Method of improving therapy response in triple-negative breast cancer

Targeting the complement cascade with complement inhibitors in combination with chemotherapy improves treatment efficacy in TNBC by inhibiting complement components like C1QA, C1QB, C1QC, C1R, CIS, C2, C3, and/or C4B, addressing chemotherapy resistance and identifying candidates through C3 protein activation detection.

WO2026213217A9PCT designated stage Publication Date: 2026-11-12RGT UNIV OF CALIFORNIA +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/US2026/021780
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-03-31
Filing Date
2026-03-31
Publication Date
2026-11-12

AI Technical Summary

Technical Problem

Triple-negative breast cancer (TNBC) is characterized by high aggressiveness and poor prognosis due to molecular and cellular heterogeneity, with the complement cascade contributing to tumor growth, metastasis, and chemotherapy resistance, and existing treatments offer limited insights into specific interactions within the tumor ecosystem.

Method used

Administering a therapeutically effective amount of a composition comprising a complement inhibitor, such as C1QA, C1QB, C1QC, C1R, CIS, C2, C3, and/or C4B, in combination with immunotherapy or chemotherapy, to inhibit the complement cascade and improve chemotherapeutic response in TNBC patients, and identifying candidates for complement C3 inhibitor therapy by detecting elevated C3 protein activation in atypical pre-invasive epithelial regions.

Benefits of technology

Enhances chemotherapeutic response and improves treatment efficacy in TNBC patients by targeting the complement cascade, particularly in those with residual disease, and provides a method to identify suitable candidates for C3 inhibitor therapy based on C3 protein activation levels.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure US2026021780_12112026_PF_FP_ABST
    Figure US2026021780_12112026_PF_FP_ABST
Patent Text Reader

Abstract

Provided herein are methods of treating triple-negative breast cancer in a subject, the method including: administering a therapeutically effective amount of a composition comprising a complement inhibitor to the subject, wherein the complement inhibitor inhibits a complement cascade component, and wherein the complement cascade component comprises C1QA, C1QB, C1QC, C1R, C1S, C2, C3, or C4B, thereby treating the triple-negative breast cancer in the subject.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0002] METHOD OF IMPROVING THERAPY RESPONSE IN TRIPLE-NEGATIVE BREAST CANCER

[0003] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to U.S. Provisional Patent Application No.

[0004] 63 / 780,895, filed on March 31, 2025. The disclosure of the prior application is considered part of the disclosure of this application and is incorporated herein by reference in its entirety.

[0005] BACKGROUND

[0006] Triple-negative breast cancer (TNBC) is a highly aggressive breast cancer (BC) subtype, characterized by <1% expression of estrogen and progesterone receptors (ER and PR) and wild-type expression of HER2 / neu. While TNBC generally carries a poor prognosis, not all TNBCs are fatal. Treatment efficacy, prognosis, and subsequent survival are increasingly acknowledged to be affected by interactions of the tumor cells with its surrounding cellular milieu (tumor microenvironment, TME). High-throughput studies over the past decade have revealed the molecular and cellular heterogeneity of TNBC with implications for therapy. Primarily based on bulk gene expression data, these studies highlight inter-patient heterogeneity and tumor immunogenicity but offer limited insights into specific interactions within the tumor ecosystem including molecular interactions between the malignant cells, stromal components (e.g., fibroblasts, adipocytes, pericytes, immune cells), vasculature, or the influence of environmental factors like hypoxia, stress, or extent of immune infiltration. Recent advancements in single-cell and spatial technologies address the limitations of bulk profiling technologies and offer increased granularity of the complex multidimensional interactions in tissue.

[0007] SUMMARY

[0008] Provided herein are methods of treating triple-negative breast cancer in a subject, the method comprising administering a therapeutically effective amount of a composition comprising a complement inhibitor to the subject, wherein the complement inhibitor inhibits a complement cascade component, and wherein the complement cascade component comprises C1QA, C1QB, C1QC, C1R, CIS, C2, C3, and / or C4B, thereby treating the triplenegative breast cancer in the subject.Attomey Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0009] Provided herein are methods improving chemotherapeutic response in a triple negative breast cancer (TNBC) patient with residual disease, the method comprising: administering a therapeutically effective amount of a composition comprising a complement inhibitor to the patient, wherein the complement inhibitor inhibits a complement cascade component, and wherein the complement cascade component comprises C1QA, Cl QB, C1QC, C1R, CIS, C2, C3, and / or C4B; and administering a therapeutic effective amount of an immunotherapy or chemotherapy, thereby improving chemotherapeutic response in a TNBC patient with residual disease.

