Content and composition of lipids in Scenedesmus obliquus
As shown in Fig. 1a, the variation rule of lipids was found out that it increases first and then followed by a reduction, it reaches the highest value on the eighth day. The total lipid content of S. obliquus was 17.7% on day 0. On the eighth day, the lipid content reached 51.6%, which was 2.9 times of the initial value. We also used GC-MS to analyze the proportion of fatty acids in lipids and the variation in the fatty acid profile of S. obliquus in stress culture are summarized (Tab. 1). The proportion of SFA has reduced by 45% and the proportion of UFA has increased by 12% from day 0 to 8. GC-MS combined with dry weight method showed that the dry weight ratio of UFA increased from 11.16% to 44.51%, while the dry weight ratio of SFA decreased from 6.90% to 6.28% from day 0 to day 8.
Microscopic examination of Scenedesmus obliquus
Image contains a lot of biological information. Obtain intuitive and clear images in the field of microbiological research, which can better analyze the characteristics and status of specific areas of cells or organisms, as well as the distribution of specific molecules. NR result indicated lipid substances widely exist in cells and cells of S. obliquus appear yellow-orange in a fluorescence microscope after lipid staining. With the presence of accumulated oil droplets, they turn bright yellow. Fig. 1b shows the fluorescence microscopic observation of different lipid accumulation in S. obliquus cells. However, due to the limited resolution of fluorescence microscopy, we used high-resolution TEM to supplement the observation of the microstructure of S. obliquus. The observed results were presented in Fig. 2. The study indicated that oil droplets were located near the cell membrane in the cytoplasm, and were mostly light gray circles or ellipses in TEM observation field.
Observation of Scenedesmus obliquus lipids by Raman spectroscopy
The effective band of 900cm-1 to 1700cm-1 was processed with the methods of cosmic rays’ removal, Savitzky-Golay smoothing and baseline correction [27-29]. As shown in Fig. 3a, the Raman spectroscopy of S. obliquus had a few peaks at 965, 1005, 1160, 1187, 1266, 1353, 1445, and 1525 cm-1, which respectively correspond to the metabolic components in S.obliquus [30-35]. Studies have shown that the corresponding H-C=bending plane at 1266 cm-1 mainly represented the fatty chain unsaturation , and the corresponding C-H2 bending bond at 1445 cm-1 mainly represented the saturated carbon chain [12,36]. After analyzing the Raman curve above, we further performed imaging analysis, and tried comparing the results with NR and GC-MS to verify the feasibility of the results and explore the advantages of Raman in the distribution of S. obliquus lipid. Interpolation method was used to map the selected characteristic peaks into pseudo-color images of lipid distribution. In general, the redder the color in the captured image, the more lipids, and vice versa. As shown in Fig. 3b, the pseudo-color map of the distribution of UFA in the cells of S. obliquus was plotted with a peak value of 1266 cm-1 and the pseudo-color map of total fatty acids (TFA) distribution was plotted with the 1445 cm-1 peak value after correction.
It is observed from the distribution of unsaturated fatty acids in S. obliquus cells that the UFA are discontinuously distributed, and most of them are distributed in the two regions of the cells. The red area in the cell indicates the accumulation of UFA. The darker the color, the higher the UFA content. Comparing the NR and TEM results with the Raman results, the oil droplets were located near the cell membrane in the cytoplasm. Over time, the red area in S. obliquus gradually became larger and darker. It was believed that the UFA content gradually increased, and the UFA content was highest on the eighth day. The changes of pseudo-color maps of TFA and UFA are similar. The results above show that Raman spectroscopy is effective for the detection of intracellular lipid in S. obliquus.
Metabolic analysis by Terahertz
Some previous studies have shown that pure categories such as proteins, lipids, carbohydrates and carotenoids can be identified by THz molecular vibrational spectroscopy techniques. The main components of S. obliquus are protein, lipids, carbohydrate and carotenoids and their terahertz absorption spectra are distributed in the different frequency ranges: 3.3-5.0 THz for proteins [15,37], 2.3, 9.3, 9.4 , 9.8, 11.4 THz for lipids [25,26], 9.0, 10.5, 12.1, 13.1 16.0, 17.2 and18.0 THz for carbohydrates [16,38], 12.1, 14.7, 15.6 and 19.6 THz for β-carotene . Terahertz has been used for the study on different types of lipids and fatty acids [25,26]. For example, saturated fatty acids (SFA) such us palmitic acid and stearic acid have distinct peaks at 9.3 and 9.4 THz, respectively. Unsaturated fatty acids (UFA) such as oleic acid, linoleic acid and linolenic acid all have two distinct peaks at 7.4 and 9.8 THz. In general, lipids usually show broader and distinct absorption peaks. In the spectra of S. obliquus cells, the peak of 9.3 THz corresponded to the C=O and -COO- vibration . This peak mainly represented the lipids. The broad band of 3.3-5 THz corresponded to the population of overlapping discrete vibrational modes from the random amino acid sequences. It is similar to other globular proteins, such us myoglobin, hemoglobin, and lysozyme, which showed that microalgae proteins lack of highly repetitive segments in their amino acid sequences [15,37]. The peaks of 15.6-18.0 THz mainly corresponded to the C-C-O, C-O-C, C-C-C covalent skeletal deformation [16,38,39]. These peaks are fused due to changes in the corresponding substance content, resulting in a significant absorption peak in the spectra of S. obliquus.
