The uptake of D31-PA and D27-MA by HMEC-1 cells was investigated first by Raman imaging, a method that uses the unique spectral signature of the C–D stretching vibration (Fig. 1). Accumulation of D-FA was analyzed in TNF-α pretreated cells in comparison to the (untreated) control. HMEC-1 inflammation was confirmed by immunochemical staining. The effect of D-FA uptake on HMEC-1 cell morphology was further investigated by AFM and two coherent Raman microscopies (SRS and CARS).

Fig. 1

The scheme of experiments and their purpose

The uptake of fatty acids by endothelial cells depends on the type and concentration of acid, and the condition of the cells

An uptake of D27-MA or D31-PA by normal and TNF-α pretreated cells was studied at 50 µM and 400 µM concentrations. Raman images provided information about the subcellular localization of endogenous FA and exogenous D-FA, which can be distinguished due to deuterium labeling. Figure 2 shows representative Raman images based on the integration of the v(C–H) band in the spectral range 2830–2900 cm−1, displaying regions rich in endogenous lipids (i.e., LD and ER), while the spectral range of 2060–2240 cm−1 (v(C–D)) was used for visualization of spots rich in D-FA. This discrimination is feasible because the C–D stretching vibration occurs in the “silent” spectral region. For the determination of the endogenous lipids and D-FA content, the Raman images based on v(C–H) and v(C–D) integral intensities were normalized. The scale bars displayed next to the Raman images indicate the maximum intensity of v(C–H) or v(C–D) Raman bands for all images in a row. This protocol was used to compare data obtained in the same conditions for D27-MA and D31-PA (presented in the same row), so the intensities indicated on the respective scale bar for the C–H and C–D image for the same experiment (in the same column) are not directly interpreted.

Fig. 2

Raman images of HMEC-1 cells upon D-FA uptake in normal and pathological conditions (TNF-α treated cells) HMEC-1 incubated with D27-MA (a) and D31-PA (b) at 50 µM and 400 µM concentrations in the reference to control (c)

HMEC-1 exhibits the ability to form LD spontaneously in normal conditions, but after treatment with the TNF-α, unsaturated LD are newly formed [1]. This effect was also observed here for all investigated cells, i.e., LD are clearly visible in the Raman images of HMEC-1 based on the intensity of v(C–H) vibration (Fig. 2a, b and c).

After incubation with D-FA at low concentration (50 µM), the v(C–D) signal can be seen widely spread in the cytoplasm (Fig. 2a and b), mostly in the area of ER (vide supra Fig. 3), with only a few evidently formed LD. Incubation with a higher concentration of D-FA (400 µM) results in the v(C–D) signal being omnipresent in the cytoplasm for D31-PA, while for D27-MA-treated cells LD are evidently visible. Additionally, looking at the intensity of the v(C–D) signal in cells incubated with D27-MA at a lower concentration, it can be noticed that their TNF-α stimulation led to increased D-FA uptake compared to unstimulated cells. On the other hand, when comparing the action of this acid in higher concentrations, the situation seems to be opposite. For D31-PA, it is difficult to interpret images collected by spontaneous Raman spectroscopy to clearly define the difference between stimulated cells in normal and pathological conditions, for both low and high concentrations. The number of cells measured was too small to make clear conclusions.

Fig. 3

KMCA of representative HMEC-1 incubated with D27-MA. The image shows four classes (blue–nucleus, brown–ER-rich area, orange–LD, gray–cytoplasm) (a). The average Raman spectra of the respective classes and spectrum of D27-MA (black, b). Scale bar equals 6 µm. All spectra were maximally extended on the y-axis

Raman images may be ambiguous and difficult to interpret when the assessment is based on visual inspection, so chemometric analysis of the lipidic part of cells was performed. To obtain a spectral profile of individual subcellular compartments, Raman data were subjected to KMCA. Representative KMCA of HMEC-1 incubated with D27-MA is shown in Fig. 3. It was possible to distinguish four cellular compartments, i.e., the nucleus (blue class, based on 785 cm−1 DNA/RNA marker band), the ER area (brown class, based on a high contribution of lipids, including phospholipids), LD (orange class manifested by high intensity of Raman features at 2852 and 2882 cm−1) and the cytoplasm (gray class, with a stronger contribution of proteins in comparison to other subcellular regions). For each class, the average Raman spectrum is shown (Fig. 3b). The characteristic signal originating from v(C-D) vibration at 2108 cm−1 is clearly seen in the spectrum of each class, with the highest intensity for the ER and the lowest for the nucleus. In the latter case, probably the signal from the cytoplasm above the nucleus contributed here.

