Bioreactor Operating Strategies for Improved Polyhydroxyalkanoate (PHA) Productivity

Microbial polyhydroxyalkanoates (PHAs) are promising biodegradable polymers that may alleviate some of the environmental burden of petroleum-derived polymers. The requirements for carbon substrates and energy for bioreactor operations are major factors contributing to the high production costs and environmental impact of PHAs. Improving the process productivity is an important aspect of cost reduction, which has been attempted using a variety of fed-batch, continuous, and semi-continuous bioreactor systems, with variable results. The purpose of this review is to summarize the bioreactor operations targeting high PHA productivity using pure cultures. The highest volumetric PHA productivity was reported more than 20 years ago for poly(3-hydroxybutryate) (PHB) production from sucrose (5.1 g L−1 h−1). In the time since, similar results have not been achieved on a scale of more than 100 L. More recently, a number fed-batch and semi-continuous (cyclic) bioreactor operation strategies have reported reasonably high productivities (1 g L−1 h−1 to 2 g L−1 h−1) under more realistic conditions for pilot or industrial-scale production, including the utilization of lower-cost waste carbon substrates and atmospheric air as the aeration medium, as well as cultivation under non-sterile conditions. Little development has occurred in the area of fully continuously fed bioreactor systems over the last eight years.


Introduction
The production of completely biodegradable materials from renewable resources is an urgent challenge for the mitigation of the cradle-to-grave environmental impacts associated with petroleum-based plastics. In 2015, the worldwide production of synthetic plastic materials exceeded 320 million tonnes [1], and at current consumption rates, it is estimated to reach 33 billion tonnes by 2050 [2]. Of the amount currently produced, less than half is recycled or landfilled [2]. An astonishing amount of these materials are unaccounted for, and persist in natural ecosystems due to their high molecular weight and recalcitrance to biodegradation [3][4][5], which can have multiple detrimental impacts on both terrestrial and aquatic ecosystems. It is estimated that 10 to 20 million tonnes of plastic material is discarded into oceans alone [6]. Once in the natural environment, plastics can absorb and concentrate persistent organic pollutants [7], and release toxic additives that are not chemically bound in the polymer [8,9]. Through mechanical abrasion, plastics are broken down into microsized to nanosized particles that bioaccumulate in filter-feeding animals, posing a significant threat to marine ecosystems [6]. Recent evidence suggests that particles <10 µm may pose significant risks to human health through inhalation and the penetration of pulmonary surfaces [10,11].
Microbial polyhydroxyalkanoates (PHAs) are a class of bio-based and biodegradable polymers that may help displace certain applications for traditional plastic materials [12][13][14]. PHAs are Table 1. Chronological summary of process developments for high-productivity fed-batch bioreactor systems in medium-chain length polyhydroxyalkanoate (mcl-PHA) research.

Ref.
Organism Substrate; Limitation [X t ] (g L −1 ) % PHA (% CDM) Y PHA/S (g g −1 ) Q v (g L −1 h −1 ) C L * Contribution [70] P. putida GPo1 Octane; N-limited 37     Despite some promising results for fed-batch mcl-PHA cultivations, as shown in Table 1, a comparison with Table 2 shows that, in general, the Q v of mcl-PHA cultivations is low when compared with scl-PHA production processes. Ryu et al. [84] reported achieving [X t ] of more than 280 g L −1 CDM with [X PHB ] of more than 230 g L −1 (% PHB = 83% CDM) using C. necator grown on glucose. While these are the highest known experimental values for [X t ] and [X PHA ], Wang and Lee [49] reported the highest Q v (5.13 g L −1 h −1 ) using A. lata (formerly Alcaligenes latus) cultured on sucrose. By comparison, MacLean et al. [48] have reported the highest mcl-PHA Q v from inoculation to harvest at 2.13 g L −1 h −1 . As shown in Table 1, the [X t ] seldom exceeded 100 g L −1 in mcl-PHA cultivations, and only one study has reported [X PHA ] of more than 70 g L −1 [75]. Generally, mcl-PHA cultivations appear to be characterized by lower [X t ] and lower % PHA . The reasons for the productivity discrepancy between scl-PHA and mcl-PHA production remains unclear, but the % PHA might be explained by differences in granule morphology [21,72].
Interestingly, the number of studies targeting cultivation strategies for achieving the highest possible Q v appeared to peak in the mid-1990s for scl-PHAs, and perhaps 10-15 years later for mcl-PHAs. While these studies are valuable tools for understanding process engineering design and limitations, they are hindered by a few key impracticalities. These include the use of high-purity carbon substrates as well as using O 2 -enriched air for aeration. More recently, there appears to be somewhat of a shift toward PHA production from waste substrates such as crude glycerol [90,99], plant-based oils [77,79,94,95,102], and lignocellulose-based hydrolysates [98], and devising strategies to achieve HCD cultures without use of purified O 2 streams [61,62,103].