[0010] Provided herein are methods of identifying a TNBC patient as a candidate for complement C3 inhibitor therapy, the method comprising: obtaining a biological sample from the patient, wherein the biological sample comprises an atypical pre-invasive epithelial region; detecting C3 protein activation in the atypical pre-invasive epithelial region; comparing the C3 protein activation to a reference value derived from a good-prognosis TNBC patient; identifying the patient as a candidate for complement C3 inhibitor therapy based on the C3 protein activation in the pre-invasive epithelial region being elevated relative to the reference value; and administering a complement C3 inhibitor to the identified TNBC patient. In some embodiments, the detecting C3 protein activation comprises immunohistochemical staining the biological sample with an anti-C3 or anti-C3b antibody. In some embodiments, the detecting step further comprises quantifying a C3-positive cell fraction by digital pathology image analysis.

[0011] Provided herein are methods comprising: providing a sample of atypical breast tissue from a subject having triple-negative breast cancer; and assaying the sample to determine a level of C3 protein activation in the sample. In some embodiments, the determining of the level of C3 protein activation comprises immunohistochemical staining the biological sample with an anti-C3 or anti-C3b antibody.

[0012] Provided herein are methods comprising: providing a breast tissue sample from a subject, the sample containing atypical epithelia; assaying the breast tissue sample using spatial transcriptomic profiling to detect a presence of anaphylatoxin-mediated priming in the atypical epithelia by identifying at least one unique ligand-receptor interaction selected from the group consisting of C3-CD81 and C3-LRP1; determining that the identified ligandreceptor interaction is localized within an area of illumination (AOI) derived from a specific spatial origin within the breast tissue sample, wherein the spatial origin is selected from the group consisting of a tumor center region and a tumor edge region; and administering a C3-Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0013] directed therapeutic agent to the subject identified as having the C3-CD81 or C3-LRP1 interaction at the tumor center or tumor edge to treat high-risk pre-cancerous breast lesions.

[0014] In some embodiments, the spatial transcriptomic profiling of the atypical epithelia comprises: hybridizing the breast tissue sample with a plurality of profiling probes, each profiling probe comprising an indexing oligonucleotide attached to a target-binding moiety via a UV- photocleavable linker; identifying at least one region of interest (ROI) within the hybridized breast tissue sample using immunofluorescent markers; segmenting the ROI into an AOI corresponding to the atypical epithelia using a digital micromirror device to direct UV illumination with ~1 pm precision; releasing the indexing oligonucleotides from the atypical epithelia segment via the directed UV illumination; and sequencing the released indexing oligonucleotides to identify the presence of the C3-CD81 or C3-LRP1 interaction.

[0015] In some embodiments, the methods comprise further identifying the subject as having (i) less than 1% expression of estrogen and progesterone receptors and / or wild-type expression of HER2 / neu gene as compared to reference non-cancerous breast tissue, and / or (ii) elevated complement C3 activation in atypical epithelial breast tissue as compared to reference non-cancerous breast tissue. In some embodiments, the complement inhibitor is a C3 inhibitor. In some embodiments, the C3 inhibitor comprises: a monoclonal antibody that specifically binds to C3 or a C3 cleavage fragment; a compstatin or compstatin analog peptide that inhibits C3 cleavage; a small molecule inhibitor of C3 or C3 convertase; an RNA-based therapeutic targeting C1QA, C1QB, C1QC, C1R, CIS, C3, or C2; or a C3a receptor (C3aR) inhibitor. In some embodiments, the C3 inhibitor comprises pegcetacoplan or compstatin. In some embodiments, the subject has been previously administered a neoadjuvant chemotherapy or an adjuvant chemotherapy. In some embodiments, the administering comprises intravenous administration, subcutaneous administration, intraperitoneal administration, rectal administration, oral administration, or combinations thereof. In some embodiments, the methods further comprise administering an anti-cancer treatment. In some embodiments, the anti-cancer treatment comprises a therapeutic a-lactalbumin vaccine, a platinum-based chemotherapy, a taxane-based chemotherapy, an immune checkpoint inhibitor, a PARP inhibitor, an antibody-drug conjugate, adoptive cell therapy (ACT), or any combinations thereof.

[0016] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Methods and materials are described herein for use in the present invention; other, suitable methods and materials known in the art can also be used. The materials, methods,Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0017] and examples are illustrative only and not intended to be limiting. All publications, patent applications, patents, sequences, database entries, and other references mentioned herein are incorporated by reference in their entirety. In case of conflict, the present specification, including definitions, will control.

