. 4 a.u. after 30 h of cultivation, at which time the DO signal also went back up toward 100%, indicating the end of metabolic substrate consumption. Please note, that it seems that approx. 80% of total GFP signal was observed after primary growth phase.
After 30 h, the GFP signal began to steadily decrease which might be explained by a change in intracellular pH: Under regular metabolic conditions, energy metabolites can be used to drive active ion channels for maintaining intracellular pH homeostasis within the neutral regime. 45 However, upon depletion of all carbon sources including storage compounds and/or secondary carbon sources, the resulting exhaustion of energy metabolite pools might result in the cessation of the activity of the ion pumps. As a result, the intracellular proton level begins to equalize to extracellular acidic levels resulting in a loss of GFP fluorescence. 46 The pulsed feeding strategy was conducted as stated: Following an initial batch phase on 5 g L −1 maltose, pulsed feeding was started after 15 h. The backscatter increased linearly until the last pulse 4.5 a.u. Afterward, the GFP fluorescence decreases strongly, as previously seen, most likely due to the loss of intracellular pH homeostasis.
With respect to the measures described, the reference batch was very similar to previous results ( Figure 6).
The fed-batch cultivations are prolonged due to carbon-limited feeding until 30 h but at harvest backscatter values comparable to the batch reference were achieved. Strikingly, the microfluidic pulsed feeding regime also resulted in GFP production patterns very similar to the batch reference which is in direct contrast to the robotic pulsed feeding where only reduced titers could be achieved compared to the robotic workflow, so that performance loss as observed during robotic feeding could be circumvented. In case of microfluidic feeding at constant rate, the product formation was further increased up to 6.02 a.u. (+33%) confirming the hunch that strict carbon limitation and thereby avoidance of glucose repression or slowing down of growth in general is highly beneficial for target protein expression.

| Lab-scale validation of fed-batch processes
To validate the increased productivity achieved by constant rate feeding at microscale and to check for its transferability, lab-scale cultivations with identical feeding profiles were conducted. After an initial batch of 5 g L −1 that lasted approx. 8  online monitoring, offline CDW and GFP fluorescence was determined ( Figure 8).
Analogous to the MBR cultivation, the batch reached the stationary phase after approx. 14 h and thereby earlier than the fed-batch process. In both cases, very similar maximal CDWs in the range of 9 g L −1 were obtained. While the batch process achieved this concentration toward the end of the maltose consumption phase at 14 h (see DO data in Figure S2) the fed-batch cultivation reached its maximal CDW right at the end of the feeding phase after approx. 27 h. In both cases, a decrease of biomass concentration down to approx. 7 g L −1 during subsequent starvation until harvest could be observed.
Even though both cultivations resulted in comparable biomass concentration, the maximum product concentration was increased by 76% for the constant feed. Clearly, this improvement goes beyond the performance differences observed at microscale (33%). As the reason for this difference remained unclear from the data available, further investigation on the microscale protocol seems promising to decipher the relevant effect and maybe achieve identical results as in stirred tank bioreactors. Nevertheless, current results strongly suggest the valid application of feeding strategies for A. niger already at microscale as the general trend could be reproduced.

| CONCLUSIONS
Microbioreactor cultivation proved to be a valuable tool for microfilamentous A. niger cultivation. Good transferability between batch microscale in shaken MTPs and lab-scale cultivations in stirred tank bioreactors enabled scalability of the results. Acquired offline parameters, such as CDW, pH and model product GFP resulted in a close overlay for both systems. Moreover, online measurements of scattered light for MBR cultivation were highly reproducible with 8.3% rmc c and showed a good correlation to offline CDW, which was not impacted by morphological changes. For reasons yet unknown approx. Two thirds of total GFP formation was observed after depletion of the initial carbon source. Based on the information available to now, it can only be speculated that this is attributable to metabolization of secondary carbon sources or storage compounds, which might affect protein expression or GFP maturation kinetics and should therefore be investigated in further detail.
In the next step, the implementation of different feeding strategies to enable microscale fed-batch cultivation was tested. Both robotic pulsed feeding as well as substrate addition via microfluidic channels resulted in reproducible fed-batch cultivation. Despite technical applicability, for the given biological system producing intracellular GFP, a clear benefit of microfluidic constant feeding with 2 μl h −1 could be seen, as protein production was increased by 33% in contrast to batch cultivation. While microfluidic pulsed feeding did not improve total product titer compared to the batch reference, robotic pulsed feeding even resulted in a reduced protein production (−50%).
It therefore seems that stringent substrate limitation during microfluidic constant rate feeding is superior to pulsed profiles where product formation is repeatedly repressed by oscillations in substrate concentration.
Overall, microbioreactor cultivation, both in batch or fed-batch mode, enable accelerated phenotyping for filamentous fungi A. niger through miniaturization and parallelization. However, further effort seems necessary to improve comparability of fed-batch processes between both scales. As a next step, the combination of other workflows, such as automated morphology analysis, omics analysis and exometabolome fingerprinting with carbon-limited fed-batch cultivations would allow for deeper insights as well as better phenotyping at microscale.

CONFLICT OF INTEREST
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this article.

DATA AVAILABILITY STATEMENT
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.