Both prokaryotic and eukaryotic microalgae are suitable candidates as producing platforms: their different features are critical for accumulation of different kinds of product. Cyanobacteria naturally accumulate bioplastic precursors (Table 1; Luengo et al., 2003) and are good hosts for small molecules not requiring eukaryotic post-translational modifications. Conversely, eukaryotic algae are valuable when such modifications (e.g., glycosylation) are essential, or when product compartmentalization is necessary. Eukaryotic algae are also a precious source of several secondary metabolites (Wijffels et al., 2013), and include some well-known oleaginous species, as P. tricornutum or N. gaditana.

The ability to change, remove or introduce gene(s) of interest in a programmable way, and to achieve a stable phenotype of the engineered strains over time, is pivotal to industrial application of photosynthetic bio-factories (Patel et al., 2019; Pérez et al., 2019). Molecular tools are better established for cyanobacteria than for eukaryotic algae, but nuclease-based genome editing tools like Clustered Regulatory Interspaced Short Palindromic Repeats (CRISPR)/Cas9 and Transcription Activator-Like Effector Nucleases (TALEN) are rapidly developing and applied to algal species of different phylogeny and commercial value (Doron et al., 2016; Nymark et al., 2016; Shin et al., 2016; Kroth et al., 2018; Verruto et al., 2018; Ortega-Escalante et al., 2019; Pérez et al., 2019; Ng et al., 2020). Yet, genetic manipulation of microalgae, particularly eukaryotic ones, still encounters several challenges. They are mostly related to the low efficiency of the transformation protocols and to the low and unstable expression levels of transgenes. Such weaknesses have been mainly ascribed to positional effects of gene integration and/or gene silencing. Secondary off-target mutations in edited strains, even when targeted genome editing has been performed with nucleases (Patel et al., 2019), are an undesirable aspect of algae transformation. To minimize issues due to off-target mutations, ribonucleoproteins (RNP-) based methods for CRISPR/Cas gene editing have been developed in substitution to vector-mediated protocols (Patel et al., 2019).

The selection of genetic backgrounds which minimizes gene silencing is a tool to achieve a stable expression of transgenes. Mutants of the model green alga C. reinhardtii that showed increased nuclear transgene expression, called UVM4 and UVM11, have been obtained with random UV mutagenesis (Neupert et al., 2009). Knowledge on molecular mechanisms underlying gene expression regulation shall be deepened to fully overcome drawbacks due to low heterologous gene expression, avoiding at the same time labor intensive screening of random mutants. The optimization of cis regulatory elements like promoters, ribosome binding sites (RBS), UTR sequences, and the optimization of heterologous sequence codon usage according to the selected cell chassis, have also been recognized as a key feature for successful transgene expression (Rasala et al., 2011; Jinkerson and Jonikas, 2015; Till et al., 2020); specifically in eukaryotic cells, mimicking a native nuclear gene sequence as accurately as possible, by interrupting the transgene coding sequence with introns, shall also contribute to a stable transgene expression (Baier et al., 2018). Finally, efforts have been made to identify and characterize the best genomic environment where to introduce transgenes, the so-called neutral sites, loci in which gene integration is not compromising other cell functions (Ng et al., 2015; Pinto et al., 2015; Kroth et al., 2018).

Once productive genome edited microalgae are obtained and ready to be exploited in large scale industrial plants, the biosafety issue must be timely addressed by specific agencies (Spicer and Molnar, 2018): both potential harm to human and animal health and potential risk for the environment (i.e., vertical or horizontal gene transfer, strain competition with wild-type strain) are to be assessed (Henley et al., 2013). Although genetically modified algae are predicted not to have fitness advantages in nature, safety measures to avoid the spill over of modified strains must be regulated and applied especially in outdoor cultivation systems (Wijffels et al., 2013; Henley et al., 2013; Spicer and Molnar, 2018).

All biological processes described in section 2, from light harvesting to metabolic energy consumption, are characterized by substantial energy losses (Figure 2 and Table 5) and are potential targets for improving energy use efficiency.