[0018] Other features and advantages of the invention will be apparent from the following detailed description and figures, and from the claims.

[0019] BRIEF DESCRIPTION OF DRAWINGS FIG. 1A shows a schematic of the study design and processing. 17 good prognosis (GPx) and 15 poor prognosis (PPx) TNBC samples were selected, and H&E stained, with regions of interest (ROIs, dots show in IF images) carefully selected using a standardized approach. 4 channel s / antibody stains were utilized to visualize the cells in IF (PanCK+, CD3+, CD45+, DAPI). Selected ROIs were subject to GeoMX profiling, and the .DCC files subsequently generated was QC-ed and processed using GeomxTools. Q3 normalized data was utilized for downstream processing and analysis, unless otherwise stated. FIG. IB shows an exemplary H&E slide with the IF imaging. FIG. 1C shows exemplary ROIs on the slide in FIG. IB. ROIs were selected by pathologist / trained clinician at the tumor interface, within the center of the tumor (with or without immune cells), and if available, regions distant from the center and edge (isolated foci). If available, regions with morphologically normal and / or precancerous changes (PanCK+), and with varying immune infiltration patterns (CD3+, Cd45+) regions were also chosen for ROIs. FIG. ID shows a bar plot showing the distribution of post processed AOIs captured from different annotated regions of the slides).

[0020] FIG. IE shows a bar plot summarizing the distribution of the post processed AOIs, showing a fairly equal number of PanCK+ and PanCK- segments (604 GPx and 547 PPx AOIs). FIG. IF shows a UMAP of AOIs from all the GPx and PPx samples post processing, showing fair separation between the segments, indicative of difference in underlying biology of the cellular landscape captured by segmenting, across a majority of the ROIs in both GPx and PPx.

[0021] FIGs. 2A-2H show immune deconvolution of the tumor microenvironment in good (GPx) and poor prognosis (PPx). FIGs. 2A-2B show quantification of CD45 and CD3 channel intensities in the PanCK- segments. FIG. 2C shows differentially expressed genes captured in the comparison of AOIs from center (C) to edge (E) from the PanCK- segments in PPx. Gene names are highlighted when p.adj < 0.05. FIG. 2D shows log -fold change (M) toAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0022] average expression (A) of GPx vs PPx PanCK- AOIs. The top 100 genes are highlighted in red, with the top 30 named on the plot. PPx shows a more distinct increase in expression of genes that are actively involved in EMT including PTMA, VIM, PTBP1 etc. FIG. 2E shows the normalized frequency of immune cell type deconvoluted in 313 GPx (top panel) and 272 PPx (bottom panel) PanCK- AOIs generated using SpatialDecon and a custom reference profile matrix. The abundance of 7 major immune cell types (T-cells, B- cells, plasmablasts, myeloid, PVLs and fibroblast (mesenchymal cell types) and endothelial cells) was estimated.

[0023] FIG. 2F shows correlation between the CD3+ channel intensity captured and T cell deconvolution in the PanCK- segments, captured in the GPx and PPx samples. FIG. 2G shows sample level estimation of deconvoluted cell frequencies for GPx, with the names corresponding the subtype identified in the earlier study. FIG. 2H shows similar results for PPx. The box highlights the reduced presence on T-cells by deconvolution in women classified as CC2.

[0024] FIGs. 3A-3E show the establishing of the transcriptional states of immune cells in the T-enr subset. FIG. 3A shows SPI spatial partial information (SPI) (measure of spatial entropy) calculated for CD3+ vs PanCK+ channels in GPx-Tenr and PPx-Tenr. FIG. 3B shows the differentially expressed genes identified by comparing the expression of GPx and PPx PanCK- AOIs. FIG. 3C shows normalized distribution of cell states identified using EcoTyper in AOIs that were enriched for immune (T) cell infiltration within GPx and within PPx. FIG. 3D shows a table summarizing the cell states that are preferentially increased within GPx-Tenr and PPx-Tenr as captured in FIG. 3C. FIG. 3E shows a chord diagram highlighting the top 50 unique ligand-receptor (L-R) interactions (at corr>0.25, q<0.05) identified using BulkSignalR, within the TME of the GPx- Tenr (left) and PPx- Tenr (right), respectively.