Control of Fed-Batch Cultivations
A significant challenge in fed-batch cultivation is predicting the nutritional requirements of the culture in real-time so that an appropriate feeding strategy can be implemented. Direct and rapid measurement of the parameters of interest (i.e., [X t ], % PHA , residual concentrations of carbon, NH 4 , PO 4 3− , Mg 2+ , etc.) is often not possible, or the existing methods require expensive analytical equipment with a considerable time delay [104][105][106]. Nonetheless, some studies have used online or rapid offline measurements of residual carbon levels using analytical techniques such as gas chromatography (GC) [73,82], high-performance liquid chromatography (HPLC) [89,91], and mass spectrometry (MS) [81]. Others have made use of online glucose analyzers [57,63,84,88,107].
Most often, feeding rates are predicted indirectly through using other parameters that can be rapidly and reliably measured using standard bioreactor equipment. Some of the more popular methods involve the use of off-gas CO 2 measurements [79,80,107,108], dissolved oxygen (DO) measurements [49,67,77,109], pH measurements [53,54], optical density [58,75], and combinations thereof [58,75,101]. In general, with these approaches, carbon limitation results in reduced oxygen demand, which manifests itself as a large drop in CO 2 production, reduced agitation rates, and also a rise in the DO signal. A rise in pH may also be detected with the exhaustion of the carbon source, assuming a negligible impact on pH from the production of acetate, amino acids, and the source of nitrogen [76]. With these methods, the carbon source is often fed pulse-wise upon detection of a limitation in order to bring the concentration back up to the desired level [49,53,58,86,94,95,98,102,110].
Interestingly, it appears that the chosen method of carbon addition can influence PHA production. For example, Możejko and Ciesielska [110] showed a significant increase in % PHA (from a maximum of 25% CDM to 44% CDM) in Pseudomonas sp. G101 when the carbon source (waste canola oil) was fed to the bioreactor by a pulse-feed strategy in comparison to a continuous drip-feed addition of an equivalent aliquot of waste canola oil. Although this type of feast-famine process is often applied to PHA-storing populations in mixed microbial cultures [40], these results may suggest that small-scale feast-famine is favorable for PHA storage in pure cultures.
Alternatively, a more proactive strategy may be applied in which the culture is fed in order to maintain a predetermined (typically exponential) growth rate. This method has been quite successful in several studies, and has achieved [X t ] nearing 100 g L −1 CDM for growth of Pseudomonas sp. on both fatty acid and carbohydrate-based carbon sources [48,61,62]. Typically, these models require the initial [X t ], the change in µ over time, as well as establishing constant [59], linear [60,61], or decaying [48,72] feeding rates during the PHA accumulation phase (see Table 3). According to MacLean et al. [48], sudden decreases in feeding rates result in cell lysis and foaming, so it is important to make the transition between different stages of feeding smoothly. Several feeding strategies have been investigated, and are shown in Table 3.
Significant efforts have gone into developing various models that allow the culture to grow as quickly as possible at the start of the process, before curbing the feeding rate to match the transient carbon uptake rate as the culture transitions to PHA storage. Sun et al. [55] have shown that controlling the growth rate via the rate of carbon addition allows higher [X t ] to be obtained before the process must be terminated due to inadequate oxygen mass transfer. Since lower growth rates are used, there is a trade-off in biomass productivity, but the constant carbon limitation is reported to cause simultaneous PHA accumulation and eliminates the need for the limitation of another nutrient to accumulate PHA. The onset of oxygen limitation inevitably limits Q v in HCD cultivations for PHA production and causes uncontrollable foaming and accumulation of the carbon source to potentially toxic levels (depending on the feeding strategy) [55,56,62,75,109,111]. Table 3. Various feeding strategies used to obtain HCD cultures in mcl-PHA production.

