Exploring the lymph node’s microenvironment for personalized management of Luminal A breast cancer

Lymph nodes (LNs) are the main doorway for tumor cell metastases from the primary site and its evaluation is a major prognostic factor. The One Step Nucleic Acid Ampli�cation (OSNA) is being adopted worldwide for sentinel-LNs (SLNs) staging in breast cancer (BC). SLNs´ OSNA lysate may be used for gene expression studies, being the potentially ideal samples to search for new markers related to immune response. Using a targeted gene expression approach, we aim to identify transcriptomic patterns of SLNs immune response and biomarkers that may improve risk strati�cation and personalized therapy for patients with Luminal A BC. and with impact in immune response. These DEGs may improve prognosis accuracy and increase the e�cacy and safety of targeted therapies in Luminal A BC patients. As OSNA assay is being implemented for SLNs staging in other cancers, RNA-seq in the OSNA lysate could also have a wider utility.


Introduction
Lymph nodes (LNs) are the main doorway for tumor cell metastases and its evaluation is a major prognostic factor. Two-thirds of breast cancer (BC) patients diagnosed with LNs metastases will develop distant metastases and 73% of these women will be dead within 5 years after diagnosis. (1,2) SLN biopsy is the standard approach for loco-regional staging in patients with clinically T1-T2 invasive BC presenting with a clinically negative axilla. (3,4). Patients with negative SLNs require no further axillary surgery.
(5, 6) Patients with 1 or 2 metastatic SLNs who meet the criteria of ACOSOG Z0011 or AMAROS trials, completion of axillary LN dissection (ALND) is not necessary if irradiation and systemic adjuvant therapy are planned. (3,(7)(8)(9)(10) However, currently, for Luminal node-positive BC, it is recommended an extension of endocrine therapy towards a duration of 10 years based on persistent risks of recurrence among such patients. (5,6,11) Conventional intraoperative histological examinations of SLNs' frozen sections are associated with 10-30% false-negative results for metastases. (12) Nevertheless, despite serial step section examination of each SLN being possible to overcome the false-negative results, it would be impractical because it requires a heavy workload for pathologists. (12) To overcome this issue, a molecular method, the One Step Nucleic Acid Ampli cation (OSNA), based on reverse transcription loop mediated isothermal ampli cation (RT-LAMP) of cytokeratin 19 (CK19) mRNA in the lysate of SLNs, is being adopted worldwide by an increasing number of BC care centers. (13) OSNA has several advantages: allows the analysis of the whole SLN, is semi-quantitative, standardized, reproducible, quicker (30 to 40 minutes from the excision of the SLN) and also diminish the pathologist workload. (12)(13)(14)(15)(16)(17) The OSNA cut-off levels were determined by Tsujimoto et al: macrometastases was de ned as > 5 000 copies/µL of CK19 mRNA, micrometastases as 250 to 5 000 copies/µL and a value < 250 copies/µL correspond to absence of metastases or presence of isolated tumor cells (ITC). (14) Total tumor load (TTL) was de ned as the sum of the total number of CK19 mRNA copies in all positive SLNs (in copies/µL). (15,18) Previous studies revealed that TTL is an independent predictor of the status of the non-sentinel LNs in BC patients and to be independently correlated with disease free survival, local recurrence free survival and overall survival. (13,15,(18)(19)(20) Nevertheless, the exact TTL cut-off to determine ALND is still under debate. (13,18,19,21,22) Although LNs metastases is among the strongest predictors of prognosis, few studies have focused on the assessment of immunoin ammatory response in the LNs and the mechanisms that underlie the local failure of effective anti-tumor immune responses remain poorly understood. In LNs, exposure to tumor-derived factors induces stromal reprogramming, modi es immune cell population dynamics and affects chemokines and interleukins levels, which have the potential to contribute to impaired immune system response. (23) Metastatic LNs are associated with an increased number of plasmacytoid dendritic cells (DCs), regulatory T lymphocytes (Tregs), immature DCs, higher expression of CD163 + M2 macrophages and a lower activation of CD8 + cytotoxic T lymphocytes, suggesting a de cient immune response. (1,24) On the other hand, CD1a DC, mature DC, or CD169 + M1 macrophages are increased in the primary tumor and LNs of patients with non-metastatic LNs, suggesting a more e cient immune response. (1,24) Also, on metastatic SLNs, expression of CD83, IL-12p40, IFN-γ, IL-10, and FOXP3 is higher than in non-metastatic SLNs. (25) Thus, prognosis seems to depend not only on whether a patient has LNs metastases or not, but also on the type of local immune-cell population. (1,24,26) Different microenvironments were recognized in each subtype of BC and, consequently, speci c microenvironments might be associated with distinct behaviors of the tumor cells and distinct prognosis and potential therapeutic targets. (27) Studies on tumor microenvironment (TME) in patients with Luminal [estrogen receptors (ER) positive] HER2 negative BC are scarce, probably due to the good global survival rates (92.5% at 4 years) and the e cacy of the already existing therapies.(28) Nevertheless, in Luminal HER2 negative BC, 31% had LN metastases and, for this group, the survival rates are lower (84.4% at 4 years). (28) Considering that about 73% of BC are Luminal HER2 negative, this demonstrates the enormous untapped potential for immunotargeted therapy in these patients. (28)(29)(30) As in OSNA assay most of the lysate sample is spared, it can be used for gene expression studies, evaluating microenvironment related genes, being the potentially ideal samples to search for new markers related to the immune SLNs response. (31,32) Using a next generation sequencing (NGS) targeted gene expression approach, we aim to identify transcriptomic patterns of immune response, at the level of SLNs, that may improve risk strati cation and the development of targeted therapies for patients with Luminal A early BC.