[0025] FIGs. 4A-4H show the characterizing of the tumor cell state. FIG. 4A shows distribution of AOIs with invasive epithelia that were captured in the center (C), edge (E) and isolated foci (F) of the tumor samples in GPx and PPx. FIG. 4B shows a comparison of consensus gene features in tumor cells (AOIs with invasive epithelia) from center of the tumor to those at the edge of the tumor showed no statistically significant differences (at FDR<0.05) in either GPx or PPx patients. FIG. 4C shows differential gene expression identified significant transcriptional dysregulation in invasive epithelia compared to normal-adjacent epithelia from GPx (204 AOIs, left) and PPx (184 AOIs, right). FIG. 4D shows a Venn highlighting the overlap in genes upregulated within the invasive epithelia from GPx (GPx-I) and PPx (PPx-I). FIG. 4E shows protein-protein interaction network generated using the uniquelyAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0026] dysregulated genes GPx-I (155 / 346) form a tightly connected network of interacting genes PPI enrichment adj p <10-16) associated with immune signaling, antigen processing and presentation (APC), and cellular adhesion including genes such as MX1, IFI6, CTSB / D, HSP90AA1, PSMB1, PSMC4, PSMC2, ANXA2 / 5. FIG. 4F shows a box plot of the mean expression of 54 immune genes identified within the PPI shows a significant increase in GPx-I AOIs compared to the PPx-I AOIs. FIG. 4G shows mean expression of the recently published optimized immune response signature (oIRS) in GPx-I and PPx-I, shows a statistically significant preferential increase in the tumor epithelia of GPx-I compared to PPx-I. FIG. 4H shows a PPI of uniquely upregulated genes in PPx-I showed evidence for extensive metabolic reprogramming and chromatin remodeling including genes such as MKI67, NDUFAB1, NDUFS6, HSPA9, AKT2, PFKP, PRDG2, MDH1 / 2, ATP5F1C, CTPS1, LDHA (mitochondrial and cellular metabolism, fdr<0.001) and AURKB, KIF2C, KIF1C, H2As, TPX2 (chromatin remodeling / organization, fdr<0.001.

[0027] FIGs. 5A-5G show the characterizing of the pre-invasive epithelial states and establishing a prognostic tumor cell signature. FIG. 5A shows differential expression analysis comparing the atypic to the invasive epithelia. PPx samples show a more dominant dysregulation of genes along this transition. FIG. 5B shows hallmark gene set enrichment of the genes identified as being upregulated in the atypia compared to invasive epithelia of PPX. FIG. 5C shows boxplots showing significantly higher expression of genes (mean expression) involved in the complement cascade within PPx -A compared to GPx-A AOIs. FIG. 5D shows boxplots showing significant increase in expression of genes in the complement cascade within AOIs with pre-invasive epithelial compared to AOIs with invasive epithelia (PPx-I).

[0028] FIG. 5E shows the L-R diagram highlighting the top 50 unique interactions captured within GPx and PPx atypia (with enrichment for epithelial cell types). FIG. 5F shows a schematic representation of the CNN model used for identification of epithelial gene signatures discriminatory of GPx and PPx to (235 and 236 genes respectively). FIG. 5G shows C3 protein activation in pre-invasive epithelial regions confirmed by immunohistochemistry. Percentage of C3-positive cells in co-registered ROIs across GPx and PPx samples, stratified by epithelial morphology (A = atypia, I = invasive, J = normal-adjacent). C3 activation is significantly elevated in PPx-A compared to GPx-A (*p < 0.05 and **p < 0.01).

[0029] FIG. 6A shows a bar plot showing the distribution of total AOIs captured pre-processing (1424 AOIs - 737 GPx and 687 PPx). FIG. 6B shows the distribution of AOIs pre-processing (left panel) and post-processing for both GPx and PPx samples. Each bar represents a patient sample. FIG. 6C shows frequency of the AOIs captured in the various regions (Center-C,Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0030] Edge-E and Isolated foci- F) of the tumor FFPE slice, pre (left panel) and post-processing (right) across both GPx and PPx samples. Each bar represents a patient sample. FIG. 6D shows frequency of the AOIs captured with various epithelial types as annotated by the pathologist, pre (left panel) and post-processing (right) across both GPx and PPx samples. Each bar represents a patient sample. FIG. 6E shows a circle dendrogram capturing the heterogeneity (intra and inter-patient) of the expression within ROIs. Every patient is represented by a unique color.