Ref.
Feeding Strategy Results [48] Exponential feeding of NA: (1) From 0 to 9 h starting at µ = 0.25 h−1 followed by a linearly decaying feeding rate over the next 21 h (2) Quadratic decay in feeding rate (5 to 15 h) with initial µ = 0.45 h−1 followed by a constant feed rate [61] Several exponential glc feeding strategies evaluated: (1) At a constant specific growth rate of 0.25 h−1 until dissolved oxygen (DO) limitation (2) At µmax (0.67 h−1) for 0 to 9 h followed by constant feeding when DO became limiting (3) At µmax (0.67 h−1) for 0 to 9 h followed by a linearly increasing feed rate when DO became limiting (4) Same as in (3)

Limitations of HCD Cultivations: The Oxygen Transfer Issue
Perhaps the biggest challenge with the control of HCD fed-batch cultures is maintaining an adequate supply of dissolved oxygen (DO). Gas-to-liquid mass transfer is driven by the concentration gradient between the liquid phase (C L ) and the equilibrium concentration (C L *), the liquid film mass transfer coefficient (k L , i.e., distance time −1 ), and the interfacial area between the gas and the liquid (a), as described in Equation (1) [113]: As shown in Equation (2), if k L a is known, the oxygen uptake rate (OUR) of the culture can be estimated from the rate of change of DO with respect to time. Assuming no buildup of O 2 in the liquid phase (i.e., steady-state conditions) the OUR can be assumed to be equal to the oxygen transfer rate (OTR), which can be written as shown in Equation (2) [114]: With the onset of O 2 limitation being problematic in strictly aerobic bioprocesses such as PHA production, many studies have focused on enhancing the OTR to achieve faster growth rates, higher cell densities, and ultimately higher productivity. Implicit from Equation (2) is that several approaches may be taken to do this, including the manipulation of C L *, a, or k L . All of the fed-batch studies use very high stirring rates to enhance the OTR by increasing both the tortuosity of the path taken by the bubble and the residence time of the bubble in the liquid phase (gas hold-up, which can be measured as the total liquid volume that is occupied by the gas). While this increases k L , stirring also helps to increase a by breaking up large bubbles. This requires a significant input of mechanical energy to the bioreactor, and is frequently used in conjunction with other techniques. Many studies also increase the driving force by increasing the C L *. As previously mentioned, this has included the use of an aeration medium with enriched O 2 content (as noted in Tables 1 and 2), as well as pressurization of the bioreactor headspace [79,115,116]. Interestingly, while pressurization increases C L *, it has also been shown to have a detrimental impact on k L a [117]. Other approaches involve increasing a using microbubbler devices [118]. Although not used in a PHA production context, Baker et al. (2016) used a microfluidic device to generate bubbles with an average diameter of around 110 µm, and obtained nearly ideal mass transfer (i.e., 90% of the oxygen in the gas phase was delivered to the liquid phase), which is a promising development for industrial biotechnology in general [119].
Other approaches have focused on modifying the medium through the addition of a second phase (liquid or solid) that effectively increases the solubility of O 2 in the medium. Magnetic functionalized nanoparticles have been used to increase OTR (both k L and a) in bioreactors [120]. However, no reports could be found on the use of functionalized nanoparticles to improve PHA productivity. Two-liquid phase systems, in which small quantities of certain immiscible liquids are dispersed into the aqueous medium, have been reported to enhance OTR [121][122][123][124]. Interestingly, soybean oil (which along with other plant-based oils are high-yielding PHA substrates) has been found to significantly improve k L a when dispersed in aqueous medium [125,126]. A few studies have investigated two-liquid phase bioreactor systems in PHA production [70,74], but in those cases, n-octane served as the carbon source, and its possible effect on O 2 solubility was not examined.
Genetic engineering strategies have also been explored to allow the synthesis of PHA to be more favorable at low DO conditions. Wei et al. [127] cloned several anaerobic promoters upstream of the phbCAB operon in E. coli. That study found that a strain harboring the promoter for alcohol dehydrogenase (in addition to phbCAB) increased the % PHB of a microaerobic (sealed flasks with no shaking) from 30% to 48% CDM, and this was further improved to 67% CDM when the acetate pathway was deleted. Although not done for the specific purpose of PHA production, Schmitz et al. [128] cloned phenazine redox mediators from P. aeruginosa into P. putida KT2440. This allowed P. putida to maintain oxygen-limited metabolism for two weeks by partial redox balancing in the presence of an anode in low DO environments, and could be a promising development for P. putida as a biocatalyst for HCD and large-scale PHA production.
In summary, maintaining adequate DO at high cell density is a significant challenge for commercial-scale PHA production. As discussed in this section, this is due to a combination of the low solubility of O 2 in aqueous medium and poor gas-to-liquid mass transfer of O 2 . Although a number of techniques, some of which appear quite promising, may be applied to delay the onset of O 2 -limited conditions and achieve higher cell densities and increased productivities, there are drawbacks to many of these techniques. In particular, the use of purified O 2 can add significant fixed and operating costs to PHA production [51,116]. Bioreactor pressurization may be a more attractive alternative, but this requires specialized equipment and also makes feeding less straightforward due to the need for bioreactor addition ports that are compatible with pressure. The use of microbubble devices can significantly improve gas-to-liquid mass transfer, but can also require high-energy input [129] and be prone to fouling [130]. Even the high mechanical energy that is required for vigorous agitation and aeration accounts for a large portion of the environmental impact and cost of PHA production [34]. An unfortunate paradox is that many of these developments are not cost-effective to implement in large-scale cultivation systems, which have inherently poor mass transfer characteristics [131][132][133]. Studies that have scaled up fed-batch processes in PHA production are discussed in the following section.

