The clinical data from the medical records of 2 patients (case 1 and case 2) diagnosed with breast ACC were retrospectively reviewed. Image-guided CNB was performed before surgical treatment between January 1, 2015 and December 31, 2022. The cases originated from the Department of Pathology of the People's Liberation Army 989 Hospital (Pingdingshan, China) or the Fenlan Laboratory (Hangzhou, China). The cases were reviewed by five of the authors and samples were anonymized prior to analysis. The present study was approved by 989 Hospital Medical Ethics Committee (approval no. WZLL-2023-016) and Fenlan Lab Medical Ethics Committee (2023–06). Patient consent was obtained if required by the protocols. All tissues were fixed in 10% formalin at room temperature for 24 h. For each tissue specimen, 4-µm sections were cut. Hematoxylin and eosin-stained sections at room temperature for 15 min of each case were reviewed by three pathologists (LCY, LFY and LGX) under the BX53 light microscope (Olympus Corporation). The inclusion criterion was a diagnosis in accordance with breast ACC.

Biopsy and surgical tissue sections (4 µm) from paraffin blocks were stained immunohistochemically using a BOND-MAX Automated IHC/ISH Stainer (Leica Microsystems GmbH). All tissue blocks were fixed in 10% formalin at room temperature and embedded in paraffin for 24 h. Sections were mounted onto slides, air dried for 20 min and heated at 60°C for 20 min. Sections were dewaxed twice with dewaxed solution (Celnovte Biotechnology Co., Ltd.) at 72°C for 2 min, hydrated three times with anhydrous ethanol at room temperature for 2 min, and washed four times with washing solution (Celnovte Biotechnology Co., Ltd.) at room temperature for 6 min, The heat (98°C)-induced antigen retrieval method was performed using Tris-EDTA buffer (1X; cat. no. K0071; Shanghai Jiehao Biotechnology Co., Ltd.) and endogenous peroxidase activity was quenched with 3% peroxidase-blocking reagent (Celnovte Biotechnology Co., Ltd.) at 37°C for 5 min. The slides were incubated with the following commercially available antibodies at 37°C for 30 min: Estrogen receptor (ER) (cat. no. CEM-0081; 1:500), progesterone receptor (PR) (cat. no. CPM-0365; 1:500), HER2 (cat. no. CCM-0844; 1:100); cytokeratin (CK)5/6 (cat. no. CCR-0982; 1:500), S100 (cat. no. CSM-0101; 1:500), GATA binding protein 3 (GATA3; cat. no. CGM-0130; 1:500), CK7 (cat. no. CCM-0992; 1:500), p63 (cat. no. CPM-0160; 1:100) and Ki-67 (cat. no. CKM-0032; 1:500; Table I). All antibodies were purchased from Celnovte Biotechnology Co., Ltd. Tissue sections were then washed twice with washing solution (Celnovte Biotechnology Co., Ltd.) for 6 min and incubated with Microstacker™ + Linker (cat. no. SD5300; 1:20; Celnovte Biotechnology Co., Ltd) at room temperature for 15 min for signal amplification. After washing twice with TBS for 6 min, samples were incubated with Microstacker™ Flex HRP-polymer detection reagent (cat. no. SD5100; 1:20; Celnovte Biotechnology Co., Ltd) at 37°C for 30 min. After incubation with the polymer reagent, tissue sections were thoroughly washed three times with TBS buffer for 6 min and incubated with Microstacker™ DAB + Chromogen (cat. no. SD5300; 1:20; Celnovte Biotechnology Co., Ltd) at 37°C for 6 min. Slides were washed twice with TBS buffer for 6 min, then counterstained with hematoxylin at 37°C for 2 min and washed with TBS and dH₂O for 6 min, respectively. Dehydration using graded ethanol solutions and 90% xylene was performed, then sections were mounted in synthetic resin and observed under a BX53 light microscope (Olympus Corporation).

IHC results were scored by two independent observers according to the percentage of positively stained cells: i) 0+, 1–25% staining; ii) 1+, 26–50% staining; iii) 2+, 51–75% staining; and iv) 3+, 76–100% staining.

HER2 FISH was performed using the Kanglu HER2 DNA Probe Kit (Wuhan HealthCare Biotechnology Co., Ltd.) according to the manufacturer's protocols. Tissue blocks were fixed in 10% formalin at room temperature and embedded in paraffin for 24 h. For each tissue specimen, 4-µm sections were cut, deparaffinized, rehydrated in ethanol gradient and heated in a pretreatment solution of 2-(N-morpholino) ethanesulphonic acid at 95°C for 10 min before being soaked twice in Tris/HCl buffer at room temperature for 3 min. The specimens were then digested in 200 µl pepsin (4 mg/ml; Guangzhou Anbiping Medical Technology Co., Ltd.) working solution at 37°C for 45 min and agitated twice in Tris/HCl buffer at room temperature for 3 min. Slides were then dehydrated in an ascending ethanol series at room temperature and air-dried. Probe (10 µl; 500 ng/µl) was applied to each slide at 85°C for 5 min and the hybridization area was covered with a coverslip. Co-denaturation of the probe and specimen was then performed at 85°C for 5 min, before the slides were incubated at 37°C for 10 h. The slides were washed in 2X saline-sodium citrate/0.1% NP-40 at 37°C for 10 min and 10 µl DAPI counterstain was applied at 85°C for 5 min before sealing. Slides were imaged using a Zeiss Imager Z2 fluorescence microscope using FISH imaging systems (V 1.3; GuangZhou MM Photoelectric Techology Co., Ltd.) software (Fluorescence microscope MF43-M (medical device model) + micro camera MC50-S/MS23+ fluorescent light source MG-100/MG-120/MG-200). The fluorescent signal in 20 cells from each slide was analyzed by two independent observers. FISH results were recorded as either negative or positive according to the American Society of Clinical Oncology/College of American Pathologists guideline by the two experienced pathologists independently in 20 tumor cells in a blinded manner (16).

