The overall goal of the studies on the biochemistry and molecular biology of lipid synthesis in microalgae was to increase the understanding of the lipid biosynthetic pathways and to identify enzymes that influence the rate of lipid accumulation and lipid quality. This information would be used to genetically manipulate the biosynthetic pathways for improvement in lipid production rates and to manipulate the nature of the lipids produced (i. e., the degree of fatty acid saturation and chain length) to optimize the production of biodiesel.

The development of a genetic transformation system for diatoms allowed NREL researchers to begin testing ways to manipulate microalgal biochemical pathways. The first target enzyme was ACCase. Previous studies at NREL had shown that increased lipid production in diatoms induced by Si starvation was accompanied by an increase in the activity of the ACCase enzyme. Therefore, it was logical to ask whether the activity of the enzyme could be increased in the cells by adding additional copies of the gene encoding ACCase (acc1), and, if so, would increased activity of the protein stimulate the production of lipids in the algal cells?

A full-length copy of the C. cryptica acc1 gene had been cloned and characterized at NREL (see Section II. B.2.f). The plasmid containing this sequence was designated pACC1 (Figure II. B.8). Before the algal transformation system, attempts were made to express the algal gene in a bacterial system to ensure that the cloned gene encoded a functional ACCase enzyme and to test for the effects of overexpression. For expression of the C. cryptica acc1 gene in E. coli, the introns were removed and the 5′ terminus was replaced with the 5′ end of the E. coli P — galactosidase gene, which included the inducible promoter region. This fusion gene was introduced into E. coli. The transformed cells were analyzed for the production of algal ACCase protein by probing blots of (Sodium Dodecyl Sulfate, SDS) polyacrylamide gels with an anti — ACCase antibody or with avidin conjugated to alkaline phosphatase. (Avidin binds to the biotin moity in the functional ACCase protein.) The bacterial cells produced full-length algal ACCase, as well as a large number of shorter polypeptides recognized by the anti-ACCase antibody. Introducting the gene into other E. coli strains deficient in protease activity also produced these shortened peptides; therefore, they were presumed to be the result of truncated transcription or translation. The full-length ACCase protein was properly biotinylated in the transformed bacteria, but not as efficiently as in the E. coli native biotin-binding ACCase subunit. No effects were observed on lipid biosynthesis in the transformed E. coli strain. Attempts were also made to introduce the C. cryptica acc1 gene into yeast, as expression in a eukaryotic system would more likely mimic the effects in algae, but these experiments were unsuccessful.

The next step was to introduce additional copies of the acc1 gene in diatoms, with the goal of increasing the activity of the ACCase enzyme and then assaying the effects of ACCase overexpression on lipid accumulation. The plasmid containing the full-length acc1 gene (pACC1) does not contain a selectable marker for transformation. Studies in other laboratories showed that nonselectable plasmids can be introduced into cells via cotransformation with a plasmid containing a selectable marker gene such as nptII. Although the exact mechanism for

this phenomenon is not clear, it is believed that during a given transformation procedure, a particular subpopulation of the cells becomes “transformation competent”. These cells may then take up multiple copies of DNA molecules present in the reaction. Introduction of pACC 1 into the diatoms was mediated by microprojectile bombardment as described in a previous section, but with pACCNPT5.1 and pACC 1 precipitated onto the tungsten beads in equimolar amounts. Transformed cells were selected based on their induced resistance to G418 and then screened for additional copies of the acc1 wild-type gene using PCR and Southern analysis. Between 20% and 80% of the G418-resistant colonies contained acc1 sequences in the cotransformation experiments. To facilitate the selection of transformants containing extra copies of the acc1 gene, a plasmid was also constructed that contained both acc1 and nptII, designated pACCNPT4; transformants generated using pACCNPT4 and selected for G418-resistance almost always contained the acc1 gene as well.

Transformed cells containing additional C. cryptica acc1 gene sequences were isolated in C. cryptica T13L, C. cryptica CYCLO1, and N. saprophila NAVIC1. Southern analysis indicated that the foreign DNA inserted into host genome, often in one or more random sites, and often in the form of tandem repeats. Several strains that contained one or more full-length sequences of the inserted acc1 gene were analyzed further to test for ACCase overexpression. The CYCLO T13L transformants showed two to three fold higher ACCase activity than wild-type cells, and there was a corresponding increase in acc1 gene transcript (mRNA) levels. However, preliminary analyses of the lipid composition of the cultures overexpressing acc1 did not indicate a detectable increase in lipid levels. These results suggest that the lipid biosynthesis pathways may be subject to feedback inhibition, so that increased activity of the ACCase enzyme is compensated for by other pathways within the cells. It was hoped that expression of C. cryptica T13L acc1 gene in other algal strains might overcome this inhibition. Numerous N. saprophila transformants were generated that contained full-length copies of the C. cryptica acc1 gene; although acc1 mRNA was detected using the RPA, the recombinant ACCase protein was not detected in any of the N. saprophila strains tested. Whether this result was due to inefficient translation of the mRNA, or degradation of the foreign protein due to improper biotinylation or targeting, is not known. Transformants were also generated in a second strain of C. cryptica, CYCLO1, but the program was discontinued before these strains could be analyzed fully.

