Exploring molecular portraits defining lymphovascular invasion in invasive breast cancer patients

BACKGROUND: Lymphovascular invasion (LVI) is a robust clinicopathological parameter of poor prognosis in invasive breast cancer (BC). The accurate detection of tumour emboli within lymphovascular cavities in invasive BC can predict the unfavourable outcome for BC patients, including distant metastasis, recurrence, and short survival. The molecular drivers of LVI in invasive BC patients remain poorly defined. We sought to determine if specific transcriptomic and proteomic profiles regulate the LVI mechanisms in invasive BC. METHODOLOGIES: Meticulous detection of LVI occurrence was performed with histological and immunohistochemical staining on invasive BC full-face tissue sections to construct the LVI-supervised (LVI+/-) cohort, (n=170). All LVI- patients in this cohort were lymph node negative (LN-). The extraction of total RNA from the LVI+/- specimens followed by RNA microarray experiment and data analysis, utilising the linear models for microarray data (LIMMA) package for differential expression, yielded a large cluster of differentially expressed genes (DEG) associated with LVI, (n=1154). The last DEGs were validated against an external cohort of invasive BC patients under the same experimental conditions. We aimed to focus and prioritise our selection of LVI candidates through the statistical assessment of copy number aberrations (CNA) focusing on the top DEGs, (n=100), associated significantly with LVI according to the LIMMA analysis. CNA gain ratios were also calculated for each DEG in the LVI+ and LVI- categories. To avoid the dismissing of crucial LVI-drivers, an exhaustive and investigative approach was also applied and included all of the 1154 DEGs to identify the upregulated gene signatures associated with the LVI+ and LVI- cases in the LVI+/- cohort. Clinicopathological relevance and LVI associations were assessed for the identified LVI-DEGs in the METABRIC cohort, (n=1980), a subset of METABRIC cohort, including LVI-/LN+/- cases, (n=296), and the LVI+/- cohort, (n=170). Fewer LVI-candidates were selected for protein studies. Immunohistochemistry (IHC) protocols were performed to assess the altered protein expression for the qualified LVI-DEGs on tissue microarray (TMA) sections derived from the Nottingham primary series (NPS) of invasive BC, (n=2497). The following experiment aimed to identify the LVI differentially expressed proteins (DEP) utilising high-performance liquid chromatography/mass spectrometry (HPLC/MS) as a high-throughput screening technique. An LVI-supervised cohort consisted of invasive BC luminal-A phenotype was constructed, (n=80). The LVI status for these patients was defined histologically and immunohistochemically, and it comprised LVI+, (n=41), and LVI-, (n=39) cases. These specimens were used for total protein extraction, digestion, purification, and injected in Eksigent ekspert nano LC 425 system, then, analysed on a Sciex TripleTOF 6600 mass spectrometer. RESULTS: The CNA/CNA gain ratios analyses revealed a promising cluster of LVI-DEGs, (n=12). These LVI-DEGs are: MCFD2 (p-value < 0.02, CNA-gain = 4.56); CCT8 (p-value < 0.03, CNA-gain = 2.73); CUL4A (p-value < 0.002, CNA-gain = 1.59); ASCC3 (p-value < 0.01, CNA-gain = 1.36); DDX59 (p-value < 0.0001, CNA-gain = 1.11); ARFGEF1 (p-value < 1.0E-9, CNA-gain = 1.07); CD46 (p-value < 0.01, CNA-gain = 1.03); HRSP12 (p-value < 0.00008, CNA-gain = 0.91); MED23 (p-value < 0.008, CNA-gain = 0.91); EDEM (p-value < 0.01, CNA-gain = 0.86); FAM103A1 (p-value < 0.007, CNA-gain = 0.30); and PPP2CB (p-value < 0.0007, CNA-gain = 0.18). Exploratory bioinformatics search identified that SNPA23 is a co-expressed gene with MCFD2 and CCT8, the top LVI-candidates in the CNA analysis. The associations between MCFD2, CCT8, and SNAP23 against LVI were interrogated on the protein level with IHC. The exhaustive transcriptomic/LVI analyses has also discovered two unique clusters of LVI-DEGs. The first cluster included upregulated transcripts in the LVI+ category, (UP-13/LVI+). These transcripts are: PPM1G (p-value < 0.0001), SEC14L1 (p-value < 0.005), P4HA2 (p-value < 0.007), CTSD (p-value < 0.009), ITPRIP (p-value < 0.016), ZDHHC9 (p-value < 0.031), PILRA (p-value < 0.031), PDLIM7 (p-value < 0.032), RNASE1 (p-value < 0.032), SLC3A2 (p-value < 0.032), IKZF3 (p-value < 0.036), SIGLEC1 (p-value < 0.039), and SLC6A9 (p-value < 0.047). The second cluster included upregulated transcripts in the LVI- category, (UP-22/LVI-). These transcripts are: LPAR1 (p-value < 0.0002), CXCL12 (p-value < 0.001), ELP5 (p-value < 0.002), ECM2 (p-value < 0.004), APCDD1 (p-value < 0.005), FBLN1 (p-value < 0.005), ANXA5 (p-value < 0.008), INKA1 (p-value < 0.011), TMEM98 (p-value < 0.012), FEZ1 (p-value < 0.016), THAP4 (p-value < 0.016), TNFRSF19 (p-value < 0.016), NBR1 (p-value < 0.018), WISP2 (p-value < 0.019), CNRIP (p-value < 0.020), ITSN1 (p-value < 0.021), C1S (p-value < 0.022), TNC (p-value < 0.028), PRRX1 (p-value < 0.030), STAT3 (p-value < 0.035), TSC1 (p-value < 0.036), and TWIST2 (p-value < 0.047). The association between SEC14L1 and LVI was investigated on the protein level with IHC. The cytosolic protein expressions of MCFD2 and SEC14L1, the nuclear protein expression of CCT8, and the membranous protein expression of SNAP23 illustrated strong associations with LVI in invasive BC. Cytosolic MCFD2 and nuclear CCT8 exhibited qualities of LVI inhibitors, (p-value < 0.003) and (p-value < 0.010) respectively. On the other hand, cytosolic SEC14L1 and membranous SNAP23 displayed strong associations with actual LVI occurrences, (p-value < 0.0003) and (p-value < 0.020) correspondingly. The HPLC/MS differential protein expression identified 370 DEP associated with LVI. Among these DEPs, we identified 56 DEP overlapped with the identifiers of DEGs produced by the LVI/RNA microarray experiment. These DEPs are: ACTB, ACTG1, ANP32A, ARHGDIA, ARPC5, CCT8, COL6A2, COMT, DBI, DDX17, DYNLRB1, EIF4E, ERH, FBLN1, H2AFY, HMGB2, HNRNPA2B1, HNRNPH3, HNRNPK, HSP90AA1, HSPA4, ILF3, KCTD12, LGALS3, LMNA, OLA1, PGAM1, PPIA, PRRC1, PSMA6, PTGES3, PTMA, RAC1, RBBP4, RPL10A, RPL7, RPL9, RPLP1, RPS11, RPS18, RPS2, RPS28, RPS4X, RPS7, RPSA, SAR1A, SEC61B, SNRPE, SRSF3, SUB1, TPI1, TUBB6, UFM1, UPF1, VDAC2, and YWHAE. As the flow of LVI-triggering genetic codes should be transcripted first into mRNA then translated into functional proteins, we utilised the METABRIC transcriptomic data and applied Mann-Whitney test for median comparison to confirm the correlations between LVI and the 56 LVI-DEP on the mRNA level. We identified a cluster of DEGs/DEPs according to the RNA microarray and HPLC/MS experiments. These LVI- DEGs/DEPs are: HSP90AA1 (p-value < 0.0001), ARHGDIA (p-value < 0.004), FBLN1 (p-value < 0.005), DYNLRB1 (p-value < 0.018), VDAC2 (p-value < 0.030), RPS18 (p-value < 0.037), RPL10A (p-value < 0.046), and PPIA (p-value < 0.047). CONCLUSION: To conclude, the LVI initiation or repression in invasive BC may be attributed to altered expression of a cluster of genes and proteins. On the transcriptomic level, the LVI definite incidences were associated with the UP-13/LVI+ cluster of mRNAs that were preferentially upregulated in the luminal-B, HER2-enriched, and basal-like intrinsic subtypes. The proteins encoded by these mRNAs are involved in several pathways including intracellular protein trafficking, adaptive immune system, signalling by nuclear receptors, transmembrane glycoprotein interactions, extracellular matrix organisation and degradation, Na+/Cl- dependent neurotransmitter transporters, SLC-mediated transmembrane transport. In contrast, the absence of LVI was associated with a different cluster of mRNAs which were overexpressed in the vast majority of normal-like and luminal-A tumours. This explorative study of LVI provided novel LVI-signature in invasive BC with relevant prognostic utilities. In the future, elaborated and well-designed functional studies may reveal better insights on the contributions for each identified LVI-candidate in driving or inhibiting LVI mechanisms in invasive BC patients