Scale-Up Fed-Batch PHA Production
As discussed above, the PHA productivity of bench-scale fed-batch operations is almost inevitably limited by the bioreactor OTR. Such limitations are exacerbated with increasing scale and represent some of the biggest engineering challenges for bioprocess scale-up. There are many criteria that can be used for the basis of scale-up, including similar geometry, constant impeller tip speed, constant Reynolds number, constant k L a, circulation time, or constant power input (either gassed or ungassed) per unit reactor volume [134]. Ultimately, all of these criteria are simply different ways of measuring energy requirements, which is the limiting factor in scale-up. For example, according to Lara et al. [133] performing a scale-up from 80 L to 1000 L on the basis of constant circulation time could be done, but would result in a 25-fold increase in power dissipation per unit volume. The lesser power dissipated for mixing in large-scale bioreactors creates non-homogeneous physical and chemical environments for microbes, and generally results in productivity losses of 10-30% at increased scale for many bioprocesses. In a conventional bioreactor design, these gradients usually occur in the vertical direction, since feeding is done from the top and aeration is done from the bottom, and significant hydrostatic pressure near the bottom can also influence gas solubility [135]. This complicates the measurement of important process variables, because the measurement results can be highly dependent on the location of the probes [131]. In order to anticipate some of these challenges, scale is increased by no more than one order of magnitude at a time [134].
Despite these challenges, the ability to demonstrate scalable PHA cultivation techniques is important to reduce costs and improve the economic viability of PHA markets. According to Choi and Lee [51], increasing scale can significantly decrease costs (both operating and capital costs) for production capacities of up to 15,000 tonnes of PHA per annum. In the literature reviewed so far, the majority of the focus in PHA research has been on strategies to improve productivity in lab-scale bioreactors, typically with working volumes of 5 L or less. Studies that have investigated some aspect of scale-up are relatively few, but are shown in Table 4 for a variety of cultivations, including both mcl-PHA and scl-PHA production at scales ranging from at least 100 L to 10,000 L. Relatively few studies have reported this kind of scale-up of at least 100 L. If the bioreactor size is set a lower limit, such as 30-50 L, Table 4 would include quite a few more studies [53,77,84,91,109,136]. Table 4. Chronological summary of studies that have examined scale-up of various processes for scl-PHA or mcl-PHA production.