NREL researcher Eric Jarvis took another approach to genetically manipulating algal pathways for increased lipid production. Previous research had resulted in the cloning and characterization of the uppl gene from C. cryptica (described in Section II. B.2.h.). This gene codes for a fusion protein containing the activities for UDPglucose pyrophosphorylase and phosphoglucomutase, two key enzymes in the production of chrysolaminarin. It was postulated that decreasing expression of the uppl gene could result in a decrease in the proportion of newly assimilated carbon into the carbohydrate synthesis pathways, and consequently increase the flow of carbon to lipids.

Two techniques that are becoming widely used for gene inactivation are ribozymes and antisense RNA. Dr. Jarvis spent 6 months working at Ribozyme Pharmaceuticals, Inc., a biotechnology company in Boulder, Colorado, learning about these new methods. Antisense RNA is a method

in which a cell is transformed with a synthetic gene that produces an RNA sequence complimentary to a specific mRNA. Although the exact mechanism is not clear, the antisense RNA prevents translation from its complimentary mRNA, effectively lowering the level of that particular protein in the cell. Ribozymes are also RNA molecules produced by synthetic genes that can bind to, and cleave, very specific RNA sequences. Ribozymes can be designed to degrade specific mRNA molecules, effectively decreasing expression of a specific gene.

Several ribozymes sequences designed to cleave uppl RNA were constructed based on computer predictions of the secondary structure of the target RNA. The ribozyme constructs were shown to specifically cleave the target RNA in vitro. The ribozyme sequences were then inserted into the pACCNPT10 vector in the untranslated acc1 sequence between the nptII stop codon and the acc1 termination sequence (see Figure II. B.8). C. cryptica T13L was transformed with these vectors as described earlier and transformants were selected based on acquired resistance to G418. Extracts were made of the transformed strains and analyzed for UGPase activity. Unfortunately, insertion of the ribozyme sequences did not result in detectable decreases in UGPase expression. Although these initial experiments were unsuccessful, gene inactivation technologies acquired during this project seemed a promising approach for manipulation of algal lipid synthesis pathways. At the time project funding was terminated, work was in progress to continue with the ribozyme experiments and to test antisense RNA constructs as an additional method for inactivating algal pathways.

A key underlying assumption in this work was that monoculture productivity is an indicator of competitiveness in mixed cultures. That is, the most productive culture will also be the most competitive. Another assumption is that productivity and competitiveness will be affected by fluctuations in environmental variables and cannot be predicted from single parameter variations or steady-state operations. Testing these hypotheses was a central objective of these experiments, which were carried out with continuous and then semicontinuous (once a day dilution) laboratory cultures (1-L bottles).

The continuous cultures were used to test productivity of several strains under simulated outdoor conditions (approximately sinusoidal diurnal temperature changes from 15°-32°C overnight, a 14-h constant light period, 0.6-0.7 d-1 dilution, and 6-h gassing with pure O2 rather than air). As in earlier studies, Chlorella and Monoraphidium exhibited higher productivities (250-300 mg/L/d) than the other strains tested (Porphyridium, Ankistrodesmus, and Chaetaceros, with some 150-200 mg/L/d productivity). Under constant conditions, similar ranking and differences were observed between the three fastest-growing strains. Mixed culture experiments between Chlorella and Chaetoceros were then carried out, with the results comparing fairly well with the APM results (Section III. B.6.b.). Figure III. B. 12. shows experiments in which a 90/10 Chaetaceros/Chlorella mixed inoculum grew in continuous cultures, with cell species changing faster experimentally than in the model, which, however, predicts the overall trend. Overall results with these strains were in accord with prior experience (Weissman and Goebel 1985, 1988).

However, the continuous cultures proved logistically and experimentally very difficult, and the experimental design was switched to a simpler, once a day (before dawn), dilution. Fluctuating conditions were achieved primarily by timers and actuators (e. g., for O2 versus air supply, for temperature control in the water baths). This allowed greater reproducibility and simultaneous operation of many more cultures, as many as 16 in four water baths, allowing collection of an extensive data set. The effects of constant versus fluctuating temperature, pO2, and light, on culture productivities and dominance in mixed cultures were studied. In summary, the overall result was that fluctuating temperature and other conditions had major effects and were different from those at constant temperature. Also, growth responses for unialgal cultures can predict the outcome of species competition experiments. Still, much more experimental work is required on these problems.

Figure III. B.12. Species competition between Chaetoceros and Chlorella in continuous cultures.

Mixed culture results for a 90/10 cell density mixture of Chaetoceros and Chlorella.

a. ) Top: Culture productivity.

b. ) Bottom: Species cell concentration (solid line represents model predictions).

(Source: Benemann and Tillett 1990.)

III. B.1. Introduction

The long-term objective of the ASP was to develop microalgae liquid fuel production processes. Since its inception, the ASP supported laboratory R&D projects (Section II) and algal mass culture projects. However, for the most part, the laboratory and outdoor projects were not integrated into a strongly unified program. This reflects in large part the difficulty of such integration. Also, during the early stages of the ASP, too close an integration would have been restrictive, as it was not yet clear at the time which research approaches, production systems or algal strains would be best.

The extensive work on strain isolation, selection, characterization, etc., carried out by the ASP was used to a significant extent by the field projects, through the testing of a number of the isolates in algal mass cultures, specifically in the projects reviewed in this section. Unfortunately, the laboratory-level screening protocols had, in hindsight, relatively little predictive power for the ability of the strains to dominate and perform in outdoor ponds. Similarly, the laboratory work on the biochemistry, genetics and physiology of lipid biosynthesis, was difficult to apply to the goal of increasing lipid productivities in outdoor systems. Greater integration of laboratory and outdoor R&D is a challenge for any future microalgae R&D program.