STUDY DESIGN AND PARTICIPANTS
This study was an investigator's initiative, observational, prospective, pilot study. The project was approved by the Ethics Patients with Luminal A early-stage BC (cT1-T2 N0) were invited to participate. The intrinsic subtype classi cation was based on international guidelines, and it was considered Luminal A BC if ER-positive, HER2-negative, Ki67 < 20% and PR ≥ 20%. (33,34) Patients were enrolled until, consecutively, obtain 16 patients with OSNA negative SLNs and 16 with OSNA positive SLNs. Inclusion criteria were de ned including women with invasive BC, Luminal A subtype, cT1-2, cN0, surgical treatment including SLN biopsy and SLN analyzed by OSNA assay. As exclusion criteria, authors de ned male BC, age under 18 years-old, pregnancy, germinal mutations associated with breast hereditary cancer, neoadjuvant treatment, cytology proven LN metastases, distant metastases, tumors not expressing CK19, patients unable to give informed consent and technical limitations to SLN biopsy.

SLN BIOPSY AND OSNA ASSAY
SLNs were identi ed under combined techniques, using patent blue and radioisotope or superparamagnetic iron oxide, as previously described. (35,36) After identi cation by the surgeon, SLNs were removed and directly sent to Pathology Department.
The detailed OSNA assay has also been previously described. (14,18,37) In the Pathology Department, after the extra nodal tissue being removed, SLNs that exceeded the speci ed maximum weight (600mg) were cut into two or more pieces and processed as separate samples. Then, fresh SLNs were homogenized in 4 ml of a mRNA-stabilizing solution (Lynorhag® solution, Sysmex Corporation) using a RP-10 system (Sysmex Corporation; 90 seconds at 12 000 rpm). The homogenate (1 ml) was centrifuged for 1 minute at 12 200 x g and the intermediate phase was collected. A volume of 20 µl of the intermediate phase was used for the OSNA assay using the LYNOAMP™ CK19 (Sysmex Corporation) on the RD-210 system (Sysmex Corporation). A standard positive control sample and a negative control sample were used in every assay. Lastly, instead of being discarded, the remaining homogenate was kept at -80˚C for RNA sequencing (RNA-seq) analysis.
The OSNA assay results were based on the calculated number of CK19 mRNA copies/µL, in accordance with the previous cut-off levels: > 5 000 copies/µL corresponding to macrometastases (pN1), 250 to 5 000 copies/µL to micrometastases (pN1mi), and values < 250 copies/µL were classi ed as negative SLN. (14) In negative SLNs, using RD-210 system (Sysmex Corporation), 160 to 249 corresponded to ITCs [pN0(i+)] and < 160 to absence of metastases (pN0).(38) TTL is de ned as the sum of the numbers of CK19 mRNA copies in all positive SLNs. (13,18) The study had two major branches for SLN microenvironment analysis: 16 patients with OSNA positive SLNs and 16 patients with OSNA negative pN0 SLNs. The OSNA positive group was subdivided in SLNs with micrometastases and SLNs with macrometastases. Whenever more than one sample of the SLN or more than one SLN were diagnosed as having metastases, the one with a higher number of copies of CK19 mRNA/µL was considered for gene expression studies.

ALND: NON-SENTINEL LNs
The decision to proceed with ALND was discussed in a multidisciplinary team for each patient. Typically, ALND was performed in patients with metastatic SLNs with > 15 000 copies/µL of CK19 mRNA or if 3 or more positive SLNs were detected. (4,5,18)