Adjusting the efficiency of photosynthetic light reactions will rise the amount of energy available for cell metabolism. A key strategy to improve light conversion into chemical energy deals with modulation of light harvesting and photoprotection. They are both very dynamic processes which readily respond to changes in light intensity, and account for substantial energy dissipation when light irradiation exceeds the photosynthesis saturation point (Figure 2 and Table 5; Peltier et al., 2010; Wilhelm and Selmar, 2011; Peers, 2014).

The reduction of light harvesting capacity is a valuable strategy, which has already been tested in strains with genetically Truncated Light harvesting Antennae (TLA). TLA approach reduces the light harvesting cross-section in photosynthetic cells and potentially triplicates the productivity of plants and algae (Melis, 2009; Kirst et al., 2017). In a dense microalgal culture, TLA phenotype diminishes the energy harvested by the external layer of cells which are directly exposed to intense illumination. Therefore, their need for energy dissipation and risk of photoinhibition are reduced, while their photosynthetic efficiency enhanced. TLA phenotype also favor the diffusion and homogeneity of light into the culture. This increases light availability for the inner layer of cells that instead, in the case of WT cultures, suffer light limitation and reduced growth. TLA benefits were demonstrated in different species, where TLA cultures showed up to 1.5 times the growth and productivity of WT (Nakajima et al., 2001; Polle et al., 2003; Mussgnug et al., 2007; Perrine et al., 2012; Kirst et al., 2014; Shin et al., 2017). However, TLA advantages have been proved under specific growth conditions (i.e., high light) which accentuate TLA positive traits. Conversely, when limiting light was supplied, TLA growth was reduced compared to the WT one (Nakajima et al., 2001; Perrine et al., 2012; Shin et al., 2017). In addition, some TLA mutants showed null advantages in their productivity, irrespective of the condition tested (Page et al., 2012; Nymark et al., 2019). Diverging phenotypes among TLA mutants have been ascribed to undesirable side effects of the mutation or to an unbalanced photoprotection caused by the altered antenna system (de Mooij et al., 2015; Nymark et al., 2019). In TLA Synechocystis cells, the reduced size of phycobilisomes was also shown to have a wide impact on cell proteome (Liberton et al., 2017). Remarkably, despite TLA approach still faces challenges to be solved in future studies, TLA approach has a further central advantage on resource use efficiency: as it reduces cellular N quota invested in light harvesting proteins, more energy can be allocated to other metabolic pathways (Figures 4, 5).

A further approach to optimize light harvesting capacity deals with modulation of photoprotection. Photoprotection results from an intricate network of processes regulating photosynthetic electron transport. It is essential to cope with variable and excess light: the light reactions promptly respond to any burst of higher illumination, increasing the amount of energy captured by the antenna system and electron transport. Conversely, the metabolic reactions are not equally fast to respond to increased energy availability. This creates a dangerous imbalance between production and utilization of ATP and reductants, leading to over-reduction of electron transport chain and photoinhibition. Photoprotection mechanisms thus provide the cell safe valves to dissipate these bursts of excess energy. If fully abolished, the cells will be prone to reactive oxygen species (ROS) formation, photoinhibition, and eventually cell damage when light is absorbed in excess (Wilhelm and Selmar, 2011; Niyogi and Truong, 2013; Gerotto et al., 2016). Thus, a valuable option to optimize photoprotection processes targets their relaxation kinetics. It is estimated that the sustained energy quenching upon a switch from high to low light accounts for about 20% reduction in crop yield (Taylor and Long, 2017): their activation/deactivation kinetic in response to light intensity fluctuations is a possible target of improvement. Validation of this hypothesis has been obtained in tobacco, where the simultaneous over-expression of three key proteins regulating heat dissipation of absorbed excess energy (i.e., PSBS; VDE; ZE) resulted in 9% increased C fixation rates and an average of 15% increase plant dry weight with respect to WT. This was due to a faster deactivation of the heat dissipation mechanisms when plants were moved from strong to limiting light (Kromdijk et al., 2016). A similar approach can be transferred to algae expecting analogous potential benefits. For example, in PBRs, culture mixing leads microalgae to experience a fluctuating light regime: cells are alternatively exposed to strong illumination in the external layer or to limiting light in the internal layer. Thus, the activation/deactivation timescale of photoprotection mechanisms shall fit with the PBR mixing kinetic, in order to protect cells from bursts of high light when cells are in the external layer, while relaxing fast when cells are back to the inner section of the PBR (Sforza et al., 2012; Simionato et al., 2013).