The ASP initiated two outdoor projects in 1980, one in California using a paddle wheel-mixed raceway pond design (“high rate pond,” [HRP]), and another in Hawaii. The Hawaii project was to demonstrate a patented algal culture system, invented by then-ASP program manager, Dr. Larry Raymond (1981). This “Algal Raceway Production System” (ARPS) used very shallow flumes (<10 cm), rapid mixing by air lifts, covers with CuSO4 filters to screen out harmful infrared radiation, and harvesting of the biomass by foam fractionation, among many other claimed attributes (Figure III. B. 1.). Very high productivities were claimed. But a review of the work (Raymond 1979), in which P. tricornutum was grown in a 0.5-m2 system, revealed that this projection derived from a single batch culture, and in fact, the last data point showed biomass density actually decreasing. Benemann et al. (1982a, b, see Section III. D.5.), carried out a comparative analysis of the ARPS and the HRP designs, concluding that the ARPS would be too expensive and energy intensive, compared to the HRP design. These two projects in California and Hawaii, carried out for more than 6 years each, are reviewed in this section, followed by descriptions of the ASP projects in Israel, New Mexico, and related projects.

After the ASP was established at SERI in the late 1970s, the emphasis switched from methane production to algal oils as the fuel product. This was based on the known ability of some microalgae species to accumulate large amounts of algal lipids, in particular under conditions of nutrient (mainly N and Si) limitations (See Section II and Section III. B.5.d.).

As discussed in Section III. A.1, initially the ASP set out to investigate both the HRP design described earlier and a patented, closed photobioreactor concept, the ARPS. In 1981, the DOE Office of Energy Research requested an in-depth engineering and cost analysis of both systems. However, by the time the final report was completed in 1982, the ARPS had already started to evolve toward a more standard design (Section II. B.2.), and the HRP project in California (Section III. B.3.) was re-instated. Thus, the comparative evaluation had become somewhat moot, and the final report (Benemann et al. 1982a) covered only the HRP system in detail, with the comparative HRP-ARPS analysis relegated to an unpublished appendix (Benemann et al. 1982b).

The HRP system followed quite closely on the earlier work of Benemann et al. (1978), but with some significant differences and much greater details for the engineering designs and cost estimates. As before, 40-ha earthwork ponds were used; however, this time with paddle wheel mixing. Productivities were now projected of 67.5 mt/ha/yr for an algal biomass containing 40% lipids (oils) by weight. This corresponded to about 90 mt/ha/yr for conventional algal biomass, yielding almost 160 barrels of crude oil/ha/yr. This was roughly twice as high as the prior study. Harvesting was again assumed to be by bioflocculation, followed by a centrifugation process to concentrate the biomass to a paste-like consistency. A solvent extraction process as used for soybean oil extraction was assumed, at three-time higher unit cost to account for the high — moisture in the paste. (However, as was pointed out, direct solvent extraction was unlikely to be feasible for such high moisture biomass.) As before, the residual biomass was to be anaerobically digested in covered ponds to produce methane gas, with the nutrients (and C) from the digester (and digester gas) recycled to the ponds.

A major emphasis in this report was the development of engineering designs for the CO2 supply and transfer systems, a major point of criticism in the Dynatech R/D Co. (1978a) study. In fact, Mr. Don Augenstein, the author of the Dynatech R/D Co study, joined EnBio, Inc. and carried out the engineering designs and calculations for CO2 supply systems for the present report. A

key assumption was that the CO2 flue gas delivery pipe from the power plant was only 5 km long, at a cost of almost $3 million (1982$). The distribution piping system, including blowers and valves, was estimated at about $2 million for the 800-ha plant. A detailed analysis of power requirements and CO2 transfer issues was also carried out.

Several cases and scenarios were analyzed, including operations at high (8.0-8.5) pH and low (7.0-7.5) pH, CO2 recycling from the methane produced, and the use of flue-gas and pure CO2. Flue gas CO2 required much higher capital and operating costs; pure CO2 required purchase of this nutrient. Various scenarios were analyzed for CO2 sources, including purchase and combustion of coal (e. g., co-siting a power plant). Another major variable analyzed was the capital and operating costs factors, such as labor costs and overhead, utilities and fuel costs, land costs, factors for buildings and power supplies, contingencies and contracting, architect and engineering fees, and capital-related cost factors (taxes, insurances, depreciations, maintenance, and returns on investment). These factors made a larger difference in final costs than most of the engineering design cost estimates (e. g., pond construction costs, paddle wheels). Somewhat surprisingly, there appears to be no standardization for such general cost factors. For the “base case” analysis, assuming no recyling of nutrients from the digester and a low pH of operation, costs were nearly identical, at almost $ 160/barrel oil, for pure CO2 and flue gas utilization cases. The higher cost of pure CO2 (@$45/mt) balanced by the higher capital and operating costs of the flue gas delivery system.

