The reagents isatin (isa) and 2-aminothiophenol (tp), in a proportion of 1:1 molar, were dissolved in ca. 50 mL of 96% EtOH. The resulting mixture was refluxed under a nitrogen atmosphere for 2 h, forming the ligands isatp {(E)-3-[(2-mercaptophenyl)imino]indolin-2-one} or nisatp {(E)-3-[(2-mercaptophenyl)imino]-5-nitro-3H-indol-2-ol}. After cooling the reaction mixture, an excess of CuCl₂·2H₂O dissolved in 96% ethanol was introduced, and the mixture was agitated for 10 min. After 2 days, the precipitated complex, [Cu(isatp)Cl], was filtered under low pressure, washed with cold ethanol, and stored in a desiccator under vacuum. Analogously, [Cu(nisatp)H₂O] was synthesized by the same procedure using 5-nitroisatin (nisa) as the starting reagent. These complexes were characterized by UV-Vis, IR, and EPR spectroscopies and had their reactivity toward HeLa cells estimated.

[Cu(isatp)Cl]·1.5H₂O: 1.: (C₁₄H₉N₂OSClCu·1.5H₂O, FW = 379.33 g/mol): Yield 45%, Elemental Anal.: Found (Calc): C: 44.36 (44.33); H: 3.01 (3.19); N: 7.37 (7.38). Matrix free MALDI-TOF MS Spectrum: m/z = 569.004 [ML-S-S-L]⁺ and 316.949 [ML]⁺. After 24 h treatment, at 37°C, for HeLa cells: IC₅₀ = 88.7 ± 1.0 µM, and for fibroblasts: >100 µM.

[Cu(nisatp)H₂O]∙0.5C₂H₅OH 2.: (C₁₄H₇N₃O₃SCu·0.5C₂H₆O, FW = 401.90 g/mol): Yield 48%, IC₅₀ = 68.4 ± 1.0 µM, and for fibroblasts: >100 µM. Elemental Analysis.: Found (Calc): C: 47.24 (47.81); H: 2.55 (3.01); N: 10.58 (10.45). After 24 h treatment, at 37°C, for HeLa cells.

The complexes were isolated in the solid phase in different tautomeric forms, depending on the pH adjusted during the synthesis: [Cu(isatp)] was obtained in the keto form, and [Cu(nisatp)] in the enol form. In both, the thiol group is ionized, as shown in Figure 1.

The chosen complexes [Cu(isatp)Cl] and [Cu(nisatp)H₂O] were not the most active in the studied series since we wanted a reactive compound that damages the cells but does not destroy them (Cerchiaro et al., 2005; Mamián-López et al., 2021). Along this text for clarity, these complexes are named [Cu(isatp)] and [Cu(nisatp)].

Semi-spherical nanoparticles were synthesized according to the Lee-Miesel procedure (Lee and Meisel, 1982). Briefly, 97 mL of a 0.626 mmol L⁻¹ HAuCl₄.3H₂O solution was brought to a boil with vigorous stirring. After thermal equilibrium was reached, 3 mL of a hot 1% (w/v) sodium citrate solution was added to the reaction mixture, and a red colloidal dispersion was obtained. The resulting colloid had an extinction spectrum with maxima at ca. 524 nm, indicating the formation of gold nanoparticles at a size of approximately 26 nm of the gold nanoparticles having citrate as stabilized ligand, Au(cit) NPs (Haiss et al., 2007). The extinction spectrum and the confirmatory transmission electronic microscopy (TEM) images are shown in Supplementary Figure S1.

It was not possible to obtain the Raman spectra of the complexes for all visible and near IR radiations (1,064 nm) due to the great luminescence presented in all spectra.

SERS spectra were registered using a 1 mmol L⁻¹ solution of the complexes dissolved in DMSO:water (1:50) solution. 100 μL of the complexes solution and 400 µL of Au(cit)NPs colloidal dispersion were mixed, and the SERS spectra were obtained. SERS and Raman spectra were recorded in a Renishaw InVia Confocal Raman Microscope spectrometer using a quartz cuvette with laser radiation at λ ₀ = 785 nm (SERS spectra) or 532 nm (Raman spectra) and a diffraction grating of 1,200 lines/mm.