Clinicopathologic results
The clinical features of the 32 patients with Luminal A invasive BC included in this study are presented in Table 1, comparing the 16 patients with OSNA negative SLNs (pN0) and the 16 patients with OSNA positive SLNs (pN1 and pN1mi). There were no statistically signi cant differences related to clinical characteristics (Table 1). Regarding the histological characteristics of the tumor, the majority were No Special Type (NST) and had a single tumoral focus ( Table 2). The OSNA positive group had a higher percentage of LVI, a higher grade and a higher Ki67 (Table 2). There were no statistically signi cant differences concerning the TILs (Table 2). The SLNs were identi ed under combined techniques and the number of removed SLNs were similar in both groups (Table 3). In OSNA positive group, the mean number of metastatic SLNs was 1.1 ± 0.3 (Table 3 and Additional le 2). Micrometastases were found in 43.8% (n = 7) and macrometastases in 56.2% (n = 9) ( Table 3 and Additional le 2). In metastatic SLNs, the mean TTL was 121 238.1 ± 213 294.7. In SLNs with micrometastases the mean TTL was 1 394.3 ± 1 750.0 and in SLNs with macrometastases the mean TTL was 214 450 ± 250 915.2 (p < 0.001). ALND was performed in 6 patients (37.5%) and 4 out of 6 patients had metastases in non-sentinel LNs (Table 3). Among patients submitted to ALND, the mean total number of metastatic LNs (sentinel and non-sentinel) was 2.7 ± 0.9 (minimum = 1; maximum = 7) and macrometastatic LNs was 2.   The result of PCA for the two principal components, including all transcripts and all samples, is shown in Fig. 1.
The results suggest a high variability and complex pattern of gene expression between samples as the two principal components only explain 41% of samples' variability. For genes heavily in uencing rst principal component (PC1) there is higher variation between groups (mainly between pN1 and pN1mi + pN0) than between samples in the same group (as sample-groups are differentiated along PC1). Genes heavily in uencing the second principal component (PC2) mainly explain intra group-sample variability, showing a high dispersion. Considering PC1, two pN1 samples stand out, S24 and S19.
A differential gene expression analysis was performed using the DESeq2 R package to compare between sample groups (pN0, pN1mi and pN1). (44) Comparing patients' SLNs with and without metastases (OSNA positive versus OSNA negative), 7 DEGs were identi ed as upregulated (1.8%) and none as downregulated (Table 4). Comparing OSNA positive with micrometastases (pN1mi) and OSNA negative SLNs (pN0), no DEGs were identi ed. Comparing OSNA positive with macrometastases and OSNA negative SLNs, 11 DEGs were identi ed as upregulated (2.8%) and none as downregulated (Table 5). Lastly, comparing OSNA positive SLNs with macrometastases and OSNA positive SLNs with micrometastases, 7 DEGs were identi ed as upregulated (1.8%) and none as downregulated (Table 6).    To allow an overview of the levels of expression of the 11 genes identi ed as being differential expressed between sample groups (pN0, pN1mi and pN1), these values are described in Table 7. A gradient of expression from pN0 to pN1 is evidenced, with some genes showing very low levels of expression in no metastatic SLNs (pN0).