The cost of oil at $160/barrel was excessive, even for the projected rising fossil fuel prices. By recycling nutrients and operating at a higher pH (reducing CO2 outgassing) and using lower cost ($22/mt) CO2, overall costs could be reduced by about 20%-25%, to about $115 for the pure CO2 case, somewhat higher for the flue gas (coal) case (Table III. D.2.). This was a significant reduction, but still represented an excessive cost. This led to a reiteration of the entire engineering analysis, by using more optimistic engineering designs and estimates at each step, including, for example, a shorter flue gas delivery pipe. This resulted in a further capital costs reduction of about 25%. The capital cost related factors were also reduced, including power supply costs, building costs (to near $1,000/ha), and return on total capital (from 15% to 10% per annum). With these most optimistic assumptions, final costs were reduced to as low as $65/barrel of oil for a high pH flue gas case. Table III. D.2. summarizes these costs, for both for the conservative (base) case and optimistic cases. Updating these costs to roughly 1997 dollars would give, even under the optimistic case, costs for the oil of about $100/barrel.

Other alternatives were also examined. Seawater systems were attractive if all the CO2 required for algal growth were to be supplied by the seawater. For the assumed productivities, this would require a hydraulic retention time of 1 day, and thus much larger settling ponds and settling velocity for the algae, but could be cost-effective.

Another conclusion was that higher value byproducts were unlikely to significantly contribute to such systems, as their production (either in scale or objectives) would not be easily integrated with algal fuel production. The costs of methane and alcohol fuels from algal biomass (high in

carbohydrates, rather than lipids) would likely be similar to that of algal oil production. Indeed the issues were not the final processing, but the primary production of the biomass.

The authors identified four major research needs to achieve the objectives of high productivity in large-scale outdoor systems:

1. Photosynthetic efficiency for light energy and high lipid production.

2. Fundamentals of species selection and control in open pond systems.

3. Mechanisms (and control) of algal bioflocculation.

4. Effects of non-steady-state operating conditions on algal metabolism.

The appendix to this report (Benemann et al. 1982b), analyzed the ARPS system as proposed by Raymond (1979, 1981) that was in development at the time at the University of Hawaii (Section III. B. 1.). First, a detailed historical review of microalgae systems designs was presented, which traced the evolution of the two concepts. The main report carried out a detailed and updated review of all prior cost analyses. The specific claims made for the ARPS systems were analyzed in detail. For example, the CuSO4-filled cover was claimed to reduce harmful IR radiation, but this was not supported by the photosynthesis literature. Also, overheating would still be a major factor even with a CuSO4 cover, requiring a cooling process. In addition, the heated CuSO4 could not be plausibly used as a power source. Mixing power inputs would be prohibitive for this design. Increased productivities caused by a flashing light effect were not plausible. Most important, the costs for even the cover and liner for such a system would be prohibitive by themselves, without considering any other factors.

This study clearly identified the major difficulties associated with microalgal mass culture for fuel production. Only a very low-cost system, based on open ponds without plastic liners, mixed at low velocities, and using a very simple harvesting process, could be considered in such a process. But even with these rather favorable, though plausible, assumptions, costs would still be well above those for current, or projected, oil prices.

I Publications:

Benemann, J. R.; Augenstein, D. C.; Weissman, J. C. (1982a) “Microalgae as a source of liquid fuels, appendix: technical feasibility analysis.” Final Report, U. S. Department of Energy, unpublished, 126 pp.

Benemann, J. R.; Goebel, R. P.; Weissman, J. C.; Augenstein, D. C. (1982b) “Microalgae as a source of liquid fuels.” Final Report, U. S. Department of Energy, 202 pp.

Raymond, L. (1979) “Initial investigations of a shallow layer algal production system.”Am. Soc. Mech. Eng., New York.

Table III. D.2. Costs of microalgal biomass production.

Productivity: 67.5 mt/ha/yr for 40% extractable lipid biomass (162 bbl oil/ha/yr).

System Description: Twenty 40-ha growth ponds, harvesting by settling, with C recycle from the digesters and use of either pure CO2 delivered to the site for $22/mt, or generation of CO2 on site from coal at half this cost.

Cost Estimates: Figures in parenthesis next to capital and operating cost items refer to the factors used for the conservative and optimistic cases, respectively. Maintenance, insurance, taxes, (6% and 4.8%, respectively), based on total capital costs, except for land and working capital, which are also not depreciated. ROI (return on total capital) based on total capital, before taxes.

(Source: Benemann et al. 1982a.)

CAPITAL COSTS — $/ha

Water and Nutrient Supply

Earthworks & Berms

Paddlewheels

Settling Ponds

Centrifuges

Oil Extraction

CO2 Supply

Electricity

Bldgs., Offsites (102,7.5Z)* A&E and Contractors (202,102)* Contingencies (102 for both)* Land Costs

Working Capital (4, 3 months)* TOTAL CAPITAL COSTS

This project was a continuation of the project described in III. B.3.b. It was performed by a new company, Microbial Products, Inc. (EnBio was dissolved when John Benemann left in 1983 for the Georgia Institute of T echnology). The pond system described earlier continued to be used for this project. The objective was to obtain long-term productivity data with a pilot-scale system and generally demonstrate the requirements of large-scale algal mass cultivation (Weissman 1984; Weissman and Goebel 1985, 1986).

The first challenge was to obtain microalgal species that could be grown on the fresh to slightly brackish water available at the site. The common experience is that either inoculated strains from

culture collections fail to grow in the outdoor ponds, or that they grow initially but become rapidly outcompeted by indigenous strains. A common practice is to make the best of a bad situation and cultivate the invading organisms. This was also the experience and approach of this project.

