by Lieve Laurens, Ed Wolfrum, and Al Darzins
At NREL’s National Bioenergy Center, we are working towards developing a new method for estimating the oil or lipid content and chemical composition in a wide range of algal strains using near infrared (NIR) spectroscopy. Algae have great potential to accumulate large quantities of lipids that can be converted into a variety of biofuels including biodiesel and jet fuel. However, the level of lipids can vary considerably between strains and changes with growth and environmental culture conditions.
A large emphasis of current algal biofuels research is focused on selecting the ‘right’ strain for large-scale cultivation. In this context, the right strain refers to several parameters that are known to influence the economics of the algal biofuels production process. Typically, the biochemical composition of algal biomass, in particular the lipid content, plays a big role. It has been observed that over the algal growth cycle there is considerable reshuffling of the cell’s biochemistry going on (i.e. the relative composition of lipids, proteins and carbohydrates changes over time). Many different environmental conditions and variables can affect the lipid content and thus there is a need for a rapid and accurate method for lipid quantification of algal biomass.
Traditional lipid analysis methods typically involve a solvent extraction protocol that is first designed to isolate the lipids from biomass. The total lipids are quantified using a gravimetric determination followed by a more detailed analysis by gas or liquid chromatography with mass spectroscopy detection. These methods yield a lot of information about the lipid content and composition, but unfortunately, the process is time and resource intensive. For these analyses, several grams of algal biomass are needed and the results are usually not available for several days.
Because of the large number of algal strains and variables that can affect algal lipid production, the number of samples to be analyzed quickly increases to levels that are hard to analyze by the traditional wet chemical methods. There is a thus a need for a rapid technique for screening and selecting algae from, for example, a culture collection. This type of screening method should be able to quickly assess the lipid content and select promising strains. Currently fluorescent lipophilic dyes, such as Nile Red and BODIPY (4,4-difluoro-1,3,5,7,8-pentamethyl-4-bora-3a,4a-diaza-s-indacene), are used for lipid quantitation and strain screening because of their selective affinity for neutral lipid droplets inside the cells. However, a major disadvantage of the dye-based assays is that they are affected by uneven dye uptake due to the inherent variability of different strains of algae and their cell wall composition which can be affected by growth conditions.
The aim of our work using infrared spectroscopy is to develop a high-throughput technique that is capable of monitoring a large number of samples with a minimal investment of time. The use of NIR for this purpose is not new and has been used to predict the chemical composition of lignocellulosic biomass such as corn stover and for the determination of trans-isomers in fats in the past. The use of infrared spectroscopy has been shown to be able to monitor biochemical changes in algal biomass and also has been useful to distinguish and classify cyanobacterial strains.
NIR spectroscopy measures the absorption of energy in the IR region of the spectrum by chemical bonds in molecules. The advantage of NIR spectroscopy is its tolerance to a certain level of variation in the samples and minimal sample preparation requirement. Overall, NIR spectroscopy can be applied as a fast, accurate and non-destructive analytical method that requires only very small amounts of homogenized biomass (~ 10 mg) using a 96-well plate set up. (Figure 1).
Because of the broad overtones rather than distinct peaks in the NIR spectra (Figure 2), this method relies on the use of chemometrics for quantification. Chemometrics in this context refers to multivariate calibration methods, routinely used in analytical chemistry. The biggest advantage of these models is the large number of samples that can be screened and analyzed, avoiding the need for laborious wet chemical analyses. With further development, there is the potential for NIR spectroscopy to be applied to actively growing algal cultures for real-time monitoring of the lipid accumulation or to screen for select strains that have high lipid content amongst a large number of strains.
Depending on the application of this screening method, one may want to develop and apply a NIR model that is optimized for a single-species or for multiple-species using a combined model. Using such models one could screen, for example, mutant or transgenic lines belonging to one species that have significantly increased lipid content, or test a range of culture conditions for one species that causes the induction of lipid accumulation. Alternatively, a combined, multiple-species NIR model could be used to screen a large number of algae strains collected from different aquatic environments to detect native, yet unusually high-lipid, producers.
We have collected data on multiple strains of algae that contain a wide range of lipid contents and used this to build and validate multivariate calibration models. In our first study, which has now been published in BioEnergy Research, we investigated whether there is enough information in the NIR spectra collected from algal biomass to build prediction models that are able to i) quantify the lipid content in algal biomass and ii) distinguish between triglyceride and phospholipid content. We took spectra of homogenized, dried biomass spanning four major divisions of photosynthetic microorganisms; the green algae (Chlorophyceae), Eustigmatophyceae, diatoms (Bacillariophyceae), and blue-green algae (Cyanobacteria). We spiked in two types of lipids, a phospholipid and a triglyceride, at nine different concentration levels. We then tested whether the spectral data could be correlated to the amount of spiked lipids, and how well these newly built models would work to predict the level of spiked lipid in new sample spectra. Because the models are highly influenced by the quality of the spectra, we investigated the effect of mathematical pretreatment of the raw spectra on the quality of the prediction models. Because we wanted to find out whether these models were applicable across different strains of algae, we looked at the accuracy of prediction of the single-species models as well as the combined multiple-species models on the respective biomass samples. We found that the combined models were actually more robust in the prediction across the different biomass samples.
More recently, we have been collecting spectra from a large number of algal biomass samples from different Chorella vulgaris strains and a culture collection set up at NREL and built and validated prediction models for these samples. The results of this work are promising and will be reported in a follow-up manuscript.
Using the NIR spectroscopic methods, we can predict the algal oil content of almost any algal species in a matter of minutes rather than days. We have developed calibration models, for single species as well as for multiple species combined, where the infrared fingerprints are correlated with lipid content. These calibration models were then subsequently used to predict the lipid content in new, unknown samples. We have created models with over 200 spectra and have successfully applied the model to new biomass samples. We will continue to develop these models to make them more robust so we can apply them across species and biomass samples obtained from different sources.
Laurens, L. M. L. and E. J. Wolfrum (2010). “Feasibility of Spectroscopic Characterization of Algal Lipids: Chemometric Correlation of NIR and FTIR Spectra with Exogenous Lipids in Algal Biomass.” Bioenergy Research: DOI: 10.1007/s12155-010-9098-y