[0031] FIG. 7A shows differential expression captured in the PanCK- segment of GPx, between center and edge AOIs. FIG. 7B shows differential expression for GPx vs PPx between AOIs captured in each region (C-center, E-edge and F-isolated foci) of the tissue. FIG. 7C (top and bottom) captures the cell type frequencies of the T-cells, B- cells, Plasmablasts, Myeloid cells and mesenchymal (PVL+ fibroblasts) in GPx and PPx PanCK- AOIs respectively. FIG. 7D shows the abundances of the seven cell types estimated by spatial deconvolution the 61 AOIs from 5 GPx patients. Each color bar represents a patient. FIG. 7E shows the abundances of the seven cell-types estimated by spatial deconvolution the 40 AOIs from 3 PPx patients. Each color bar represents a patient. FIG. 7F shows average expression of CD8T cell markers in S03 (exhausted T) shows a significant (p<0.05) increase in PPx - Tenr AOIs over GPx-Tenr AOIs.

[0032] FIGs. 8A-8B show ssGSEA on PPx and GPx epithelia (PanCK+ AOIs), wherein the ssGSEA results highlighted a prominent increase in mechanisms associated with proliferation, DNA repair and metabolism within the invasive epithelia, in both GPx and PPx. The AOIs with non-invasive epithelia in both subsets were more prominently associated with signaling cascades in volved in in EMT including WNT Beta catenin, Notch, and KRAS signaling. FIG. 9A shows frequencies of AOIs with each epithelial type captured in the different regions within GPx and PPx. FIG. 9B shows enrichment (Hallmark gene sets) of the genes upregulated within AOIs within invasive epithelia compared to the normal adjacent in GPx(left) and PPx (right panel). FIGs. 9C-9D show enrichment of genes uniquely upregulated in PanCK+ AOIs (left) and TF target enrichment as identified by DecoupleR including STAT3, API TFs, and TP63. FIGs. 9E-9F show a similar analysis on the DEGs unique in PPx AOIs with invasive epithelia, with enrichment of MYC, E2F, HIF1A and NFKB transcription factors.

[0033] FIG. 10 shows boxplots showing expression(log2) of complement genes in GPx and PPx PanCK+ AOIs with atypia and invasive types.Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0034] FIG. 11 shows a network representing the protein-protein interactions captured between genes that were identified to be highly dissimilar between the two ranked feature lists by the CNN as necessary to predict the labels GPx and PPx. The enrichment of this network highlights that the genes are mostly associated with cellular stress response and cell cycle control / DNA damage repair centered on MYC emphasizing the difference in the interactions that exist in epithelia which may contribute to subsequent prognostication.

[0035] DETAILED DESCRIPTION

[0036] Triple-negative breast cancer (TNBC) is a prevalent breast cancer subtype with the lowest 5-year survival. TNBC may be characterized by lack of expression of both estrogen receptor (ER) and progesterone receptor (PR), as well as absence of human epidermal growth factor 2 (HER2) upregulation. Several factors contribute to its treatment response, but the inherent molecular and cellular tumor heterogeneity are increasingly acknowledged as crucial determinants.

[0037] The complement cascade is known to play a paradoxical, often pro-tumorigenic role in triple-negative breast cancer (TNBC), contributing to tumor growth, metastasis, and chemotherapy resistance. While typically part of the innate immune system, complement components are often overexpressed in TNBC, promoting an immunosuppressive microenvironment that enables evasion of immunosurveillance.

[0038] Thus, several methods are described herein, including methods of treating triplenegative breast cancer in a subject, the method including administering a therapeutically effective amount of a composition comprising a complement inhibitor to the subject, wherein the complement inhibitor inhibits one or more of the complement cascade components C1QA, C1QB, C1QC, C1R, CIS, C2, C3, and / or C4B. Also provided herein are methods of improving chemotherapeutic response in a triple negative breast cancer (TNBC) patient with residual disease, the method including (a) administering a therapeutically effective amount of a composition comprising a complement inhibitor to the patient, wherein the complement inhibitor inhibits one or more of the a complement cascade components C1QA, Cl QB, C1QC, C1R, CIS, C2, C3, and / or C4B, and (b) administering a therapeutic effective amount of an immunotherapy or chemotherapy. Also provided herein are methods of identifying a TNBC patient as a candidate for complement C3 inhibitor therapy, the method including (a) obtaining a biological sample from the patient, wherein the biological sample comprises an atypical pre-invasive epithelial region, (b) detecting C3 protein activation in the atypical preAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0039] invasive epithelial region, (c) comparing the C3 protein activation to a reference value derived from a good-prognosis TNBC patient, (d) identifying the patient as a candidate for complement C3 inhibitor therapy based on the C3 protein activation in the pre-invasive epithelial region being elevated relative to the reference value, and (e) administering a complement C3 inhibitor to the identified TNBC patient.

[0040] Various non-limiting aspects of these methods are described herein and can be used in any combination without limitation. Additional aspects of various components of the methods described herein are known in the art.