HeLa cells (cervical tumor cells, CCL-2, obtained from ATCC) were incubated on small quartz Petri dishes with DMEM (Dulbecco’s Modified Eagle’s Medium) containing 10% FBS (Fetal Bovine Serum), at 37°C, in a 5% CO₂ atmosphere. After 24 h, the culture medium was removed, and the adhered cells were washed three times with PBS (Phosphate Buffer Saline), maintained under PBS solution, and used as control cells. Treated cells were incubated for another 16 h with the respective assemblies, at 37°C, then washed three times with PBS and maintained under PBS solution during the imaging procedure. Previous HeLa cell viability data, in the presence of studied copper complexes, were evaluated by the tetrazolium dye reduction assay (MTT), compared to non-tumorigenic human foreskin fibroblasts P4 cells (see Supplementary Figure S2). These fibroblasts P4 cells were isolated from tissues at Faculty of Pharmaceutical Sciences-USP in the laboratory of Professor Silvya Stuchi Maria-Engler (Kim et al., 2013). Intake of copper ions in Hela cells after 6 h and 24 h with complexes 1 and 2, in comparison with aqua-copper ions, [Cu(H 2 O) 4 ] 2+, at 37°C is shown at Supplementary Figure S3. The determination of intracellular copper concentrations was done by ESI-MS measurements.

For image acquisition, the culture medium was removed and substituted with a fresh PBS solution. A Renishaw InVia Confocal Raman Microscope spectrometer was used with excitation at λ ₀ = 785 nm for SERS imaging, with the diffraction grating centered at 1,340 cm⁻¹, and λ ₀ = 532 nm for Raman imaging, with the diffraction grating centered at 3,020 cm⁻¹. The objective used was a water immersion Leica ×63 with a N.A of 0.90. The airy disk radius (r) was calculated as r = 0.61x (wavelength/N.A). For 785 nm the airy disk has a diameter of 1.064 µm, and for 532 nm, 0.721 µm. Thus, a 1 µm pixel is enough to resolve the Raman mappings.

First, the SERS images were obtained. After this procedure, Raman images of the same cells were acquired. The spatial resolution between each pixel was 1 µm and the integration time for each spectrum was 0.5 s. This two-radiation strategy was adopted due to the complementary information they supplied: the first radiation source allows detecting SERS intensification from the Raman reporter, while the latter one allows collecting information from lipids, proteins, and water vibration modes, which is useful for building chemical maps from specific cells regions, as the membrane, nuclei, and water content, for example, as widely described in the literature (Talari et al., 2015; Li et al., 2019). Besides, irradiating with λ ₀ = 532 in the presence of AuNPs could generate a luminescence background in the spectra (Dilruba and Kalayda, 2016).

All spectra processing and image building were performed in Matlab (MATLAB 9.11.0.1837725 (R2021b) Update 2). PCA analyses and preprocessing were performed on PLS Toolbox (Eigenvector Inc.). The MCR-ALS method was run using the MCR Toolbox. (Jaumot et al., 2015; de Juan and Tauler, 2021).

The image background was removed using a digital mask based on a threshold value statistically chosen from the raw chemical maps, in a way that those pixels with relative abundance values than it were kept, and the remaining were filled with zeros. Due to the natural variability found in each cell, the threshold is necessarily different for each one and its choice was automated within the digital mask command lines.

Datasets were arranged as matrices, spikes (instrumental noise) were manually removed, and the baseline was corrected using an Asymmetric Least Squares procedure with a polynomial of order: 2, λ = 10,000, p = 0.001.

Different cell samples were investigated: control samples (untreated), samples (treated with [Cu(isatp)], [Cu(nisatp)], Au(cit) NPs, Au(cit) NPs with attached DTNB, Au(cit) (DTNB) NPs and finally Au(cit) (DTNB) NPs attached with one of the complexes, giving Au(cit) (DTNB) (Cu(isatp) and Au (cit) (DTNB) (Cu(nisatp). Firstly, an exploratory analysis was applied for each of the seven different groups of samples separately to estimate a preliminary number of principal components and investigate intragroup variability. Before this procedure, spectra were mean-centered.