Ref.
Organism, Product Conditions  Table 4, scl-PHA productivities of around 1 g L −1 h −1 have been demonstrated in a few studies at the 300-L (100-150 L working volume) scale [139,140], whereas 0.75 g L −1 h −1 has been demonstrated for mcl-PHA production at a similar scale [112]. These values are less than half the Q v values reported in Tables 1 and 2 for similar processes. This is likely because many of the studies in Tables 1 and 2 have used purified O 2 as the aeration medium, which can increase the driving force for oxygen transfer nearly five-fold [146], and is usually impractical in larger scale vessels. This is exemplified by none of the scale-up studies in Table 4, other than one study that used a slightly pressurized headspace, using an O 2 -enriched aeration medium. A recent study performed a k L a-based scale-up and obtained a Q v of 1.66 g L −1 h −1 (P(3-HB-co-4-HB) in a 5000-L bioreactor, which is perhaps the most promising pilot-scale result in PHA production to date [145]. Interestingly, in that same study, the Q v obtained at 5000 L was slightly higher than that which was achieved at the 1000-L scale, and was significantly better than the result achieved at lab-scale (7.5 L, see Table 4). However, in this study, the estimated k L a in both the 1000-L and 5000-L bioreactors was similar, and significantly better than the 7.5-L bioreactor.

As indicated in
Shang et al. [141] evaluated growth and PHB synthesis from C. necator using air compared to purified O 2 as the aeration medium in 5-L and 30-L bioreactors, as well as a 300-L bioreactor where air was the sole aeration medium. Although the focus of that study was the effects of CO 2 inhibition arising from poor liquid-to-gas mass transfer at lower gas flow rates (used to minimize the use of purified O 2 ), the study presents a useful comparison to help deconvolute the effects of increasing the driving force for oxygen transfer by gassing the reactor with pure O 2 . The use of pure O 2 instead of atmospheric air doubled [X t ] nearly four-fold (from 96 g L −1 to 208 g L −1 ) in the 5-L bioreactor and more than tripled [X t ] (from 49 g L −1 to 186 g L −1 ) in the 30-L bioreactor. Furthermore, Q v displayed a two-fold increase from approximately 1 g L −1 to 3 g L −1 h −1 in both the 5-L and 30-L bioreactors. Interestingly, the PHB content was elevated when pure O 2 was used at both the 5-L and 30-L scale. This is somewhat unexpected, since PHB is known to be synthesized from glucose in low-DO environments [147][148][149]. Oxygen limitation became very prevalent at the 300-L scale, which achieved two-fold less biomass than the 30-L scale, and four-fold less biomass than the 5-L scale when air was used as the sole aeration medium.
It is worthwhile mentioning that Shang et al. [141] concluded that there was a significant effect of CO 2 inhibition on both growth and PHA synthesis in C. necator. Follonier et al. [115] also evaluated the effect of elevated CO 2 concentrations in a pressurized bioreactor system, and concluded that dissolved CO 2 concentrations of up to 540 mg L −1 did not negatively affect mcl-PHA synthesis, but did result in decreased µ in cultures of P. putida KT2440. Both of these studies suggest that the buildup of CO 2 from liquid-to-gas mass transfer limitations also merit consideration during process scale-up, since they have been shown to negatively impact both growth and PHA synthesis.

Continuous and Semi-Continuous Bioreactor Systems
In continuous culture, fresh medium is constantly supplied to the bioreactor, and a portion of the culture is removed at the same rate: the dilution rate (D). The conditions inside the reactor (substrate, cell, and product concentrations) remain at steady state, and because of this, continuous cultures are often referred to as chemostats, which is short for "chemical environment is static" [150]. Steady-state operation has a number of advantages. It is an attractive platform to study bioprocess physiology and implement control strategies due to the constant growth rate, and is less laborious to operate and maintain once steady-state operation is reached [15,150,151]. Steady-state operation also circumvents the major challenge of the optimal control of fed-batch strategies, which involves predicting growth rates over time and under highly dynamic conditions [50].
However, the main reason that continuous bioreactor operations may be of interest is that they can theoretically achieve higher average productivity (over time) in comparison to discontinuous processes [152]. This is done by maintaining the culture at a moderate but constant growth rate for prolonged periods and avoiding the non-productive downtime that is required for harvest, cleaning, preparation of new medium, sterilization, cooling, and culture lag in a discontinuous process [15,150].
Ultimately, higher productivities could result in a smaller reactor volume requirement, less capital investment (for both the production and downstream processing requirements), lower operating costs, and would avoid some of the typical mass transfer issues associated with the scale-up of discontinuous processes, which have already been discussed [150,[152][153][154]. Furthermore, the manipulation of chemostat cultures has also been shown to reproducibly tailor monomer composition [153,155,156] and M w [157], giving the process considerable flexibility.
Despite the significant upsides of continuous cultivation processes, the experimental values for PHA productivities have not reached the levels that have been reported for fed-batch cultivation. [153].
There are a number of other challenges with continuous culturing that have perhaps limited its industrial application. These include: (1) the higher risk of infection due to a large number of fittings and pumping processes; (2) the risk of back-growth into the feed medium reservoir; (3) maintaining genetic stability of the production strain over many generations; and (4) cell washout as the D approaches µ max [50,158,159].