For ETS variant transcription factor 6 (ETV6) FISH, 4-µm 10% formalin-fixed at room temperature for 24 h, paraffin-embedded tumor sections were heated at 60°C for 1 h, rinsed in 100% ethanol and pretreated with 0.2 N HCl for 20 min at room temperature. Slides pretreatment, hybridization, washing and counterdyeing steps followed: Place slides in a 65±5°C incubator for 24 h, room temperature in xylene twice, 10 min each time → 100% ethanol at room temperature for 10 min → 90% ethanol at room temperature for 3 min → 70% ethanol at room temperature for 3 min → purified water at room temperature for 3 min → purified water at 100°C for 25 min → pepsin (4 mg/ml; Guangzhou Anbiping Medical Technology Co., Ltd.) working solution at 37°C for 15 min → 2×SSC at room temperature for 3 min → 70% ethanol at room temperature for 2 min → 90% ethanol at room temperature for 2 min → 100% ethanol at room temperature for 2 min → add probe (500 ng/µl; Guangzhou Anbiping Medical Technology Co., Ltd.) to the slides→ denature at 85°C for 5 min → hybridization at 37°C for 18 h → 0.1% NP-40/2X SSC at 37°C for 5 min → 70% ethanol for at room temperature 3 min → hybrid blue staining solution at room temperature. Slides were imaged using a Zeiss fluorescence microscope (Carl Zeiss AG). IMSTAR software (version 2.0; ProgRes CT3, Jena, Germany) was used to analysis the results.

The two patients underwent molecular analysis using the Illumina NGS platform (Illumina, Inc.), which can detect mutations, copy number variations and fusions in 685 genes that are known to be oncogenic drivers (17). NGS was performed for the two patients using DNA from both tumor samples. NGS was performed according to Illumina's standard protocol. DNA was extracted using the MeiJi Tissue & Blood DNA Kit (cat. no. IVD3101-200; Guangzhou Magen Biotechnology Co., Ltd.) according to the manufacturer's protocol. A paired-end DNA library was prepared according to the manufacturer's protocol (Agilent Technologies, Inc.). Samples with >1 µg DNA were used for library preparation. Qubit quantification was used to verify the quality and integrity of samples and that the extraction amount was ≥50 ng. The loading concentration of the final library was 0.7 nM. The concentration of Cng/ul was measured by Qubit, and the average size of the library was double base pair) analyzed using the Agilent 4200 bioanalyzer (Agilent Technologies, Inc.). The molar concentration was calculated according to the formula: M=(Cng/µl ×10⁶)/(660 g/mol × D) nmol/l. Genomic DNA was fragmented using a sonicator (Covaris, LLC) to a length of 200–300 bp. The ends of the DNA fragments were then repaired, before the Illumina Adaptor was added (Fast Library Prep Kit; iGeneTech Bioscience Co., Ltd.) Using the 2003200-P2 Unique Dual Index Adapters (2 nmol) and KAPA Library Construction kit kk8514 (F. Hoffmann-La Roche Ltd; cat. no. 07962428001) KAPA Hyperplus Kit. The DNA fragments were end-polished, A-tailed and ligated with the full-length adapter. After the sequencing library was constructed, whole exomes were captured with the NadPrep Hybrid Capture Reagents (Nanodigmbio Biotechnology Co., Ltd.) and double-ended double-label sequenced using a Nextseq 6000 sequencer (Illumina, Inc.) with the NovaSeq 6000 S1 Reagent Kit v1.5 (300 cycles; cat. no. 20028317; Illumina, Inc.). The maximum sequencing read length was 2×150 bp (Paired-End 150 bp) and double-ended double-label sequencing was performed. Fastp software (version 0.11.9) (18) was used for quality control and preprocessing and the Burrows-Wheeler Aligner software (version 0.7.16) (19) was used to align clean reads with the hg19 reference genome. Mutect software (version 1.1.6) (20) was used to detect somatic variants in the tumor samples. Single nucleotide variants were annotated and filtered using snpEff software (version 5.1) (21).