Before running the MCR method, Singular Value Decomposition (SVD) and SIMPLe-to-use Interactive Self Modeling Analysis (SIMPLISMA) were applied to determine the number of “purest” components in each dataset and the profile of the spectral factors. For calculating this, MCR-ALS was constrained with no negativity in spectra and concentrations and using 1.0 as the convergence criterion. The calculation stopped when no significant variation was observed in the model for two consecutive iterations.

Raman and SERS images were built with the relative concentration matrix calculated by MCR-ALS. These values were normalized to the maximum value and reshaped to the xy dimension used during the spectroscopic mapping, where x and y are the numbers of pixels.

The interaction between the surface of Au (cit) NPs and the [Cu(nisatp)] and [Cu(isatp)] complexes were investigated through their respective SERS spectra. Supplementary Figure S4; Supplementary Table S1 presents a tentative vibrational bands assignment.

Despite that the SERS spectra of the oxindolimine copper(II) complexes with a moderate signal-to-noise ratio could be acquired, this scenario changes when the system Au(cit)-complex NPs requires to be incubated with cell culture for periods no longer than 16 h, at high concentrations. These complexes easily induce nanoparticle aggregation in the aqueous medium with time, and in consequence, a significant lowering in the signal intensity is observed. Therefore, this undesirable effect was avoided using complex concentrations lower than the corresponding IC₅₀ values (68.89 ± 3.31 and 93.95 ± 5.78 µmolL⁻¹ for [Cu(nisatp)] 2 and [Cu(isatp)] 1, respectively). Also, to overcome those disadvantages, a Raman reporter, which has a high scattering cross-section, was used to obtain SERS images of living cells.

The SERS spectrum of 5,5′-dithiobis-2-nitrobenzoic acid, DTNB, chosen as the Raman reporter in the colloidal solution of Au(cit) NPs, is depicted in Figure 2. The DTNB is widely used in protein quantification through UV-Vis spectroscopy, and one of its main structural characteristics is the presence of a disulfide bridge that can be chemically cleaved, for example, in the presence of sodium borohydride in an alkaline medium (Wang et al., 2015). Similarly, upon interaction with the gold surface, its high affinity towards the sulfur atom can break the -S-S chemical bond, allowing the AuNPs functionalization, as shown in Scheme 1.

The SERS spectrum of DTNB shows a main band located at 1,340 cm⁻¹, corresponding to the ν_s (NO₂) mode, chosen as the marker mode for the cell mappings. In addition, as observed in Supplementary Figure S5, it is worth mentioning the presence of an Au-S band at 322 cm⁻¹, confirming the disulfide bridge cleavage and the formation of a chemical bond between sulfur and gold.

Then, the SERS probe was completed by the adsorption of the complexes on the Au(cit) (DTNB) NPs. The SERS spectrum of the whole nanostructured architecture was acquired in the colloidal aqueous medium, as shown in Figure 2. It is noticeable that the ν_s (NO₂) band at 1,340 cm⁻¹ due to DTNB remains in both Au(cit) (DTNB) (complex) NPs spectra and, in addition, a band with contribution from the imine modes [ν_s (NC) + ν(CO)] at 1,393 cm⁻¹ appears, attesting the presence of isatp and nisatp ligands.

A closer view of the spectral region near the strong band of νs (NO₂) at 1,340 cm⁻¹ due to DTNB shows slight differences in shape and frequency values for each SERS probe. For Au(cit) (DTNB) (Cu(isatp) NPs at frequencies higher than 1,340 cm⁻¹, a band is observed at 1,393 cm⁻¹, assigned as the imine. In the case of the Au(cit) (DTNB) [Cu(nisatp)] NPs, the spectrum also shows this band with a lower intensity and at lower frequencies from the strong band of the DTNB νs (NO₂) mode, a shoulder at 1,323 cm⁻¹ is observed (Supplementary Figure S5), which could be assigned to νs (NO₂) located at the nisa moiety. Besides, the absence of wavenumber shifts in these spectra suggests that the metal complexes and the DTNB do not interact chemically. The findings at this stage indicate that the DTNB choice is suitable for SERS imaging of living cells, and a deeper understanding of the spectral behavior will be discussed in the sections ahead.

