What if one optical workflow could both map HER2 on an individual cancer cell and read the cell’s broader metabolic state—and then test what changes when HER2 is silenced? Breast cancer is a heterogeneous neoplasm characterized by several subtypes, so patients with apparently similar disease can still differ biologically.

One aggressive form with high metastasis rates presents overexpression of the human epidermal growth factor receptor 2, or HER2. A quantitative evaluation of HER2 levels is essential for a correct diagnosis, selection of the most appropriate therapeutic strategy, and monitoring the response to therapy.

That makes accurate measurement more than a descriptive exercise. The HER2-positive subtype accounts for twenty to thirty percent of all invasive breast cancers. It is characterized by remarkably high HER2 expression: from five hundred thousand to more than two million molecules per cell, compared with twenty-five thousand to one hundred eighty-five thousand in normal epithelial cells or non-amplified tumours.

HER2 overexpression or amplification is associated with unrestricted tumour growth, greater aggressiveness, a higher metastasising rate, and resistance to chemotherapy, influencing patient management. Quantifying HER2 positivity is therefore crucial for a correct and early diagnosis.

Routine methods include immunohistochemistry, fluorescence in situ hybridization, and chromogenic in-situ hybridization. These techniques have several limitations and require tissue biopsies. Methods and technologies for precise HER2 quantification are important for proper breast cancer diagnosis, for guiding the therapeutic strategy, and for predicting response to treatment.

A method able to assess the exact HER2 concentration at single-cell level, together with associated cellular metabolic changes, would benefit studies on cell lines in vitro and diagnosis of HER2-positive breast cancer in vivo. Surface-enhanced Raman spectroscopy, or SERS, has emerged as a tool for detecting and quantifying membrane biomarkers, including HER2, through a nanoparticle-based approach.

It is characterized by high sensitivity and molecular specificity. Unlike conventional Raman spectroscopy, SERS uses the plasmonic properties of noble metal nanoparticles, particularly gold and silver, to amplify Raman scattering signals from analytes adsorbed on their surfaces.

The amplification factor can reach six to ten orders of magnitude, allowing detection of trace amounts of analytes with unprecedented precision. Artificial Intelligence-assisted Raman spectroscopy can provide efficient cell discrimination and identify the specific metabolic signature of cancer cells.

SERS analysis can provide detailed visualization of biomarker distribution on the cell surface. The information obtained by Raman spectroscopy and SERS complements each other, enabling a synergistic enrichment of clinical relevance. The study uses SERS-Raman screening of representative breast cancer cell lines for quantitative identification of HER2 levels and for correlating HER2 expression and membrane exposure with metabolic features associated with malignant transformation.

The same SERS-Raman approach was applied to cell clones in which the HER2 gene was stably silenced using RNA interference technology. This permits investigation of metabolic changes associated with HER2 expression. The association of SERS-Raman spectroscopy with RNA interference technology provides insights into HER2-positive breast cancer biology and mimics the effect of HER2-silencing targeted therapies.

The tumor cell lines selected were SKBR3 and MDA-MB-468 because they exhibit the highest and lowest HER2 expression, respectively, and represent the most severe prognostic scenarios in breast cancer. MCF10A cells, derived from non-tumorigenic breast epithelium, were selected for comparison.

The gold nanoparticles were functionalized sequentially with the Raman reporter and the TZ antibody. Forty-nanometre gold colloidal nanoparticles were combined with borate buffer and then mixed with 4-mercapto-benzoic acid. The solution was mixed by inversion and incubated at four degrees Celsius for ninety minutes.

EDC and NHS were then added to activate the particles’ carboxyl groups. The particles were then immunosensitised by adding TZ antibody and allowing it to react at four degrees Celsius for ninety minutes. After washing to remove non-conjugated antibody, bovine serum albumin was added to promote colloidal stability and block unreacted binding sites.

The TZ-AuNPs were stored at four degrees Celsius. Table one is a reference map linking Raman bands in normal breast epithelium and cancer breast-cancer cells to molecular vibrations. The authors attribute bands to components including proteins, lipids, collagen, nucleic acids, carbohydrates, and carotenoids—for example, the band at one thousand two to one thousand four is associated with phenylalanine and carotenoids, while one thousand eighty-one to one thousand ninety-six includes phosphodiesters and phospholipids.

This matters because it provides the biochemical vocabulary for interpreting Raman spectra from breast tissue and cells. Figure one establishes a HER2-expression panel across six breast cell lines, including non-tumorigenic MCF10A and clinically distinct subtypes.

Western blots measure total HER2 and soluble HER2 in conditioned media, while flow cytometry measures cell-surface HER2 using fluorescence intensity and isotype controls. Together, these readouts show that HER2 abundance can be assessed at the total-protein, released, and membrane-exposed levels, with several comparisons marked as statistically significant at p less than zero point zero zero zero one.

The approach clearly distinguishes HER2-positive SKBR3 cells from triple-negative MDA-MB-468 breast cancer cells and from non-tumorigenic MCF10A breast epithelial cells. TZ-AuNPs assess and quantify HER2 directly on the cell membrane at single-cell level, identifying and discriminating cells on the basis of HER2 expression.