Continuous Processes in PHA Production
The D is an important parameter in continuous systems, and is defined as the quotient of the bioreactor working volume to the incoming flow rate (h −1 ). This can have a significant impact on µ, % PHA , and hence Q v [160,161]. Unless cell retention techniques are employed, the D is equal to both the hydraulic retention time and solids retention time, and must be lower than µ max to prevent cell washout or wasting the carbon source (which unless recycled must be subject to post-treatment to remove the biochemical oxygen demand of the waste effluent). On the other hand, µ must be maintained high enough to allow productive growth, keep toxic substrates diluted, and remove any accumulation of toxic end products.
It must be acknowledged that the application of continuous culture for PHA production is a significant challenge. In general, PHA synthesis is not growth-associated, and cannot therefore be simultaneously optimized (along with growth) in a reactor operating at steady state [50,150]. This may be the reason that continuous cultures are often applied as a tool to study physiology rather than an operating strategy to advance PHA productivity. One solution has been to implement a two-stage process in which the two phases (growth and PHA synthesis) are separated spatially by adding a second reactor in-series, where conditions favorable to PHA accumulation (i.e., high C/N or C/P ratios) can be implemented. This has been done in multiple instances with promising results [153,162,163]. One study producing PHB from glucose using C. necator has used a cascade of five bioreactors in series (i.e., to simulate plug flow) in order to optimize % PHA [164].
These, and other past studies of PHA production (both mcl-PHA and scl-PHA) in continuous cultures are summarized in Table 5. It is worth noting that Ramsay et al. [159] carried out the original study of mcl-PHA production in continuous culture as a way to circumvent the toxicity effect of related substrates using P. putida GPo1 grown on octanoate. While this study showed a relationship of decreased % PHA with increased D, % PHA was low, and ultimately, it was concluded that a fed-batch strategy should be pursued. This group of researchers subsequently developed high-productivity fed-batch strategies (Table 1).
Reviewing Tables 1 and 2 reveals that Q v (for both mcl-PHA and scl-PHA) is significantly lower in continuous processes (to date) than in fed-batch processes. However, it is important to note that fed-batch productivities are always reported on a time-scale from inoculation to harvest. To be able to compare these values with those in Table 6, productivity should be amortized over time from one harvest to the next, factoring in the downtime for cleaning, preparation, sterilization, cooling, and preparation of the inoculum for the subsequent cultivation. Follonier et al. [115] investigated the effects of increased pressure on mcl-PHA production from a combination of octanoic and undecenoic acid in a chemostat culture. Although a reasonable productivity was obtained (Table 5), it was estimated that with some further optimization to increase the OUR (higher aeration), the mcl-PHA productivity in this system could be pushed to as high as 11 g L −1 h −1 , assuming O 2 to be the only limiting factor. While this number would indeed be a remarkable development, to our knowledge, this has not yet been done. Table 5. Chronological summary of continuous culture studies in PHA production.

Ref.
Organism, Substrate (Polymer) Conditions Contribution [149] A. beijerinckii, glc (PHB) Single-stage, O 2 -limited with D = 0.05 h −1 -45 0.13 -First application of PHA production in a chemostat. [165] C. necator DSM 545, glc + propionate, Single-stage with N-limited conditions D = 0.15 h −1 -33 --Demonstrated copolymer production dependent on the ratio of feed components, also first implementation of a two-stage system for improved % PHA .  First study to use continuous scl-PHA production as a platform for improved Q v . Showed effect of the dilution rate on HV content.    First time more than two in-series CSTRs used for PHA production. Highest reported Q v for any continuous PHA production system.