The clinical data of the 2 patients were recorded (Table II). Both of the patients were female, and the patients were aged 42 and 69 years old at the time of diagnosis. Both patients reported the presence of a mass with no other complications. The preoperative duration was 3 days and 1 month for case 1 and case 2, respectively. Follow-up information was available for case 1 and case 2 for 16 and 14 months, respectively. After surgery, case 1 was administered with adjuvant chemotherapy of the TC taxel + Cyclophosphamide) scheme (120 mg/m² docetaxel and 800 mg/m² cyclophosphamide) intravenously every 3 weeks for 4 cycles. Subsequently, 1 year after surgery, the patient (case 1) developed a liver metastasis (Fig. S1), accompanied by lumbar vertebra and sacral vertebra metastasis. In addition, 15 months after surgery, the patient presented with ascites. Case 2 declined any further therapy and showed no recurrence or metastasis after surgery. Patients denied the use of novel antibody-drug conjugates (ADCs).

In case 1, a hypoechoic nodule of irregular shape with dimensions of ~4.6×2.3×2.7 cm was found in the left breast (‘6 o'clock’) beneath the nipple with unclear boundaries (Fig. 1A). The lesion was characterized as Breast Imaging Reporting and Data System (BI-RADS) category 4 (22).

In case 2, a hypoechoic mass measuring ~1.9×1.4×1.9 cm could be observed in the right breast (‘11 o'clock’), 2 cm away from the nipple. The tumor boundary was visible, and the shape was regular and was accompanied by a lateral sound shadow with a strong echo. Blood flow signals could be detected in the mass. The lesion was characterized as BI-RADS category 4 (Fig. 1B).

Both of the tumors exhibited solid or nested growth patterns. The solid or nested growth patterns interconnected to form the burrowing labyrinthine network, ‘hand-holding-hand’ pattern or crawling phenomenon, which is similar to the carcinoma cuniculatum in the mandible (22) or the crawling carcinoma in the stomach (23) (Figs. 3A-E and 4A-E). The neoplastic cells exhibited abundant basophilic cytoplasm with centrally-located nuclei and prominent nucleoli, but also occasionally exhibited a clear cytoplasm (Figs. 3F and 4F). Atypia of tumor cells was prominent with marked mitotic figures (Figs. 3F and 4F insert). The cancer cells had high-grade nuclei with moderate-to-severe pleomorphism, coarse chromatin and a prominent singular nucleolus (Figs. 3F and 4F). An in situ component was not present either within or around the tumor. There were no invasive lymphocytes in the tumor tissues (Figs. 3A and 4A). Tumors were positive for S100 protein (Figs. 3G and 4G) in both patients. The neoplastic cells exhibited moderately positive (2+) HER2 expression (Fig. 3H and 4H). Both of the cases were CK5/6- (Fig. S2A and B) and CK7-positive (Fig. S2C and D). The tumor was positive for GATA3 in case 1 (Fig. S2E), but negative in case 2 (Fig. S2F). Staining for p63 (Fig. S2G and H), ER (Fig. S3A and B) and PR (Fig. S3C and D) was negative in both tumors. The Ki-67 index was 30% (Fig. S3E) and 55% (Fig. S3F) in the tumors from case 1 and case 2, respectively.

Targeted NGS results from case 1 demonstrated the presence of mutations in the tumor protein p53 binding protein 1 (NM_001141980.1; c.1362_1367del; p.I455_P456del), partner and localizer of breast cancer 2 (NM_024675. 3; c.2799del; p.C933Wfs*2), fibroblast growth factor 3 (NM_005247.2; c.542G>A; p.R181H), α thalassemia/mental retardation syndrome X-linked (NM_000489.4; c.5690A>G; p.E1897G), mTOR (NM_004958.3; c.6414G>T; p.L2138F), Kelch-like protein 6 (NM_130446.2; c.1589C>G; p.A530G), cyclin E1 (NM_001238.3; c.479A>G; p.K160R), MED12 (NM_005120.2; c.3817G>T; p.A1273S; abundance, 2.17%) and E74-like ETS transcription factor 3 (NM_004433.4; c.485G>A; p.G162D) genes.

The gene mutations present in case 2 included TP53 (NM_000546.5; c.323del; p.G108Vfs*15), nuclear receptor-binding SET domain 1 (NM_022455.4; c.3157A>T; p.K1053*), spen family transcription repressor (NM_015001.2; c.7480C>T; p.P2494S), disruptor of telomeric silencing 1-like (NM_032482.2; c.3580G>A; p.E1194K), rearranged in transfection (NM_020975.4; c.538C>T; p.R180*), janus kinase 3 (NM_000215.3; c.3118T>A; p.C1040S), mitogen-activated protein kinase kinase kinase 14 (NM_003954.4; c.115G>A; p.V39I), MED12 (NM_005120.2; c.3817G>T; p.A1273S; abundance, 3.25%) and lysine methyltransferase 2C (NM_170606.2; c.1814-1G>A; p.*). The MED12 exon 27 mutation was found in both cases (Tables III and IV).

FISH showed the average copy number of the red signal/average copy number of the green signal in case one and case two was 1.0 and 1.11, respectively, and revealed that the tumor tissues in the two case were negative for HER2 amplification (Fig. S4A and B). The two cases showed 2G(green signal)2R(red signal) signal mode and indicated no ETV6 break-apart probe detection (Fig. S4C and D).