Two main sets of experiments (in the absence and presence of Au NPs) were employed to study the effect of the metallodrugs in the intracellular environment, as detailed in Table 1. In the first set, cells were treated with either [Cu(isatp)] and [Cu(nisatp)] to evaluate their effect in the absence of Au(cit) NPs. In the second set, Au(cit) NPs were present in all cases, with their assemblies modified as indicated. Also, Raman spectra were acquired for cells without any treatment as a control experiment. Therefore, the discussion will rely on the chemical images built from mappings registered under 785 (for acquiring SERS signals) and 532 nm radiation (for acquiring Raman signals from lipids, proteins, and water).

A preliminary data mining procedure was performed on the Raman spectra λ ₀ = 532 nm of each group of cells, separately, through PCA analysis, to identify the main intercellular differences upon each treatment applied. The standard deviation (SD) for the variance explained by each principal component (PC) was calculated to estimate this. The results depicted in Figure 3 give an initial hint about the extent of damage suffered by the cell groups D, E, F, and G compared to control cell group A. It is expected to observe a certain degree of variability due to the natural characteristics of cells, i.e., intracellular processes at different stages during the mapping acquisition, as seen in A cell’s subplot. This behavior is similar for the D and E groups, where the cells are in the presence of Au(cit) NPs or Au(cit) (DTNB) NPs, respectively. For this last group, though, it seems that for components 2 and 3, there is a slight increase in variability, suggesting that the presence of Au(cit) (DTNB) NPs could cause a different response in the cells when compared to Au(cit) NPs alone or control cells.

For F cells containing Au(cit) (DTNB) [Cu(isatp)] NPs, it is interesting to notice a higher intracellular variability for the second component. The % of explained variance remains constant from the third component, contrary to the behavior of A, D, and E groups, where components 1 to 3 store a significant amount of explained variance, which could indicate that at least for a fraction of the cells, chemical changes are occurring that could be observed through the calculated spectral profile. Finally, the remarkable difference among the cells of group G is clearly showing that the Au(cit) (DTNB) [Cu(nisatp)] NPs system is causing chemical modifications observable through the Raman spectra.

As Raman signals are highly specific and, in the case of living cells, can offer fingerprints of their chemical components, the first approach was performed through a curve resolution method, MCR, to separate the most important fingerprints in the datasets.

Three main MCR factors or spectral profiles could be resolved for the control experiments spectra (group A, cells without any treatment), whose spatial distribution is depicted in Figures 4E–G. Component 1 represents the OH symmetric and asymmetric stretching modes from water content (Figure 4B). It is noticeable, in image 4E, that its distribution, as expected, is higher close to the internal part of the membrane. Component 2 (Figure 4C) is related to regions with high protein content, and it is distributed along the entire cell but concentrated in the nucleus/nucleolus region, as seen in Figure 4F. Finally, the lipid-rich regions are described in component 3 (Figure 4D), located mainly in the cytoplasm, image 4G. The optical image (Figure 4A) confirms the location of these regions. The chemical images and spectra for cells upon treatment will be analyzed using this cells group as reference, used as a control experiment in our study.

The Raman images, obtained by using 532 nm exciting radiation, for both B and C groups of cells, treated with [Cu(isatp)] or [Cu(nisatp)], showed no damage to the cells’ structure, also confirmed by their respective optical images, as can be observed in Supplementary Figures S7–S16.

When the cells were incubated with Au(cit) NPs (group D), a noticeable signal enhancement was observed for several bands, mainly in the region between 1,200 and 1,500 cm⁻¹. This SERS enhancement occurs due to Au NPs and the proper exciting radiation at 785 nm. In addition, vibrational modes from the cytoplasm, corresponding to lipids and proteins, as well as characteristic bands assigned to nucleic acids such as guanine and adenine, were identified. An example of this enhancement in the intracellular environment is shown in Supplementary Figure S6, and a more detailed band assignment can be found in Supplementary Table S2. Also, notice the signal enhancement obtained compared with the spectra from A cells, where Au(cit) NPs are absent.

Again, the chemical image obtained under 532 nm radiation showed that the membrane structure seems to be without damage; changes in the spectra between 3,000 and 3,450 cm⁻¹ suggested an increase in the intracellular water content, which could mean a slight membrane permeation induced by the AuNPs. Other regions inside the cell show no apparent damage, as illustrated in the chemical maps for D cell in Figure 5.