Statistical analysis between DEGs and relevant clinicopathologic parameters
The results of Spearman correlation between relevant clinicopathologic parameters and the normalized expression levels of the 11 identi ed DEGs are shown in Table 8 (Table 8) as independent variables was attempted, but none of the independent variables were statistically signi cant. However, multiple linear regression with normalized expression levels of NECTIN2, LRG1, CD276, FOXM1 and IGF1R as dependent variables and the corresponding signi cant clinicopathologic parameters (Table 8)

Clusters
To identify patterns of gene expression within the 32 samples, a hierarchical clustering heatmap was constructed using the 11 identi ed DEGs (Fig. 2). Three main clusters were identi ed: cluster 1 including two cases with macrometastases (pN1), cluster 2 with most pN1 cases but also with two micrometastases cases (pN1mi) and cluster 3 aggregating N0 and most pN1mi cases ( Fig. 2; Table 9). The two cases with macrometastases of cluster 1 were from patients that also had non-sentinel LNs with macrometastases and vessels embolization on the histologic examination of the non-sentinel LNs. These samples were the two pN1 outliers previously identi ed in the PCA analysis (Fig. 1). The mean TTL of these two cases was high (375 950.0 ± 500 702.3) but some cases in cluster 2 showed higher values (Table 9 and Additional le 2). In cluster 1, the mean number of total LNs (SLNs and non-sentinel LNs) with macrometastases was also signi catively higher (5.0 ± 2.8 versus 0.3 ± 0.5 in cluster 2 + cluster 3; p < 0.001). Globally, all the 11 DEGs had higher expression in cluster 1 (Table 10). Cluster 2 included only OSNA positive cases: 6 with macrometastases and 2 with micrometastases. The mean TTL was 147 218.8 ± 173 519.8 (Table 9). Importantly, the two cases of SLNs with micrometastases that clustered with this group had a high TTL (3300 and 4500, respectively) (Additional le 2), close to the established cut-off of macrometastases (5 000). Globally, the 11 DEGs were less expressed than in cluster 1 but more expressed than in cluster 3 (Table 10).
On the other hand, the predominantly OSNA negative cluster (cluster 3) included 16 pN0 patients, 5 of the 7 pN1mi patients and one pN1 patient. The mean TTL of the six OSNA positive patients in this cluster 3 was low (1 693 ± 3 189.9). The patients with micrometastases had low TTL (minimum = 280; maximum = 620) and the only patient pN1 (with macrometastases) in this cluster had the lowest TTL among the patients with macrometastases (8 200) (Additional le 2). Cluster 3 had the lowest gene expressions of this 11 DEGs (Table 10). Moreover, considering only the patients with metastatic SLNs (n = 16), in cluster 3 (n = 6) the gene expression levels of the DEGs were globally signi cantly lower when compared to gene expression levels of clusters 1 and 2 (Additional le 3).
Finally, besides tumor diameter and tumor grade, the comparison of clinical and other tumor pathological characteristics (hormone receptors, LVI, Ki67 and TILs) between the three clusters did not reveal any statistically signi cant difference (Table 9).