Conversely to light harvesting system and photoprotection processes, photosystem RCs are highly conserved among oxygenic photosynthetic organisms and they are already extremely efficient machines (0.8 and 1 electron/photon for PSII and PSI, respectively) (Caffarri et al., 2014; Raven et al., 2014; Romero et al., 2017). Thus, instead of modifying PSI and PSII, synthetic biology shall design novel architectures of the electron transport chain, in which protein complexes from multiple natural sources are merged together to improve photosynthetic performance (Ort et al., 2015; Pérez et al., 2019). For example, the introduction of bacterial type reaction centers holding bacteriochlorophyll pigments will allow PAR to be expanded up to 1100 nm. It requires a complete redesign of the electron transport chain, thus relying on extensive application of synthetic biology. The modified oxygenic photosynthesis is suggested to potentially double the efficiency of its electron transport (Ort et al., 2015).

Once light energy is temporary stored as chemical energy, ATP and NADPH are used to fix C. Rubisco enzyme causes a considerable loss in energy conversion due to its competitive oxygenase activity: when 25% of the reactions catalyzed by Rubisco use molecular oxygen as substrates rather than CO₂, the energy requirement for mol of C fixed increases by 60% compared to the sole carboxylation activity (Table 5). So far attempts to engineer the enzymatic properties of Rubisco by enhancing carboxylation kinetic while limiting oxygenation reactions have failed, despite a massive number of studies focused on the characterization of Rubisco crystal structures in different cyanobacterial, algal and plant species and a detailed knowledge on the catalytic reaction mechanisms both for carboxylation and oxygenation (Taylor et al., 2001; Andersson, 2008; Tcherkez, 2013, 2016). Thus, if Rubisco has already been “optimized” by the Nature during photosynthetic life evolution (Tcherkez et al., 2006; Savir et al., 2010; Kerfeld, 2015), it will possibly remain a major source of energy loss. What can research still pursue? Following a synthetic biology approach, the “best” available natural option of Rubisco (i.e., showing the highest K_cat^c, as in cyanobacteria, or the best τ, as in red algae, Table 2) may be combined with the most energy-saving and efficient CCM set up (as in some green algae and diatoms) in the cell chassis. Another suggested solution to improve Rubisco catalytic efficiency, particularly under fluctuating irradiation, deals with engineering Rubisco activase, a Rubisco regulatory protein which facilitates the release of inhibitors from Rubisco active site in an ATP-dependent manner (Carmo-Silva et al., 2015). Moreover, being Rubisco a relatively big and highly expressed enzyme (Ellis, 1979), the removal of high-N content amino acids such as arginine from its sequence may represent a way to increase N and energy use efficiency (Figure 4).

RuBP regeneration is also recognized as a main bottleneck reducing CBB cycle efficiency (Falkowski and Raven, 2007). A path to improve C fixation thus relies on optimization of CBB enzymes’ stoichiometry to ease the regeneration of RuBP. Mutants harboring modified amounts of Rubisco, sedoheptulose-1,7-bisphosphatase, fructose-1,6-bisphosphatase (the latter two as a bifunctional enzyme in cyanobacteria), fructose 1,6-bisphosphate aldolase or transketolase, have been obtained both in eukaryotic algae and cyanobacteria (Fang et al., 2012; Ogawa et al., 2015; Liang and Lindblad, 2016; Yang et al., 2017; De Porcellinis et al., 2018). As described above for TLA mutants, manipulated abundance of CBB enzymes led to variable results, according to the species and to the specific enzyme under investigation. Yet, these mutants often displayed increased photosynthesis and growth. For example, overexpression of sedoheptulose-1,7-bisphosphatase or sedoheptusole-1,7/fructose-1,6-bisphosphatase increased glycerol accumulation in Dunaliella bardawil and wax esters in Euglena gracilis (Fang et al., 2012; Ogawa et al., 2015). In Synechococcus, the overexpression of the bifunctional enzyme led to a large adjustment of C metabolism, enhancing photosynthetic C fixation while decreasing respiration (De Porcellinis et al., 2018). These findings prove that modifying the expression of specific CBB enzymes shall contribute to an improved C fixation and cell productivity.