After inoculation of Scenedesmus obliquus strain 1450 from the SERI Culture Collection, a strain of Scenedesmus quadricuada invaded. This turned out to be the most successful organism, cultivated for 13 months in fresh water and an additional 3 months in brackish. After an Oocystis sp. (Walker Lake isolate) was inoculated, a Chlorella sp. became dominant and was maintained (or maintained itself) for 2 months under semi-continuous dilution. However, some strains provided by SERI researchers could be grown for at least a few months outdoors, including an Ankistrodesmus falcatus and a freshwater Scenedesmus sp. So2a.

Productivity for S. quadricuada grown semi-continuously which is harvested every few days (a “sequential batch” growth mode), averaged about 15 g/m2/d for the 8 month period of March through October, with monthly averaged solar conversion efficiency ranging from 1.2% to 2.2% (Figure III. B.6.). “Typical” real biomass density just before harvest ranged from 60 to 100 g/m2, except for May, which recorded the highest standing biomass (160 g/m2) and productivity (20 g/m2/d). The continuously diluted cultures (diluted during the entire light period) exhibited approximately 20% higher productivity.

From a large number (39) of experiments, a correlation of Tmax, Tmin, and total insolation with productivity reduced the variance in the prediction of productivity by about 50% when using any single variable, but not in combination. This suggested that one of these three factors generally dominated (e. g., too high or too low a temperature or too little insolation). Similar experiments were carried out with the other microalgae in combination with the study of variables such as mixing speed, O2 outgassing, CO2 addition, and N limitation (for lipid induction).

The main conclusions of the extensive experimental program were:

1. Productivities of 15 to 25 g/m2/d were routinely obtained during the 8-month growing season at this location. However, higher numbers were rarely seen.

2. Continuous operations are about 20% more productive than semi-continuous cultures, but the latter densities are much higher, a factor in harvesting.

3. Culture collection strains fare poorly in competition with wild types.

4. Temperature effects are important in species selection and culture collapses, including grazer development.

5. Nighttime productivity losses increased to 10% to 20 % in July, when grazers were present; nighttime respiratory losses were high only at high temperatures.

6. There is a significant decrease in productivity in the afternoons, compared to the mornings, in the algal ponds.

7. Oxygen levels can increase as much as 40 mg/L, over 450% of saturation, and high oxygen levels limit productivity in some strains but not others. Oxygen inhibition was synergistic with other limiting factors (e. g., temperature).

8. Increasing TDS from 0.4 to 4 ppt decreased productivity, depending on strains.

9. Mixing power inputs were small at low mixing velocities (e. g., 15 cm/s) but increased exponentially. Productivity was independent of mixing speed.

10. The strains investigated in this study did not exhibit high lipid contents even upon N limitation.

11. The transfer of CO2 into the ponds was more than 60% efficient, even though the CO2 was transferred through only the 20-cm depth of the pond.

12. Harvesting by sedimentation has promise, but was strain specific and was increased by N limitation.

13. Initial experiments demonstrated that media recycle is feasible.

14. Project end input operating costs for large-scale production (@ $50/mt CO2,

70% use efficiency, etc.) was $130/mt of algae, of which half was for CO2 and one-third for other nutrients, with pumping and mixing power only about $10/mt.

This project answered a number of issues that had been raised about this process. One initially controversial observation was the finding that mixing speed had no effect on productivity (Figure III. B.7.). However, this experiment used a strain of Chlorella that did not settle, and care was taken to keep other parameters identical (in particular pH and pO2 levels). Thus, the increased productivities seen in some experiments (e. g., those of Hawaii), could possibly be accounted for by differences other than those of mixing, such as changes in outgassing of O2.

From the perspective of large-scale biomass production, one conclusion from this research was that mixing power inputs make any mixing speed much above about 30 cm/s impractical, as the energy consumed would rapidly exceed that produced. The rate of mixing should only be between about 15 and 25 cm/s, sufficient to keep cells in suspension and transfer the cultures to the CO2 supply stations in time to avoid C limitations in large-scale (>1-ha) ponds.

For low-cost production higher productivities would reduce capital, labor, and some other costs, but nutrient (e. g., CO2) related costs would not change. This suggested the need for low-cost CO2, and other nutrients, as well as a high CO2 utilization efficiency. Efficient utilization of CO2 appeared feasible based on the results obtained with even this unoptimized system.

Another major conclusion was that competitive strains would be required to maintain monocultures. The need for feedback from the outdoor studies to development of laboratory screening protocols was a major recommendation. Specifically, the relatively controllable parameters of CO2, pH, and O2 were of interest in determining species survival and culture productivity. Also, harvesting was identified as a specific area for further research. Finally, lipid induction remained to be demonstrated. These were the general objectives during the final year of this project, described in Section III. B.3.d.

Sustained	Maximum
Numbers above bars indicate photosynthetic efficiency (% of total)
XvX
“Ш!
Ш
Oct	NOV	May
Mar

Month

Figure III. B.6. Long-term productivity of S. quadricauda in freshwater (100-m2 pond). (Source: Weissman and Goebel 1985.)

Average linear mixing velocity (cm/sec)

Figure III. B.7. Chlorella mixing velocity experiments.

Data from 3.5-m2 ponds operated under constant pH and pO2 levels and compared to a control culture operated at 30 cm/s mixing velocity. Numbers above the data points indicate productivity of the control. Max/min temperatures 20/30 oC (15/25 oC for the 60 cm/s data set). (Source: Weissman and Goebel 1986.)