The spectrum of LD (orange line, Fig. 3b) contains bands that are seen in the reference spectrum of D27-MA, but also other bands due to non-deuterated lipids (1666 cm−1 and more), as well as a small intensity shoulder at 1746 cm−1 can be noticed and assigned to the esterified form of lipids. From that observation we can hypothesize a passive absorption of free FA by cells (without the participation of transporter proteins, such as albumin) resulting in synthesis and accumulation of triacylglycerols in the cytoplasm of cells. It should be noticed also that all LD show a similar composition, so newly formed LD are composed not only of incorporated D-FA but also of cellular lipids, i.e. phospholipids.

To compare quantitatively the uptake of D-FA by inflamed cells in reference to control, the Raman spectra taken from cells treated with D-FA were grouped into two classes, i.e., the nucleus and the cytoplasm class, the latter includes ER region and LD together with other organelles (Fig. 4a). This approach enabled us to analyze the overall signal from D-FA originating from the v(C–D) vibration since not always LD were distinct enough and lipid signal was rather dispersed in the whole cytoplasm. For each experimental group, an average spectrum is presented in Fig. 4a and b together with the reference spectra of D27-MA and D31-PA. HMEC-1 treated with a low concentration of D-FA (50 µM) show only slight changes in the intensity of the v(C–D) band, i.e., a small increase in the case of cells incubated with D31-PA. For cells treated with a high concentration of D-FA (400 µM), for both D27-MA and D31-PA, the intensity of v(C–D) band is lower in the spectra from cells pretreated with TNF-α in comparison to control. To quantify the uptake of D-FA, the ratio of integral intensities of Raman bands originating from v(C–H) and v(C–D) was calculated (Fig. 4d and e). By doing so, the other effects which could influence the background of the spectra and hence bands intensity was eliminated. For the low concentration of D-FA (Fig. 4d), the uptake of D27-MA was greater than D31-PA. Such difference in the uptake of those two D-FA may result from the differences in their aliphatic chain length [53]. R.W. Mitchell et al. [53] classified D27-MA to medium-length saturated FA, while D31-PA for a long chain one. They have also shown that the uptake of FA through endothelium monolayer is easier for medium-length FA. Pretreatment with TNF-α enhanced the uptake of D27-MA used in low concentration, while did not affect the uptake of D31-PA.

Fig. 4

Quantification of the uptake of FA by inflamed HMEC-1 cells. Raman images obtained by integration of bands in the regions: 3030–2830 cm−1 (organic matter), 2880–2830 cm−1 (v(C–H)), 2150–2050 cm−1 (v(C–D)) and 800–770 cm.−1 (nucleic acids), together with KMCA image (blue class–nucleus, orange–cytoplasm) (a). The average Raman spectra of the cytoplasm of HMEC-1 cells treated with D27-MA (b) and D31-PA (c), respectively. Calculated integral intensity ratio for bands originating from v(C–D) and v(C–H) vibrations for low (d) and high concentration (e) of used FA. The spectra of reference D-FA are also shown. All spectra were maximally extended in the y-axis. *p < 0.05

For HMEC-1 cells incubated with D-FA in high concentrations, the observations are different. Treatment of EC with a concentration of 400 µM D27-MA resulted in its uptake; however, this effect was more pronounced for D31-PA. In turn, a decrease of the v(C–D) band intensity observed in the spectra collected from cells pretreated with TNF-α and incubated with D31-PA may suggest a slight downregulation of the D-FA uptake that occurs upon EC inflammation. The pretreatment with TNF-α did not influence the uptake of D27-MA applied in high concentrations.