Discussion
The cross-talk between immune cells and tumor cells modulates tumor metastases and response to therapy.  Tables 4 to 6. Remarkably, in our study, genes with an expression that could be mainly attributed to microenvironment cells, as granzymes (GZMA, GZMB, GZMH, GZMK), CD3 (CD3D, CD3E, CD3G), other immune system-response related genes codifying interleukins, IFN-γ, T cell receptors (TCRs) or immune checkpoint molecules such as Programmed Cell Death-1 (PD-1) or Cytotoxic T-Lymphocyte Associated Protein 4 (CTLA4), did not show differential expression between pN0, pN1mi and pN1 samples. This lack of evidence of the involvement of other genes associated with immune system activation may possibly be related to the increased expression of the inhibitory immune checkpoints VTCN1 and CD276.
Analysis of the expression levels of the 11 DEGs within the three groups of samples (pN0, pN1mi and pN1), as shown in Table 7, highlights the correlation with the metastatic load of SLN. Some genes, such as KRT7, VTCN1, CD44 or ALOX15B had very low or even no expression in pN0 samples (Table 7), which is in accordance with the low levels of the respective proteins in LNs,  (Table 5).
In our study, we used a target RNA-seq to study the transcriptomic patterns of immune response at the level of SLNs. Because RNA-seq does not rely on a predesigned complement sequence detection probe, there are no limitations such as crosshybridization with extremely similar sequences.(56, 57) A recent study also revealed that the information provided by microarray and RNA-seq data is not completely the same; however, the transcript abundance assayed by RNA-seq provides a higher quality estimate of protein abundance, because the signal-to-noise is improved, compared to data obtained using a microarray.(58) RNAseq is a very sensitive and speci c method, as evidenced by the detection of differential expression of transcripts with very low expression, such as VTCN1, supporting the reliability of our results. Moreover, when compared to the whole transcriptome, beyond reduced costs, targeted RNA-seq protocols are optimized for the selected transcripts, showing increased sensitivity.
Considerable evidence suggests that BC metastases arise from cells undergoing epithelial-mesenchymal transition and cancer stem-like cells. A previous study using single-cell RNA-seq in BC cell lines revealed that migratory BC cells exhibited overall signatures of epithelial-mesenchymal transition and cancer stem-like cells with variable expression of marker genes, and they retained expression pro les of epithelial-mesenchymal transition over time. (59)  Furthermore, the therapeutic target expression on normal tissues must be restricted, preferably at levels below the ones required for effector mechanism activation, in order to minimize toxicity.(62, 63)