The problem with integrating microalgae biodiesel production with any high value coproducts or byproducts, such as pigments, vitamins, or specialty chemicals, is that these would be produced in very large amounts, saturating any likely markets. And, of course, the requirements for producing such high value products are quite different from the needs for biomass fuels. One example for coproducts comes from fuel ethanol production from corn, which is economically dependent on animal feeds (distillers dried grains), byproducts for economic viability, in addition to the more than $1/gallon in subsidies. Indeed, only large byproduct markets, such as animal feeds, could be realistically considered in the context of biodiesel production. However, although it may be possible to coproduce proteins with algal lipids, such an optimization (e. g., for high protein feeds) is likely to be difficult. Another major problem, as in distillers dried grain, is the drying costs. Overall, higher-value feed coproducts cannot, and should not, be a major driving force in developing this technology.

Of course, the likely route for the future development of practical and commercial large-scale microalgae culture technology will be through development of specialty foods and animal feeds coproduction. For example, Spirulina with two farms in the United States, comprising more than 100-ha of ponds, is becoming a commodity product, with bulk prices declining by almost half

(from the high 20s to the mid teens in dollars per kg) during the past 2 years. If this trend were to continue, to below $10/kg, this algal biomass could become a significant ingredient in aquaculture and other specialty animal feeds. Larger-scale systems for poultry feed production (microalgae high in xanthophylls, for example), or even cattle and hog feeds, could be foreseen, but require a decrease in costs (prices) to about $1,000/ton.

But to be considered for fuel production, costs for microalgal biomass production would need to be reduced to the absolute minimum. This implies that productivities must significantly increase, and costs decrease from current levels. We argue that any future technology development effort for microalgae biodiesel production should exclude higher value byproducts or coproducts as a specific target. We believe that the needs of biodiesel production, specifically for high lipid productivities, must be the objectives of such a program. One exception is the combination of such a process with wastewater treatment, as there are few likely alternative uses for the biomass. The other is the utility of such processes in greenhouse gas mitigation, also the objective of the Japanese R&D program in microalgae CO2 utilization, discussed briefly in the following section.

Little is known about the regulation of gene expression in diatoms, partly because genetic transformation was not possible in this group of algae before NREL’s transformation system was developed. The availability of this transformation system now allows the study of the roles of regulatory DNA sequences in gene expression. As a first step toward a better understanding of gene transcription in diatoms, NREL researchers Paul Roessler and Steve Milstrey used the firefly luciferase reporter gene (discussed earlier), to study the level of gene expression as controlled by various DNA regulatory sequences from the diatom C. cryptica and other organisms. They also used this system to try to define the regions of the ACCase gene promoter involved in the Si-depletion response.

Various plasmids were constructed in which different combinations of 5’ regulatory DNA sequences (promoters) and 3 ’ regulatory DNA sequences (terminators) were linked to the firefly luciferase gene (luc). The regulatory sequences used in this study included both the ACCase promoter and the UDP-glucose pyrophosphorylase/phosphoglucomutase promoters from C. cryptica. Also tested were the simian virus 40 (SV40) promoter, which drives high levels of gene expression in mammalian cells, and the cauliflower mosaic virus 35 S RNA promoter (CaMV35S), which is a strong constitutive promoter in plants. These plasmids were introduced into C. cryptica via cotransformation with the selectable marker plasmid pACCNPT5.1 as described earlier. Approximately half of the transformed strains produced in this manner contained the luc gene, as determined by PCR analysis. Based on past results, it is expected that the plasmid DNA was integrated into the genome of the cells. Luciferase activities in randomly

chosen transformants (eight from each plasmid type) were determined by the use of a luminometer.

As expected, the promoter regions of both C. cryptica genes drove luciferase expression in the transformed C. cryptica cell at high levels. Less predictable was the finding that the SV40 mammalian promoter also drove luciferase expression in C. cryptica at relatively high levels (although lower than seen using the homologous promoters), but the CaMV35S promoter was much less effective. In most of the constructs used in this study, the 3′ terminator regulatory region was from the C. cryptica aac1 gene. Replacement of this sequence with the SV40 terminator did not affect the levels of luciferase expression driven by the accl promoter, indicating that the source of the terminator sequence used may not be a critical determinant of gene transcription efficiency.

Previous results at NREL indicated that Si deficiency may affect the expression of the acc1 gene in C. cryptica (see Section II. B.2.d.). To try to identify regions of the acc1 promoter that might be responsive to Si levels, three plasmids that contained varying lengths of the acc1 promoter region (900, 445, and 396 bp, respectively) fused to the luc gene were used to transform C. cryptica. Under Si replete conditions, the average luciferase activities of transformants containing these plasmids were very similar. Furthermore, the luciferase activity increased to the same extent (approximately twofold) 6 hours after transfer into Si-free medium. This suggests that the Si-responsive elements are either within the shortest (396-bp) promoter region tested or in a separate area of the genome.

These results indicate that the firefly luciferase gene can be expressed in recombinant C. cryptica cells, to provide a sensitive reporter system for analyzing gene expression and promoter function in diatoms. This and similar systems will likely be extremely useful for gaining a better understanding of the molecular biology of this important group of organisms.