Since the number of cells measured by Raman microscopy was relatively small to draw definite conclusions, coherent anti-Stokes Raman scattering (CARS) was used, which allowed the measurement of a large population of cells in a much shorter time (Fig. 5). It can be seen that the uptake of D31-PA depends on its concentration and the condition of the cells. Inflamed cells show increased D31-PA uptake at higher concentrations (400 µM), but not when its concentration is low (50 µM). Here we applied the CARS microscopy that provides information on the overall content of lipids in cells, however, obtained in a label-free manner. Figure 5 presents the quantification of CARS signal at 2850 cm−1 intensity for experimental conditions as follows: control (untreated cells), TNF-α treated, D31-PA used in concentrations of 50 µM and 400 µM, without and after pretreatment with TNF-α. The CARS signal based on C-H lipid stretching vibrations shows the highest content of lipids in cells treated with TNF-α followed by the addition of a high concentration of D31-PA. In turn, the HMEC-1 cells treated with an inflammatory agent followed by the addition of D31-PA in low concentration show the content of lipids at the level of control (untreated cells). We anticipated that CARS imaging will provide similar results as extracted from fluorescence images since in both cases the overall lipid content was analyzed. On the other hand, the CARS intensity comes from the intrinsic vibrational modes of lipids (specifically C-H vibration), not from the introduced externally fluorescent label (BODIPY).

Fig. 5

Analysis of CARS images of lipid content in HMEC-1 incubated with D31-PA at 50 µM and 400 µM. The quantification of CARS signal at 2850 cm−1 (a). and representative CARS images (b) of HMEC-1 cells upon D31-PA uptake in normal and pathological conditions (TNF-α treated cells). Values given as mean ± SEM are shown in box plots: mean (horizontal line), SEM (box), minimal and maximal values (whiskers). *p < 0.05

An uptake of D31-PA by untreated and TNF-α pretreated cells was also studied employing SRS (Fig. 6) at 50 µM and 400 µM concentrations, incubated for 24 h. SRS images provided information about the subcellular localization of endogenous FA and exogenous D-PA, which can be distinguished due to deuterium labeling. SRS microscopy allowed imaging lipid and protein and lipid distributions (2930 cm−1), mainly lipids (2850 cm−1), and solely incorporated D31-PA (2110 cm−1). As shown in Fig. 6, there is a very large difference in the lipid content in cells depending on the concentration of D31-PA employed. The uptake of D31-PA is much higher at the higher concentration (400 µM), and this agrees with the Raman and CARS data. At lower concentration of D31-PA, both for control and inflamed cells, the signal from D31-PA is not clearly visible. Pre-stimulation of EC with TNF-α enhanced the uptake of D31-PA used at the higher concentration, while it did not affect the uptake of D31-PA at the lower concentration. It is worth mentioning, that the CARS and SRS techniques differ slightly in terms of the interaction of molecular dipoles with the light pulses phenomena and electron transitions. In the CARS process, the sample interacts consecutively with the electric fields of the pump, Stokes and probe (often replaced by a replica of the pump field itself), there is the interaction of 3 photons, and the generation of a new photon of the anti-Stokes frequency (ωas = 2ωp – ωs) takes place. On the other hand, in the SRS, the measured signal is a slight modification of the beam intensity (SRG on the Stokes beam or SRL on the pump beam). CARS microscopy was used to study the intercellular distribution of lipids in a large group of cells as this technique is well suited for selective imaging of lipids.

Fig. 6

Analysis of SRS images of lipid content in HMEC-1 incubated with D31-PA at 50 µM and 400 µM. Representative SRS images of HMEC-1 cells upon D31-PA uptake in normal and pathological conditions (TNF-α treated cells)

Uptake of saturated fatty acids decreases the inflammation in endothelial cells

To verify whether D-FA stimulate ICAM-1 expression in HMEC-1 cells, fluorescence microscopy was applied. All experimental groups of HMEC-1 treated with D27-MA or D31-PA for 24 h, for control cells and the one preincubated with TNF-α, were tested on the ICAM-1 expression and lipid content (BODIPY staining). The representative fluorescence images are presented in Figs. 7a and 8a, followed by quantitative analysis of the ICAM-1 surface expression (Figs. 7b and 8b), as well as the analysis of the area covered by lipids (Figs. 7c and 8c) for cells treated with both D-FA at low and high concentrations, respectively.