Relating changes in Terahertz band under nitrogen stress
After confirming the corresponding absorption peak of main components in microalgae, analysis of several main absorption peaks in the S. obliquus spectra can reflect the changes of the main components, such as proteins, carbohydrate, lipids, carotenoids in the S. obliquus. Changes in the terahertz band and its corresponding components during microalgal metabolism were analyzed under nitrogen stress. During the cultivation under nitrogen stress, the protein content of the microalgae cells continues to decrease, the lipid content increased first and then decreased with its maximum on the eighth day. The carbohydrates first increased and then stabilized, and the β-carotene content of the S. obliquus cells continued to increase. As shown in Fig. 4a, the absorption band of 3.3-5 THz was red-shifted, which may be due to the changes in protein types of the cells. A study of microalgae showed that nitrogen starvation triggers the accumulation of multiple proteins associated with oxidative phosphorylation , such as NADH dehydrogenase, ATP synthase and cytochrome coxidase. In general, this absorption band strength is gradually reduced because of the increase in nitrogen stress time. Absorption peaks of β-carotene and carbohydrates are difficult to distinguish at the band 15-18 THz. But from the red and blue shift of the absorption peak, the carbohydrate and β-carotene response and cumulative speed can be found. The carbohydrates rapidly accumulate, resulting in a blue shift, such as on the fourth day. The β-carotene accumulates rapidly, resulting in a red shift, such as on the tenth day. Nitrogen deficiency induces β-carotene accumulation, but the response speed is significantly slower than that of carbohydrates.
PCA and PLS modeling based on THz spectroscopy
Spectra in the range of 2-20 THz for each day were further analyzed by PCA algorithm to distinguish these ingredients. THz spectra of the cells in different cultivation days showed obvious clustering in PC-1 and PC-2 directions (Fig. 4b). PC1 and PC2 explained in total 97% of the variation in the spectra (85% by PC1 and 12% by PC2). The aggregation of algae samples is mainly affected by protein, carbohydrate, lipid and pigment content. Among them, the decrease of protein content and the increase of lipid content have the greatest influence. Combining the results shown in Fig. 2a and Fig. 2b, the absorbance value of lipids changes most obviously at 0-10 days, which is consistent with the clustering effect of samples analyzed by PCA at 0-10 days. PCA analysis shows that the THz spectra of S. obliquus can be effectively distinguished for different growth days under nitrogen stress, and it is feasible to analyze the changes of microalgae metabolites.
Terahertz data were divided into modeling set and predicting set with a ratio of 2:1. Three different modeling schemes were used: full band modeling, lipids characteristic band modeling (2.3, 4.6, 5.3, 6.0, 9.3, 9.4 , 9.8, 11.4 THz), and other characteristic band modeling (3.3, 9.0, 10.5, 12.1, 13.1, 14.7, 15.6, 16.0, 17.2, 18.0, 19.6 THz). Using PLS algorithm to model the data of modeling set to discriminate different lipid content of S. obliquus under nitrogen stress. When building PLS prediction model, the accuracy of prediction model is determined by comparing correlation coefficient and root mean square error of calibration set and prediction set. It can be seen from Table 2 that the lipid characteristic band modeling and full band modeling are better than other band modeling. The correlation coefficient (r) of the lipid characteristic band modeling is close to that of the full band modeling, but considering RMSE and RPD, the lipid characteristic band modeling is the best. These results indicated that the model constructed by terahertz spectrum of lipid characteristic band can effectively distinguish Scenedesmus obliquus with different lipid content under nitrogen stress, and the accuracy of the model was relatively high. On this basis, we combined terahertz bands to model and obtained better results, which proves that this prediction method was effective.
Quantitative analysis of lipid by lipid characteristic peak area
In the quantitative analysis of the spectrum, two methods can be used: the peak intensity of spectrum (the peak height in the spectrum) and the integration of the characteristic peak area. Next, we will carry out the quantitative analysis of the lipid of S. obliquus based on the characteristic peak area. In the terahertz spectrum, there is a lipid characteristic peak that changes obviously over time at 9.3 THz. In the Raman spectrum, the lipid characteristic peak of 1445 cm-1 is considered to be used for semi-quantitative monitoring of the lipid content of microalgae [12,28,36].
Fig. 5 shows a comparison between estimated and actual lipid content for the same samples. The estimated lipid contents were obtained from characteristic peak areas at 9.3 THz and 1445cm-1. The estimated lipid content using the characteristic peak of 9.3 THz ranged from 17.3% to 52.6% (standard error≤6.3%), the estimated lipid content using the characteristic peak of 1445cm-1 ranged from 18.0% to 57.2% (standard error≤11.0%). The terahertz model had r of 0.960 and the Raman model had r of 0.783. For the prediction of lipid content in microalgae based on lipid characteristic peaks, the accuracy of the Raman model in this paper is similar to the results of other studies [12,28], and the prediction accuracy of the terahertz model based on lipid characteristic peaks is higher than that of the Raman model based on lipid characteristic peaks.