III. C.1. Introduction

One major concern of the ASP, as for any renewable energy option being developed by DOE, is the resource potential for the technology. How large an impact will it have on future U. S. energy supplies, or, in today’s units, fossil CO2 or equivalent greenhouse gas reduction potential? This is required both by the mission of the DOE as well as the inherent need to justify budgetary decisions. However, potential contributions to fuel supplies should be only one, albeit an important, parameter in such an assessment. Other factors must also be considered, such as economics, time frame for implementation, and the possibility of success for the R&D effort. If large resource potential is the main criterion, this could result in focusing too many resources on a few technologies with a low probability of succeeding in practice.

Microalgal biodiesel is one of many different biomass energy options, extending from co-firing of wood in power plants to energy farms and a myriad of fuel conversion options (Hughes and Benemann 1997). One important attribute of microalgae systems is that they need not compete with other biomass alternatives, but must be able to use water and land resources generally not considered for crop production. Also, microalgae fuel production systems could meet other objectives, from waste treatment to salinity management. Resource potentials are important, and have been studied extensively by the ASP.

The ASP emphasized the production potential of microalgae in the southwestern United States. This choice was based in large part on the perspective that this area offered the best and largest resource potential for algal culture systems, including available saline water supplies, land area, sunlight, and CO2 sources. Each of these resources individually was projected to have a potential of many quads (1015 Btu) of energy, and overall potentials of several quads were projected (e. g., Chelf and Brown 1989). This certainly justified the emphasis by the ASP on this geographical area. In fact, a significant activity of the ASP was to document this resource potential. It was not a trivial task to assess these resource potentials. This required the development of suitable assessment techniques. This was a major accomplishment of the ASP, as reviewed in this section. However, estimating an actual resource potential is not a simple matter, as the juxtapositions of the requirements for microalgae production—water, land, CO2, and climate— are very difficult to quantitate, and would greatly reduce the resource potential estimated from single factor analyses.

III. B.2.a. Hawaii ARPS Project Initiation, 1980-1981

As mentioned earlier, the concept for the ARPS project derived from the Raymond (1981) patent. The project was initiated in early 1980s, with construction of a single, 48-m2 raceway system completed in early 1981 (Laws 1981). During this first year, chemostat experiments using two strains of P. tricornutum were carried out. The tests revealed large differences in protein and lipid productivity between the strains. This laboratory work also investigated cell harvesting by “foam fractionation” in which the foam formed by the aeration of the cultures via air lift mixing was collected and found to contain some 10-40 times the cell concentration of the liquid. However, the harvesting efficiency was not reported.

One difficulty noted in the laboratory experiments was the low cell densities achieved, compared with the original reports by Raymond for the ARPS system. Researchers tried to increase cell density by increasing the pond depth to 0.6 m, rather than 0.1 m as proposed by Raymond (1979). This resulted in other problems (low cell density, shading-see below), and the depth was again reduced to 30 cm. The laboratory experiments were extrapolated to predict an outdoor productivity of almost 130 mt/ha/yr, at least for the best P. tricornutum strain, similar to the Raymond prediction. However, this extrapolation was based on the invalid assumption that such laboratory growth rates can be used to predict outdoor productivities.

The initial work also studied the effect of CuSO4 filters, concluding that although any productivity increases would be minor, installation of CuSO4 filters is advisable in the APRS, “as it would help manage overheating of the culture and store a potentially valuable by-product, heat.” Finally, the “flashing light” effect was investigated. The time constants (1 s light: 1 s dark), and low light intensities used were quite different from the classic flashing light effect of Kok (1953), which uses approx. 1-5 psec high intensity light flashes, followed by about five times longer dark periods. Only small, “not particularly encouraging”, productivity increases were noted. Still, a 70% increase in productivity was predicted, though the basis for this was not stated.

Laws (1981) also reported on initial results with the 48-m2, 0.6-m deep, airlift-mixed flume system. Cell densities were much lower than predicted, likely because of the great depth of the culture, which was later reduced. The report concluded that, assuming $30,000/ha/yr production cost, a biomass production of 180 mt/ha/yr AFDW would allow oil production (with protein byproducts) competitive with fossil fuel. However, this productivity figure was extrapolated from the indoor chemostat work, and increased by one-third, as “effects of modulated blue light on the system will allow the extra production to be achieved,” so the reliability of this prediction is questionable.

In 1986, a report was generated by SERI ASP in-house researchers that analyzed factors and costs involved in microalgae biomass production systems (Neenan et al. 1986). The report reviewed the various system components and requirements for algal biomass production, and summarized and extended the available resource analyses for water, land, and most importantly, CO2 required for large-scale microalgal biomass production. It also reviewed the various fuel product alternatives from microalgae biomass, including ethanol, methane, PVO (“pseudovegetable oil”), biodiesel (methyl ester fuels), and even gasoline, (see also Feinberg 1984).

Although generally following the HRP system concept and design described earlier, the authors raised a number of issues and questions. For example, they concluded that paddle wheel mixing was not “the optimal design,” although paddle wheel mixing was used in their analysis. Some modifications were made in the engineering analyses, such as replacing channel dividers by plastic fences, and the use of clay liners in the ponds. For harvesting, microstrainers or belt filters were used as primary harvesters, followed by centrifugation. However, the overall design and cost estimates were essentially based on the Benemann et al. (1982) analysis, including scale (860-ha of pond surface).