Cyclic/Repeated (Semi-Continuous) Processes Culture in PHA Production
Repeated or cyclic batch and fed-batch processes could also be adopted as a method to increase Q v by decreasing the non-productive downtime between batches. Although still a discontinuous process, the growth and PHA production phases can be accomplished in a single bioreactor operated under transient conditions. With this type of operation, a portion of the reactor is decanted and refilled with fresh medium to restart the process immediately after the end of the last cycle [47]. This principle is similar to sequencing batch reactors (SBRs), which have become a popular culture selection tool with subsequent scl-PHA production using mixed microbial cultures (MMCs), allowing for the quick and easy manipulation of process conditions to favor PHA production [40]. Although developed primarily as a selection tool for robust scl-PHA-producing populations from MMCs, they may also be a reliable method for producing mcl-PHA effectively using pure cultures. Although harvest is done at intervals (i.e., discontinuously), there is still a considerable time-saving advantage by using the remaining broth as a seed for the subsequent reactor, eliminating the time required to clean, sterilize, and prepare new inoculum [173]. Several instances of this type of process are shown in Table 6, and some have reported highly competitive productivities. However, in several cases, only a few cycles were done [169,171], and the cultures did not appear to achieve a steady state. Ideally, more prolonged operation would be required to take advantage of the time saved by using a cyclic approach.
Since waste carbon sources are usually diluted (lactose in whey solution, for example), often these cyclic processes are employed in conjunction with a cell-retention technique (typically an external membrane module) to overcome the volumetric limitation imposed by feeding a diluted carbon source [168,175,176]. This is generally not the case with the studies reported in Tables 1 and 2. Cell-retention techniques (i.e., a membrane filter) allow the decoupling of the solids' retention time from the D, meaning that cells can be exposed to more carbon and thus accumulate higher % PHA , while avoiding washout with the effluent. As shown, some of these studies have reported cell densities approaching 150-200 g L −1 , with PHB content of up to 87% CDM.

Sterility Challenges
One of the challenges of semi-continuous or continuous operation is the increasing risk of contamination [178]. Although pure cultures may be preferred in PHA production because of the reproducibility of the process and products, the cost of sterilization processes is one of the main disadvantages compared with mixed culture PHA production [179]. However, if an environment can be established that precludes or minimizes the growth of competitors, then the lack of a need for sterilization could be a cost-saving measure for pure culture production as well. Such techniques are used in PHA production with MMCs, in which feast-and-famine cycles (aerobic dynamic feeding) are used to select for PHA-producing bacterial populations, because the ability to store carbon gives them a competitive advantage under starvation conditions [180]. Tan et al. [181] isolated a halophilic bacterium (Halomonas bluephagenesis TD01) from a salt lake in China. This organism was cultured under non-sterile conditions (37 • C, pH 9.0), and produced 80 g L −1 CDM containing 80% PHA in 56 h using a fed-batch method. In a two-stage continuous bioreactor, the steady-state concentration of biomass in the first reactor was 40 g L −1 with 60% PHB content, and PHB content rose to 70% CDM after being transferred to the second stage with nitrogen-limiting conditions. In an uninoculated blank study, 3.5 g L −1 of cellular contaminants was observed after 48 h, but no contaminants could be detected in the presence of H. bluephagenesis TD01. Subsequently, a genetically modified derivative of this strain (TD40, with the succinate semi-aldehyde dehydrogenase deleted) was cultivated in fed-batch mode under non-sterile conditions at the 5000-L scale [145]. In that study, [X t ] of nearly 100 g L −1 were achieved along with P(3-HB-co-4-HB) productivities of 1.66 g L −1 h −1 (see Table 4).
In a similar study, Yue et al. [172] were able to maintain relatively steady biomass and PHB concentrations, with no evident contamination using wild-type and recombinant H. campaniensis LS21 in a 65-day open process that was periodically fed a mixture of fats, proteins, and cellulose-based substrates designed to model kitchen waste. Lillo and Rodriguez [166] made use of the salt-tolerant properties of the haloarcula Hfx. mediterranei ATCC 33500 to maintain a continuous culture for three months under non-sterile conditions with stable PHB production. Other studies have employed methanotrophs [182] as well as thermophilic strains [171] for the production of PHA under non-sterile conditions.