For the E group, containing the DTNB assembled on the Au(cit) NPs, it was verified through both optical and chemical images that the location of Au(cit) (DTNB) NPs through the intensified ν_s (NO₂) band match inside the cell, confirming that, firstly, Au(cit) (DTNB) NPs can enter the cell and secondly, the DTNB remains attached to the metallic surface. However, the profile of the MCR component describing regions with high lipids concentration appears modified regarding the A cells group, suggesting that lipids content in cells in the presence of DTNB was affected. At the same time, the intracellular water content increased, as shown in Figure 6, and the image shows that the cellular wall was severely damaged. Besides, the component 2 distribution shows unusually big protein agglomerates, suggesting that the nuclei content was scattered inside the cell, mainly in the cytoplasm. Cells in the same E group showed similar behavior; even for some of them (Supplementary Figure S13), the MCR method could not resolve the lipidic component, probably due to modifications on the cell or influence of the DTNB attached to the Au(cit) NPs on the spectra themselves.

The optical and chemical maps analysis allowed identifying remarkable characteristics upon the cell’s incubation with the metallodrugs. The first one was related to the number of MCR factors resolved as describing the whole variability in the cell’s spectra. Contrary to all the other cell groups studied, the number of spectral factors resolved by the MCR method was only two, with a clear absence of the lipids factor. This result matches with a second observation where the optical image shows that the Au(cit) (DTNB) (metallodrug) NPs are located preferentially in lipid-rich regions, suggesting that lipids could be at least one of the main targets during the metallodrug mechanism of action. This is shown in Figure 7 for cells from group F [cells with Au(cit) (DTNB) (Cu(nisatp)] NPs.

Another important effect of this treatment was related to the changes found in the protein distribution, i.e., the factor 2 from MCR. Their distribution appears highly dispersed inside the cell, indicating possible damage and leaking of the nucleus content. It is also interesting that the water pattern for cells F and G [treated with Au(cit) (DTNB) (Cu(nisatp) and Au(cit) (DTNB) (Cu(isatp) NPs respectively] is comparable to the control cells, which means that the assembly Au(cit) (DTNB) (metallodrug) NPs does not break the membrane structure to the point of allowing entering water through it. The same behavior is observed in both images and spectral MCR profiles in Figure 8, corresponding to a G cell Au(cit) (DTNB[(Cu(isatp)].

More chemical images and MCR profiles obtained for other cells from groups A to G are included in the supplementary information, Supplementary Figures S7–S16.

Regarding the SERS spectra in the region between 1,200 and 1,400 cm⁻¹ where the ν_s (NO₂) mode is observed, interesting differences were found upon application of a curve resolution method MCR-ALS (run on the same number of spectra from groups treated with Au(cit) (DTNB), Au(cit) (DTNB) (Cu(isatp) and Au(cit) (DTNB) (Cu(nisatp) NPs. This was done in order to confirm that the SERS probe (Figure 9G) reached the inside of the cells. When the Au(cit) (DTNB) NPs is inside the intracellular environment, the NO₂ band shows that it can be decomposed into two main features centered at 1,333 and 1,340 cm⁻¹, as seen in Figure 9A. The same processing applied to groups F and G, showed that when the [Cu(nisatp)] complex (Figure 9H) is included, there is a 7 cm⁻¹ shift for both components to lower frequency values (Figure 9C). No frequency shifts were observed for the [Cu(isatp] complex (Figure 9B), although factor 2 was predominant. These observations were corroborated by comparison to the behavior of the relative concentrations (Calculated by MCR-ALS) for each factor inside the cells (Figures 9D–F), where these values can be understood as the abundance of each factor in each spectrum. As shown, Factor 1 in the cells containing Au(cit) (DTNB) (Cu(isatp) NPs, one of the NO₂ modes, seems not to be enhanced, while for Au(cit) (DTNB) and Au(DTNB) (Cu(nisatp) NPs the abundance of the two factors in each spectrum is comparable. This result reinforces that the complexes remained attached to the metal surface, so the SERS probe is stable inside the cells.