[0041] It must be noted that, as used in the specification and the appended claims, the singular forms “a,” “an” and “the” include plural referents unless the context clearly dictates otherwise.

[0042] As used herein, the terms “about” and “approximately,” when used to modify an amount specified in a numeric value or range, indicate that the numeric value as well as reasonable deviations from the value known to the skilled person in the art, for example ± 10%, or ± 5%, are within the intended meaning of the recited value.

[0043] Methods of Treating Triple-Negative Breast Cancer

[0044] Provided herein are methods of treating triple-negative breast cancer in a subject that include administering a therapeutically effective amount of a composition comprising a complement inhibitor to the subject, wherein the complement inhibitor inhibits a complement cascade component, and wherein the complement cascade component comprises C1QA, C1QB, C1QC, C1R, CIS, C2, C3, or C4B. In some embodiments, the method further comprises identifying the subject as having (i) less than 1% expression of estrogen and progesterone receptors and / or wild-type expression of HER2 / neu gene compared to reference non-cancerous breast tissue, and / or (ii) elevated complement C3 activation in atypical epithelial breast tissue as compared to reference non-cancerous breast tissue.

[0045] Triple-Negative Breast Cancer

[0046] Triple negative breast cancer (TNBC) refers to a well-defined clinically relevant subtype of breast carcinomas that account for approximately 15% of all breast cancer cases. TNBC tumors score negative (i.e., using conventional histopathology methods and criteria) for expression of estrogen receptor (ER) and progesterone receptor (PR) and do not express amplified levels of HER2 (i.e., the tumor cells are ER-, PR-, HER2-). In someAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0047] embodiments, tumor cells are considered negative for expression of ER and PR if less than 5% (e.g., less than 4%, less than 3%, less than 2%, or less than 1%) of the tumor cell nuclei are stained for ER and PR expression using standard immunohistochemical techniques. In some embodiments, a triple-negative breast cancer tissue is compared to a reference non-cancerous breast tissue, wherein the TNBC tissue has less than 1% expression of estrogen and progesterone receptors and / or wild-type expression of HER2 / neu gene compared to reference non-cancerous breast tissue.

[0048] The expression levels of ER, PR, and HER2 can be determined using methods known in the art, e.g., using standard molecular biological or clinical trial methodology and statistical analysis, or using reference levels obtained from normal (i.e., non-cancerous) tissues. For example, the expression level of ER, PR, and HER2 proteins can be evaluated using methods that include, but are not limited to, Western blot, enzyme linked immunosorbent assay (ELISA), biotin / avidin type assays, protein array detection, radioimmunoassay, immunohistochemistry (IHC), immune-precipitation assay, FACS (fluorescent activated cell sorting), or mass spectrometry. In some embodiments, the expression levels of ER, PR, and HER2 can be determined by immunohistochemistry (IHC) and fluorescent in situ hybridization (FISH).

[0049] Complement Inhibitors

[0050] The complement cascade system refers to a central component of innate immunity comprising an extensive network of plasma and membrane-bound proteins, cofactors, and receptors. Inappropriate complement activation is associated with a variety of pathological conditions, including autoimmune disorders, neurodegenerative diseases, and cancer. The complement cascade system is involved in tumor progression and response to cancer treatment, and local dysregulation of complement proteins in the tumor microenvironment. For example, complement activation induces the production of the highly potent anaphylatoxins C3a and C5a, which have been shown to have profound effects on immunity and on tumor-promoting inflammation. In some embodiments, a complement cascade component is selected from C1QA, C1QB, C1QC, C1R, CIS, C2, C3, and C4B.

[0051] In some embodiments, a method of treating triple-negative breast cancer in a subject includes administering a therapeutically effective amount of a composition comprising a complement inhibitor to the subject, wherein the complement inhibitor inhibits a complement cascade component, and wherein the complement cascade component comprises C1QA,Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0052] CIQB, C1QC, C1R, CIS, C2, C3, or C4B, and identifying the subject as having elevated complement C3 activation in atypical epithelial breast tissue as compared to reference non-cancerous breast tissue. In some embodiments, the subject is identified as having triplenegative breast cancer based on elevated complement C3 activation in atypical epithelial breast tissue from the subject as compared to the level of complement C3 activation in a reference non-cancerous breast tissue.