But assessing biochemical composition, tumorigenic mechanisms, and links with tumour aggressiveness and metastasising potential requires additional information. Raman analysis was performed on the same breast cell lines to correlate quantitative HER2 expression from SERS analysis with a more general metabolic and physio-pathological picture.

Raman analysis of single cells used the same microscope-assisted equipment. The study first screened a panel of breast-cancer-derived cell lines for HER2 expression levels by western blot analysis. The panel included SKBR3 and BT474 cells, representative of HER2-positive subtypes, and MCF7 and MDA-MB-468 cells.

Classification used two machine-learning techniques: principal component analysis, or PCA, and linear discriminant analysis, or LDA. PCA decomposed the data matrix into loadings, scores, and variances for each component. The first four principal components accounted for more than ninety percent of cumulative variance.

Their loadings were compared with reference bands for vibrational modes of the main cellular constituents. PCA scores for principal components two through four were plotted in three-dimensional space, and the components best explaining class differences were evaluated with a Wilcoxon non-parametric test.

For LDA, the model used spectra with six inverse-centimetre resolution, trained on fifty percent of the data, and evaluated mean misclassification error over one thousand iterations on a ten-fold random test set with equal class-size distribution. Figure two pairs brightfield images of selected MCF10A, MDA-MB-468, and SKBR3 cells with SERS intensity maps at Raman shifts of one thousand eighty and one thousand five hundred eighty inverse centimeters.

The maps retain pixels exceeding the noise threshold for both bands, while the color scale reports SERS intensity in counts per second. Panel D summarizes the relative HER2 quantification on the cell membranes, showing how the probe-based SERS readout distinguishes these cell lines.

Figure three shows that Raman spectra capture distinct biochemical fingerprints across the three breast-derived cell lines, with marked peaks and biologically relevant spectral bands. In the PCA score plot, the cell lines form separate patterns using principal components two through four, while the PC4 box plots show statistically tested distributions, with asterisks indicating p less than zero point zero zero zero one.

The LDA panels further visualize supervised separation and the reported misclassification rate. The supervised LDA used fifty percent of the data as a training set and fifty percent as a test set, without mixing spectra from a given cell line between the training and test sets.

The analysis achieved one hundred percent overall accuracy in breast-cancer-cell-line classification and a mean misclassification error of less than five percent. Similar results were achieved using only one spectrum per cell and validating it on an independent batch of cells.

RNA interference produced two cell lines, SCR-SKBR3 and the HER2-silenced counterpart, which differ only for HER2 expression. SERS was then applied to differentiate them. The SERS approach provides precise HER2 quantification at single-cell level and the specific spatial distribution of HER2 on the cell membrane.

In SCR-SKBR3 cells, the receptor is highly abundant at the cell boundaries, whereas it is completely absent in HER2-silenced cells. This was documented by a ninety-three percent reduction of SERS signalling. Figure four validates the HER2-silenced model across several measurements.

Western blots show reduced HER2 in the shRNA clones, while flow cytometry confirms altered cell-surface HER2 in clone three; the same clone also shows reduced AKT phosphorylation relative to the scrambled control. SERS maps at one thousand eighty and one thousand five hundred eighty inverse centimeters, together with membrane biomarker quantification, demonstrate that the technique can assess HER2 at the single-cell level and visualize its spatial distribution.

Raman fingerprints were evaluated for thirty cells from each line, with nine spectra from different cytoplasm regions per cell and an integration time of sixty seconds for each spectrum. The two cell lines displayed only a few spectral differences, consistent with the lack of morphological changes.

The most evident differences occurred at bands assigned to phenylalanine, phosphodiesters, lipids, phospholipids, nucleic acids, proteins, pyrimidine rings, heme-containing proteins, and aromatic amino acids. Figure five compares Raman spectra from scrambled SKBR3 cells and HER2-silenced clone three.

Panel A overlays their median interquartile-range spectra with principal-component loadings, highlighting peaks and biological bands including the phospholipid- and nucleic-acid-associated region around one thousand eighty-five to one thousand ninety-five inverse centimeters.

Panel B shows distinct PC4 score distributions with a Wilcoxon p-value below two point two times ten to the power of minus sixteen, while panel C visualizes their separation in PC1–PC4 score space. Quite unexpectedly, HER2-silenced cells exhibit an increase in lipid content.

The proliferation potential related to HER2 overexpression therefore appears dissociated from metabolic changes, especially the lipid counterpart. The lipid content and fatty-acid synthesis may be regulated through pathways likely independent from HER2 expression, because they persist or even increase upon HER2 silencing.

Whether this link is direct or indirect needs to be clarified. Raman analysis currently underscores lipid-metabolism rewiring no longer tied to HER2 expression and aggressive behaviour. The study shows the potential of associating gene silencing with combined SERS-Raman technology to classify breast cancers, discriminate subtypes at single-cell level with high specificity and selectivity, and characterize general and distinctive metabolic changes.

The reliable and consistent results represent a basis for translating the procedure into clinical practice by analysing patients’ biopsies or tumour tissue specimens to support diagnosis or follow response to therapy. The combined SERS-Raman approach identified HER2 at single-cell level while revealing broader metabolic differences, including lipid changes that persisted or increased after HER2 silencing.