Fig. 7

Analysis of inflammation and lipids contribution in HMEC-1 cells incubated with a low concentration of D-FA (50 µM). Representative fluorescence images of HMEC-1 cells treated with D-FA (a) showed surface expression of ICAM-1 (red areas), lipids distribution (green areas), and the number of nuclei (blue areas) together with calculated average fluorescence intensity of ICAM-1 (b) and area covered by lipids presented as a % of control cells (c). Values given as mean ± SEM are shown in box plots: mean (horizontal line), SEM (box), minimal and maximal values (whiskers). *p < 0.05

Fig. 8

Analysis of inflammation and lipids content in HMEC-1 incubated with D-FA at 400 µM. Representative fluorescence images of HMEC-1 cells (a) showed surface expression of ICAM-1 (red areas), lipids (green areas), and nuclei (blue areas) together with calculated statistics of ICAM-1 fluorescence intensity (b) and area covered by lipids presented as a % of control cells (c). Values given as mean ± SEM are shown in box plots: mean (horizontal line), SEM (box), minimal and maximal values (whiskers). *p < 0.05

The analysis of ICAM-1 expression via fluorescence microscopy (Figs. 7a, b and 8a, b) showed that TNF-α evoked the inflammation of EC [21]. Inflammation caused by TNF-α is mediated by the TNFR1 receptor followed by a series of changes in the cell leading to the release of the nuclear factor (NF-kβ) and AP-1 responsible for the initiation of the inflammatory response [54]. The influence of FA on the endothelium is a complex process and is still under investigation. PA is generally considered as lipotoxic FA, which can contribute to inflammation due to induction of interleukin IL-6 [55] and in certain concentrations may lead to cell autophagy [54]. On the other hand, W. Li et al. [56] have shown that TNF-α stimulates PA transcytosis across the endothelial barrier. MA, on the other hand, is one of the FA known for the acylation of cellular proteins and does not inhibit cell growth or cause cytotoxicity [57].

In our study, D27-MA and D31-PA, independently from used concentration, did not cause inflammation of EC (Figs. 7b and 8b). In the pairwise comparison (D-FA vs. D−FA + TNF-α) the effect of inflammation was clearly seen, however inflammation induced by TNF-α did not seem to be enhanced by D-FA.

The changes in lipid content in EC due to the uptake of D-FA were investigated based on BODIPY staining (Figs. 7a and 8a) and calculations of the cell area covered by lipids (Figs. 7c and 8c). EC incubated with TNF-α show a higher content of lipids in a form of newly synthesized LD, as previously reported [1]. In each pair, D-FA vs. D−FA + TNF-α, some differences between both D-FA in a concentration-dependent manner can be noticed. For a low concentration of both D-FA, a similar area of cells was covered by lipids. The cell area covered by lipids calculated for EC treated with 400 µM concentration of D27-MA is higher when compared to cells treated with D31-PA (Fig. 8c). The results obtained for all samples pretreated with TNF-α showed a higher percentage of cell area covered by lipids. This is most likely caused by the additive effect of TNF-α and D-FA on LD formation. In general, fluorescence microscopy showed that TNF-α is responsible for both the inflammation and in great part for the increase in lipid content in EC. However, using BODIPY staining, no information about D-FA uptake alone was observed. Using fluorescence, the information is given from all lipids in general, both those that are normally present in the cell and those up-taken by cells due to stimulation, while in Raman spectroscopy we infer specifically about the uptake of D-FA.

Myristic and palmitic acids change differently the morphology of the cellular membrane of endothelial cells

Atomic force microscopy (AFM) imaging was used to investigate the morphology of the cells after D-FA uptake in the state of their inflammation and control conditions. Images of the amplitude, phase and topography of representative cells incubated with D-FA after TNF-α stimulation as well as without that treatment are shown in Fig. 9.

Fig. 9

AFM images of HMEC-1 (with and without stimulation with TNF-α) incubated with D-FA at 400 µM for 24 h. Representative images of cells topography, phase, and amplitude (a) The height profile (marked in green line) enabled to define changes in the morphology of cells as shown for example (b) together with the cell size statistics (c)

EC in a normal state and those incubated with D27-MA show a characteristic elongated shape, while the cells treated with D31-PA are more oval and the entire cell stretches out on the substrate. The differences also manifest themselves in the phase images. D31-PA treated cells have a more rugged cell membrane than those from other groups. PA modulates the estrogen receptor alpha variant (ER46), which is located in the plasma membrane, through the nitric oxide synthase (eNOS) pathway in endothelial cells [58, 59] and PA may also lead to cell apoptosis [60]. These two facts may be involved in the changes of the cell shape. The analysis of cell size (Fig. 9b, c) indicates an increase in height of EC treated with D31-PA and both TNF-α and fatty acids probably due to an uptake of D-FA followed by cell inflammation (Fig. 9c). Those findings may suggest a different way of interaction of D-FA with the cell membrane depending on the length of their hydrocarbon chains, which may affect the transport of FA into the cell interior.