[0053] Complement inhibitors refer to therapeutic agents that target and block specific components of the complement cascade system (e.g., C3, C5, or CIS). In some embodiments, the complement inhibitor is a C3 inhibitor. In some embodiments, a C3 inhibitor includes (a) a monoclonal antibody that specifically binds to C3 or a C3 cleavage fragment, (b) a compstatin or compstatin analog peptide that inhibits C3 cleavage, (c) a small molecule inhibitor of C3 or C3 convertase, (d) an RNA-based therapeutic targeting C1QA, Cl QB, CIQC, C1R, CIS, C3, or C2, or (e) a C3a receptor (C3aR) inhibitor. In some embodiments, the C3 inhibitor comprises pegcetacoplan or compstatin.

[0054] In some embodiments, a complement C3 inhibitor is capable of reducing, blocking, or neutralizing C3 activation, C3 cleavage, or C3 -mediated downstream signaling. In some embodiments, a C3 inhibitor includes monoclonal antibodies targeting C3 or C3 cleavage fragments. In some embodiments, a C3 inhibitor includes anti-C3 antibodies that block C3 convertase activity or neutralize C3a or C3b. In some embodiments, a C3 inhibitor includes compstatin and compstatin analogs, wherein the compstatin and compstatin analogs are cyclic peptides that bind C3 and inhibit its cleavage by C3 convertase. In some embodiments, a C3 inhibitor includes AMY-101 (Cp40) and pegcetacoplan (APL-2). In some embodiments, a C3 inhibitor includes small molecule inhibitors of C3, C3 convertase components (C2, C4B, C1R, CIS), or C3a receptor (C3aR). In some embodiments, a C3 inhibitor includes RNA-based therapeutics (e.g., siRNA, antisense oligonucleotides) targeting C3 or upstream complement regulators C1QA, Cl QB, C1QC, C1R, CIS, or C2. In some embodiments, a C3 inhibitor includes inhibitors of C3a receptor (C3aR) signaling, wherein the C3 inhibitor blocks downstream anaphylatoxin-mediated effects on cell migration and EMT. In some embodiments, a C3 inhibitor includes agents that selectively promote C3b-mediated opsonization while suppressing C3a-mediated anaphylatoxin signaling.

[0055] In some embodiments, the subject has been diagnosed with triple-negative breast cancer. In some embodiments, the subject has been diagnosed with triple-negative breast cancer and previously administered a neoadjuvant chemotherapy or an adjuvant chemotherapy. In some embodiments, the subject has residual disease, wherein residualAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0056] disease refers to the presence of residual cancer cells or tumor-derived fragments that persist after treatment (e.g., neoadjuvant chemotherapy or adjuvant chemotherapy).

[0057] In some embodiments, the complement inhibitor is administered by intravenous (systemic) administration, subcutaneous injection, or localized intratumoral or intratissue injection into pre-invasive lesions. In some embodiments, the administering of the complement inhibitor comprises intravenous administration, subcutaneous administration, intraperitoneal administration, rectal administration, oral administration, or any combinations thereof. Administration refers to the administration of a composition to a subject or system to achieve delivery of an agent that is, or is included in, the composition. Those of ordinary skill in the art will be aware of a variety of routes that may, in appropriate circumstances, be utilized for administration to a subject, for example a human. For example, in some embodiments, administration may be ocular, oral, parenteral, topical, etc. In some embodiments, administration may be enteral, intra-arterial, intradermal, intragastric, intramedullary, intramuscular, intranasal, intraperitoneal, intrathecal, intravenous, intraventricular, within a specific organ (e. g. intrahepatic), mucosal, nasal, oral, rectal, subcutaneous, sublingual, topical, tracheal (e.g., by intratracheal instillation), vaginal, vitreal, or any combinations thereof. In some embodiments, administration may involve only a single dose. In some embodiments, administration may involve application of a fixed number of doses. In some embodiments, administration may involve dosing that is intermittent (e.g., a plurality of doses separated in time) and / or periodic (e.g., individual doses separated by a common period of time) dosing. In some embodiments, administration may involve continuous dosing (e.g., perfusion) for at least a selected period of time.

[0058] In some embodiments, the administration of the complement inhibitor is initiated at initial diagnosis of TNBC. In some embodiments, the complement inhibitor is administered prior to neoadjuvant chemotherapy. In some embodiments, the complement inhibitor is administered concurrently with neoadjuvant chemotherapy. In some embodiments, the complement inhibitor is administered prior to adjuvant therapy. In some embodiments, the complement inhibitor is administered concurrently with adjuvant therapy. In some embodiments, the complement inhibitor is administered after confirmation of residual disease after primary therapy.

[0059] Pharmaceutical CompositionsAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0060] The methods described herein include the administration of pharmaceutical compositions and formulations comprising a complementary inhibitor. In some embodiments, the complement inhibitor is a C3 inhibitor.