Another attracting option to enhance C fixation efficiency relies on engineering or replacing the photorespiratory pathway, in order to metabolize glycolate produced by Rubisco with lower energy cost and resource consumption. A first functional synthetic metabolic bypass to photorespiration has been successfully introduced in Synechococcus elongatus PCC7942. Six heterologous genes of the C fixing 3-hydroxypropionate pathway from Chloroflexus aurantiacus, a thermophilic anoxygenic phototroph, were expressed in S. elongatus (Shih et al., 2014). Next steps on the same path rely on further engineering of cell metabolism to allocate the metabolic intermediates generated by the bypass pathway to the production of industrially relevant molecules (Shih et al., 2014).

More ambitious routes would be to import oxygen-insensitive pathways for the key reaction of C fixation to bypass Rubisco altogether. The possibility to introduce (at least in addition to carboxylation) CO₂ reductive pathways into photosynthetic cells has also been considered. Verifying the actual feasibility and potential of these synthetic pathways in vivo is a major task for the future (Bar-Even et al., 2010; Bar-Even, 2018; Cotton et al., 2018).

For some specific purposes, a fully opposite option can be envisioned, which takes advantage of Rubisco oxygenation side reaction instead of trying to limit it. Under specific growth conditions, glycolate is a main sink of fixed C and a very high amount of glycolate is produced and secreted by C. reinhardtii cells. Glycolate can be recovered from the culture media and used as a substrate for other biotechnological applications, as methane production (Taubert et al., 2019; Table 1).

In parallel with maximizing biomass productivity, several research groups worldwide are constructing genetically modified strains with the goal of improving the yield of natural products or introducing the ability to synthesize non-native molecules (examples in Table 6; Case and Atsumi, 2016; Sun et al., 2019; Wichmann et al., 2020). Multiple strategies are under investigation to enhance the accumulation of biofuel feedstocks like alcohols or TAG. They range from overexpression of native or heterologous enzymes of the fatty acid biosynthetic pathway, like diacylglycerol acyltransferase (DGAT) enzyme; increased accumulation of metabolic intermediates used for lipids synthesis, like boosting acetyl-CoA supply; disruption of competing pathways; up to modulation of transcription factors accumulation (Sun et al., 2019; Figure 3 and Table 6).

Terpenoids, also known as terpenes or isoprenoids, are another valuable class of hydrocarbon-derived molecules for industrial purposes. They are naturally produced by photosynthetic cells in the form of pigments, plant hormones and species-specific secondary metabolites with defensive role or mediating interactions with the environment (Figure 3; Pichersky and Raguso, 2018). They include thousands of chemicals with a wide range of human applications, from fuels, to pharmaceuticals (e.g., artemisinin has anti-malarial properties), to food and cosmetic additives, as menthol, limonene and squalene (Davies et al., 2015; Vavitsas et al., 2018; Wichmann et al., 2020). However, terpenoids are usually accumulated in low amount, harnessing an economically viable extraction from their natural producer. Currently, some terpenoids are produced in heterotrophic hosts like E. coli and S. cerevisiae, which require multiple genetic modification to establish the biosynthetic pathway. Conversely, microalgae already hold core terpenoid biosynthetic pathways for production of primary molecules like pigments, being a valuable host for heterologous production of non-native plant terpenes (Vavitsas et al., 2018; Wichmann et al., 2020; Table 6 lists some examples).

In addition to hydrocarbons, microalgae, particularly green algae, have been used as valuable hosts for heterologous protein productions. Proteins expressed in C. reinhardtii, but also Dunaliella salina, D. tertiolecta and Chlorella ellipsoidea include antibodies, immonotoxins, enzymes and subunits of vaccines (for a review see Rasala and Mayfield, 2015), further demonstrating the potential of microalgae as an important source of fuels, food, feed and pharmaceutical products.