Using these inputs, an Algal Production and Economic Model was developed using various unit costs (costs of fertilizers, land, power, water, CO2, etc.) and design parameters (module size, depth, nutrient concentrations and losses, mixing velocities, etc.), financial factors, and operating parameters (retention times, algal biomass composition, growing seasons, efficiencies, etc.). As in the prior studies, the residues from the fuel extraction/processing would be converted to methane gas by anaerobic digestion. Schematics and process diagrams for the various processing options were developed. The reference case assumed a biomass with a 30% lipid content and a 17 g/m2/d average annual productivity (62.5 mt/ha/yr). The overall projected system costs for the reference case were $43,283/ha of ponds and $433/mt of algal biomass. This compares to $39,850/ha and $274/t in the “conservative” case of the Benemann et al. (1982) analysis (Table III. D.2., pure CO2 case). That was for a somewhat higher productivity (67.5 mg/ha/yr) and lipid content (40% versus 30% in this study). It is difficult to extract the specific cost differences between these analyses. However, the cost routines and parameters used had larger effects on costs than the engineering estimates. For one example, in Neenan et al. (1986) water costs were 12% of overall costs, compared to less than 4% in Benemann et al. (1982).

Compared to a biomass cost of $433/mt estimated for the reference case, the allowable feedstock costs for the various fuel options were calculated to be only somewhat above $ 100/mt (and even less for ethanol). Table III. D.3. provides a summary of this analysis, for a 36,000 mt/y (33,000 t/y) microalgae biomass to fuels processing plant. The table provides capital and operating costs only for the fuel processing units, and derives an allowable algae cost, in $/t. There is a large discrepancy between these estimates and the projected biomass production costs. This led to the conclusion that such a process “is currently not commercially viable.” In fact, the main and

coproduct credits in this analysis were projections for the year 2010, when diesel or methane was expected to cost some four times current (1998) costs (even without inflation adjustment). This makes the economics of this process even less attractive.

As in the prior analysis (Benemann et al. 1982a), a cost reduction and process improvement effort was undertaken and “attainability targets” developed. First, sensitivities were run for 13 resources (such as power costs and evaporation), 15 facility design parameters (e. g., culture depth and mixing), three biological parameters (such as growing season) and eight financial parameters (cost escalations, etc.). Taken one at a time, most factors did not reduce costs significantly (except for growing seasons, culture depth and source water CO2 content). Although some of the results are difficult to interpret (for example, the large decrease in costs with increasing depth), the major conclusion was that no single parameter dominated costs sufficiently to achieve the goal of low-cost fuel production. Of course, several parameters in combination could do so. In particular, by increasing productivity to as high as 8% of total solar conversion efficiency and 50% lipids (50 g/m2d), and by assuming a capital investment of $48,000/ha, an algal production cost of $211/mt was estimated. Using this cost, the microalgae biomass fuel processing costs were again estimated, allowing calculation of allowable fuel product costs, of $1.65/gal of biodiesel.

The fuel processing cost estimates, the major contribution of this report, were very preliminary and based on many assumptions. For example, the costs of the transesterification plant were rather high, and might be reduced in the future. But, as a central conclusion, productivity was again the most important parameter: increasing production efficiencies by a factor of about four decreased production costs by almost half. The report concluded with a detailed analysis of the “attainable” process improvements, emphasizing the need for achieving multiple cost reductions, in addition to significantly increased photosynthetic efficiencies. The report concluded that “aggressive research is need to fulfill the performance requirements defined by this analysis”.

I Publications:

Feinberg, D. A. (1984) “Fuel options from microalgae with representative chemical compositions.” Report, Solar Energy Research Institute, Golden, Colorado, SERI/TR-231 -2427.

Neenan, B.; Feinberg, D.; Hill, A.; McIntosh, R.; Terry, K. (1986) “Fuels from microalgae: Technology status, potential, and research requirements.” Report, Solar Energy Research Institute, Golden, Colorado, SERI/SP-231-2550, 158 pp.

Table III. D.3. Costs of microalgal biofuels production: reference case.

(Source: Neenan et al. 1986.)

Summary of reference capital, operating, and allowable feedstock costs for microalgae fuel processing options. Of the five fuel options, the first four produce a high (30%) lipid biomass; the last (ethanol) fermentable carbohydrate, representing only 13% of biomass weight.

Notes:

a. Not including algae feedsock

b. Process and cooling waters

c. By-product prices: $0.07/m3 for CO2 (captive use only), $7.40/MMBtu for methane, fuel gas, or LPG, and $13.30/MMBtu diesel fuel (exports)

d. Main product prices: $7.40/MMBtu for methane, fuel gas, LPG, and $16.60 ($1.75/gal) for gasoline and diesel fuel, $1.20/gal for ethanol and $1.75/gal for biodiesel (methyl ester) or PVO

e. The value ($/t) of the algal feestock in producing the main and coproduct mixes. For example, for the ester fuel case, the facility would produce 2.1 million gallons of biodiesel in addition to some 600,000 GJ of methane and 760 t of glycerol for export (plus 11 million m3 CO2 and 2,400 tons N, recyled internally) at prices listed in c. and d. above.

Table III. D.4. Costs of microalgae biofuels production: attainability cases.

(Source: Neenan et al. 1986.)

Summary of Attainable Capital and Operating Costs for Fuel Processing Options.

Notes:

a. Includes microalgae biomass feedstock production costs of $211/mt

b. Process and cooling waters

c. By-product prices: $0.07/m3 for CO2 (captive use only), $7.40/MMBtu for methane, fuel gas, or LPG, and $13.30/MMBtu diesel fuel (exports)

d. Main product, million gallons/yr.