Strengths and limitations
As far as we know, this is the rst study in human BC patients that intended to analyze the immune-related DEGs in the whole SLN, comparing the global microenvironment of non-metastatic and metastatic SLNs. Furthermore, the metastatic SLNs were classi ed as micrometastatic and macrometastatic and the microenvironment was compared according to the tumor load.
To the best of our knowledge, this is also the rst study in BC patients using a target RNA-Seq potentially useful for clinical translation. Previous gene expression studies in LNs used mainly microarray-based datasets. In our study, instead of microarrays, we used targeted RNA-seq, a more sensitive and speci c method, with a higher quality estimate of protein abundance.
It is known that distinct transcriptomic pro les across molecular subtypes is associated with inter-tumoral heterogeneity of BC. (45) In this study, selecting a homogeneous cohort of patients with Luminal A early BC, we established a differential immune transcriptomic pro le of Luminal A BC metastatic SLNs and we were able to de ne three different clusters. Moreover, as the aggressive behavior of BC seems to derive from LNs metastases, these ndings could help to take a further step in de ning a more precise prognosis of Luminal A BC patients and improving methods of personalized treatments towards higher effectiveness and less side effects. To the best of our knowledge, this study is the rst to delineate the immune transcriptomic pro le of Luminal A BC SLNs metastases.
Finally, as another major strength, this study is the rst to use the OSNA lysate spared sample in the search for prognosis and treatment related markers associated with tumor-microenvironment interplay and not only tumor markers. Hence, this study, using a targeted RNA-seq in the OSNA lysate spared samples, evidenced that a selected SLN gene expression pro le can identify molecular markers useful as a new prediction tool for prognosis evaluation and treatment selection. The RNA extraction from OSNA lysate spared sample is easier and allows a higher RNA concentration, with higher quality when compared with formalinxed para n-embedded tumor samples and may constitute an alternative to tumor RNA characterization, particularly when the primary tumor size is small, as in the majority of current patients. This approach also has the additional advantage of maintaining the integrity of the primary tumor samples for eventually necessary future studies. Furthermore, since, by law, OSNA lysates samples are, currently, not required to be preserved, OSNA lysates would otherwise be wasted. Therefore, OSNA lysate samples have less ethical and legal implications and, nowadays, have no other utilities besides SLNs staging. Additionally, accordingly to previous DEGs studies, there is a transcriptomic similarity between primary BC and its corresponding LN metastases. (52) Thus, this similarity may translate into new prognosis data and therapeutic strategies using the OSNA lysate spared sample from each BC patient. Lastly, as OSNA is being adopted worldwide by an increasing number of centers in other type of cancers besides BC, RNA-seq in the OSNA lysate spared samples could have a wider utility.
On the other hand, as a limitation, the OSNA lysate samples are obtained from homogenized SLNs, and thus, this approach performs a global evaluation of SLNs' microenvironment, including tumoral and non-tumoral cells. This limitation is inherent to OSNA sample. RNA-seq deconvolution analysis, a computational method that can simultaneously estimate both sample-speci c cell-type proportions and cell-type-speci c gene expression pro les using bulk tissue samples, is not feasible in this study because it would require a signi cantly higher number of targets.(64) However, as already discussed, regardless of the cell of origin, the DEGs can be useful biomarkers.
The overexpression of several potential targets for immunotherapy in metastatic Luminal A BC SLNs represents a promising therapeutic target. However, as this was a RNA-seq study, successful targeting would require further knowledge about the amount and distribution of protein expression, because the RNA-protein correlation may be distorted by posttranscriptional regulation. Finally, this study had a small sample size and no follow-up data. A larger cohort of patients with subsequent long-term follow-up will be necessary to enrich these results, especially regarding clusters implications.

Conclusions
Using a targeted RNA-seq, in OSNA lysate of SLNs from Luminal A BC patients, it was found that, in metastatic SLNs, there were upregulated immune-related genes. Globally, in metastatic SLNs, KRT7, VTCN1, CD44, GATA3, ALOX15B, RORC and NECTIN2 were upregulated. In macrometastatic SLNs, LRG1, CD276, FOXM1 and IGF1R were also upregulated. In metastatic SLNs, higher metastatic load and higher levels of TTL were correlated with higher expression levels of the majority of the DEGs. Hierarchical clustering analysis revealed three different clusters, not coincident with pN0, pN1mi and pN1 classi cation, suggesting that the expression pro le of these genes may bring further information on current SLN evaluation.
The 11 identi ed DEGs codify proteins mainly involved in cancer aggressiveness and with impact in immune response. Some of the found DEGs are biomarkers potentially useful to take a further step in improving personalized treatment strategies towards higher effectiveness and less side effects.
As OSNA is being adopted worldwide in another cancers besides BC, RNA-seq in the OSNA lysate could also have utility for other cancer types. In the future, the SLN's gene-signature study could be used in order to de ne a more precise prognosis and choose the best therapy according to patient characteristics, as the complex interaction between cancer and the host immune system will be the main strategic key to future personalized treatment tools. CHUC-045-20).

Consent for publication
Not applicable.

Availability of data and materials
The datasets generated from RNA-seq data and analysed during the current study are available in the NCBI's Gene Expression Omnibus repository, with the accession number GSE210006.

Competing interests
The authors declare that they have no competing interests.  First two principal components analysis with all genes and samples. Each group of samples has a speci c color.