The above-mentioned examples and findings on the genetic engineering of single processes in photosynthetic cells show a few folds impact on algal productivity. According to the estimations on the energy conversion efficiency and CO₂ sequestration reported in Tables 3, 5, this increase is still not enough to reach a major productivity breakthrough in multiple applications. Further, many studies highlight important side effects on other cell functions. This can be due to secondary effects of the mutations, but also to unpredicted interactions among physiological processes, leading e.g., to C allocation into different ready-to-use or storage macromolecular pools, luxury uptake of nutrients, functional and compositional homeostasis (see review Giordano, 2013 for details on homeostasis; Giordano et al., 2015a; Finkel et al., 2016; Ruan et al., 2018; Giordano and Wang, 2018). Our lacunose knowledge on cell strategies regarding homeostasis, acclimation and adaptation responses, even in the case of extensively studied species (e.g., diatoms), is still a major issue in the manipulation of photosynthetic cells (Wagner et al., 2017).

It is thus clear that any attempt to improve CO₂ fixation and biomass quality should be based on the holistic view of cell physiology, by simultaneously adapting (i.e., balancing) the activity and the regulation of multiple cellular functions.

First, chemical energy (ATP) and reductants produced by photosynthesis light reaction must be carefully tied with the energy cost of C fixation and other metabolic pathways (Figures 1, 4) (Cotton et al., 2015; Giordano and Wang, 2018). Remarkably, the source/sink balance is not the only way in which photosynthetic light reactions and cell metabolism are interdependent. The redox poise in chloroplast stroma generated by the electron transport, in the form of reduced ferredoxin or NADPH, is also a crucial regulatory signal for several metabolic pathways through redox regulation of multiple enzymes by the thioredoxin system (Kikutani et al., 2012; Morisse et al., 2014; Nikkanen and Rintamäki, 2014).

Further, any change driven by metabolic engineering design must also be carefully customized to the selected genetic background and the specific growth environment cells will experience in large scale production, to avoid downstream issues during the scale up of the process, which is currently among the biggest obstacles to economically viable bio-factories (Borowitzka, 2016; Newby et al., 2016).

Even though some common features of bio-factory strains can be outlined, like the ability to thrive in extreme environment to limit contaminations, a fast growth rate, a high cell density, a high biomass productivity, it is conceivable that multiple ad hoc cell chassis should be designed according to each specific purpose, to minimize all the bottlenecks of each specific production system.

Rapidly developing -omic sciences, i.e., transcriptomic, proteomic and metabolomic, shall support a comprehensive understanding of cell physiology and of the consequences of each targeted mutation on cell functions. Such a holistic view of cell functioning shall then drive the selection of additional processes which require metabolic energy optimization or which can be minimized to save metabolic energy (Figure 5). Performing direct validation of the productivity and the drawbacks of different metabolic variants with experimental studies on living cells is a labor-intensive and time-consuming procedure. Mathematical models which mimic the consequences of a mutation in silico are thus extremely relevant to develop a successful, wide-ranging, metabolic engineering approach to design photosynthetic cells as a function of the selected species and growth environments.

Biochemical simulations for photosynthetic C assimilation started in the 80s (Farquhar et al., 1980, 2001), then being constantly improved and expanded. Available models now include reactions from multiple pathways, like CBB cycle, photorespiration, starch and sucrose synthesis, as a function of nutrient availability (Zhu et al., 2007, 2013). Given the high metabolic cost of N assimilation into amino acids (Figure 4), a further advantage of computational modeling is, for example, that a constant N quota can be allocated to the whole set of enzymes considered. This will avoid additional energy investment of the cell in N acquisition, while optimizing enzymes relative stoichiometry (Zhu et al., 2007). Similar computational predictions are recently under development also for economically relevant algal species. They simulate the complex network of photosynthetic light reactions and their regulation, as well as algal metabolic reactions in different environmental conditions, like illumination regime, nutrient and CO₂ availability (Chang et al., 2011; Du et al., 2018; Perin et al., 2019; Fachet et al., 2020; Toyoshima et al., 2020). Remarkably, robust mathematical models rely on a deep understanding of the cell physiology, to include as many parameters as possible in the simulation. This multidisciplinary approach shall allow to design living factory where all the house-keeping functions are present but minimized and most of the energy is directed to the commercial products (Table 5 and Figure 5).