Discussion and Conclusions
The widespread use of PHAs in biopolymer applications is currently prevented by their high production cost [34], the cost of downstream processing [29], as well as the questionable life cycle-based performance [32]. The upstream production costs can largely be broken down into the cost of the carbon source [39], the cost for mechanical agitation and the aeration of the bioreactor itself [34], as well as the capital and operating costs that are required to maintain a sterile environment [178]. Therefore, the development of efficient cultivation strategies is an important milestone toward producing PHAs in a cost-effective manner. In this review, we have attempted to summarize the bioreactor operational strategies that are used to target high-productivity PHA production using pure cultures. The paragraphs that follow are a summary of the findings.
Fed-batch strategies appear to be the most popular high-productivity cultivation method, and boast some of the highest reported Q v values for both mcl-PHA and scl-PHA production. Many earlier studies, particularly from the mid-1990s in scl-PHA production and perhaps 10-15 years later in mcl-PHA production, report some of the highest known Q v values to this day. However, recently, we have seen a shift away from those studies that simply target reporting the highest known Q v , perhaps because many of those studies were done under conditions that were only possible at the lab-scale and unrealistic for pilot-scale and industrial-scale bioreactors. These conditions include the use of highly refined carbon substrates and increasing the driving force for oxygen transfer through sparging with O 2 -enriched air.
A look at more recent studies shows a shift toward addressing these issues. Firstly, several recent studies have achieved reasonably good results using more realistic waste carbon substrates, including agricultural residues [98,108], waste glycerol [99,183], VFAs [61,62,96,101], and plant oils [94,102]. Secondly, many of these studies are doing so without use of O 2 -enriched air, which can significantly improve productivity, but also adds significant expense, and is therefore not generally recommended [51,139]. This is exemplified in this review; one study from over 20 years ago used a 50-L bioreactor with pure O 2 [53], but no such recent studies of scaled-up systems with working volumes of approximately 100 L or greater were found. This suggests that ongoing developments for pilot-scale and industrial-scale PHA production operations should rely on efficient cultivation strategies that are not dependent on pure O 2 usage. This is currently being addressed at the lab scale, and several promising high-productivity cultivation systems have been developed for both mcl-PHA and scl-PHA production that use only air [61,62,97,103,176] as well as in industrial-scale bioreactors [139,145]. Although other methods for increasing the driving force of oxygen transfer are available, including pressure [115], microbubble devices [118], and microfluidic devices [119], they have not yet been applied on a large scale.
Time is a significant factor in advancing PHA productivities. While the preferred mode of operation in PHA production certainly appears to be fed batch, continuous processes may have higher average productivity over prolonged operation due to the lack of downtime between batches. Continuous cultures of both scl-PHA and mcl-PHA have resulted in productivities in excess of 1 g L −1 h −1 , but little process development has occurred in this area within the last eight years, which is a problem that was also identified in a 2015 review of strategies for large-scale PHA production, and appears to remain unaddressed at present [45].
Cyclic/repeated processes also use time as a factor to advance productivity, although not as effectively as fully continuous bioreactors. There have been a number of promising developments in this area within the last 10 years. Within this context, cell retention techniques can be employed to overcome the volumetric limitation that comes from feeding waste carbon streams in which the substrates are often diluted. When these cell-retention techniques are employed, some studies have achieved productivities in excess of 3 g L −1 h −1 from waste substrates [168] or diluted carbon streams intended to represent waste streams [176]. While these processes have recently shown great promise, they were generally not operated long enough to take advantage of the time saved from the minimal downtime in a cyclic process. Finally, the capital and operating costs of sterilization are being considered, as is indicated by a recent review of non-sterile biotechnological processes [184]. This is particularly important for continuous and semi-continuous bioreactor systems, which are especially susceptible to contamination [50], but is also being considered in fed-batch cultivations by using halophiles and thermophiles [171,172,181].
In summary, with the exception of the lack of process development for full continuous bioreactor systems, the overall trends toward scaling-up, utilizing available inexpensive waste carbon streams, and non-sterile cultivations are promising. The recent work of Ye et al. [145] perhaps best represents the culmination of these technologies, and establishes what is, in our view, a new benchmark to give optimism to the industry: industrial-scale PHA production with high Q v (1.7 g L −1 h −1 ) from an available waste carbon source (corn steep liquor) under non-sterile conditions. These and other similar developments keep PHAs a relevant and worthwhile pursuit for novel biopolymer applications.