[0061] In some embodiments, the compositions are formulated with a pharmaceutically acceptable carrier. The pharmaceutical compositions and formulations can be administered intravenously, subcutaneously, or by oral administration. The pharmaceutical compositions can be formulated in any way and can be administered in a variety of unit dosage forms depending upon the condition or disease and the degree of illness, the general medical condition of each patient, the resulting preferred method of administration and the like.

[0062] Details on techniques for formulation and administration of pharmaceuticals are well described in the scientific and patent literature, see, e.g., Remington: The Science and Practice of Pharmacy, 21st ed., 2005.

[0063] In some embodiments, the complement inhibitor can be administered alone or as a component of a pharmaceutical formulation (composition). The compounds may be formulated for administration, in any convenient way for use in human or veterinary medicine. Wetting agents, emulsifiers and lubricants, such as sodium lauryl sulfate and magnesium stearate, as well as coloring agents, release agents, coating agents, sweetening, flavoring and perfuming agents, preservatives and antioxidants can also be present in the compositions.

[0064] Pharmaceutical formulations can be prepared according to any method known to the art for the manufacture of pharmaceuticals. Such drugs can contain sweetening agents, flavoring agents, coloring agents and preserving agents. A formulation can be admixtured with nontoxic pharmaceutically acceptable excipients which are suitable for manufacture. Formulations may comprise one or more diluents, emulsifiers, preservatives, buffers, excipients, etc. and may be provided in such forms as liquids, powders, emulsions, lyophilized powders, sprays, creams, lotions, controlled release formulations, tablets, pills, gels, on patches, in implants, etc.

[0065] Pharmaceutical formulations for oral administration can be formulated using pharmaceutically acceptable carriers well known in the art in appropriate and suitable dosages. Such carriers enable the pharmaceuticals to be formulated in unit dosage forms as tablets, pills, powder, dragees, capsules, liquids, lozenges, gels, syrups, slurries, suspensions, etc., suitable for ingestion by the patient. Pharmaceutical preparations for oral use can be formulated as a solid excipient, optionally grinding a resulting mixture, and processing the mixture of granules, after adding suitable additional compounds, if desired, to obtain tabletsAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0066] or dragee cores. Suitable solid excipients are carbohydrate or protein fillers include, e.g., sugars, including lactose, sucrose, mannitol, or sorbitol; starch from corn, wheat, rice, potato, or other plants; cellulose such as methyl cellulose, hydroxypropylmethyl-cellulose, or sodium carboxy-methylcellulose; and gums including arabic and tragacanth; and proteins, e.g., gelatin and collagen. Disintegrating or solubilizing agents may be added, such as the crosslinked polyvinyl pyrrolidone, agar, alginic acid, or a salt thereof, such as sodium alginate. Push-fit capsules can contain active agents mixed with a filler or binders such as lactose or starches, lubricants such as talc or magnesium stearate, and, optionally, stabilizers. In soft capsules, the active agents can be dissolved or suspended in suitable liquids, such as fatty oils, liquid paraffin, or liquid polyethylene glycol with or without stabilizers.

[0067] Aqueous suspensions can contain an active agent (e.g., complement inhibitor) in admixture with excipients suitable for the manufacture of aqueous suspensions, e.g., for aqueous intradermal injections. Such excipients include a suspending agent, such as sodium carboxymethylcellulose, methylcellulose, hydroxypropylmethylcellulose, sodium alginate, polyvinylpyrrolidone, gum tragacanth and gum acacia, and dispersing or wetting agents such as a naturally occurring phosphatide (e.g., lecithin), a condensation product of an alkylene oxide with a fatty acid (e.g., polyoxyethylene stearate), a condensation product of ethylene oxide with a long chain aliphatic alcohol (e.g., heptadecaethylene oxycetanol), a condensation product of ethylene oxide with a partial ester derived from a fatty acid and a hexitol (e.g., polyoxyethylene sorbitol mono-oleate), or a condensation product of ethylene oxide with a partial ester derived from fatty acid and a hexitol anhydride (e.g., polyoxyethylene sorbitan mono-oleate). The aqueous suspension can also contain one or more preservatives such as ethyl or n-propyl p-hydroxybenzoate, one or more coloring agents, one or more flavoring agents and one or more sweetening agents, such as sucrose, aspartame or saccharin. Formulations can be adjusted for osmolarity.

[0068] In some embodiments, oil-based pharmaceuticals are used for administration of nucleic acid sequences. Oil-based suspensions can be formulated by suspending an active agent in a vegetable oil, such as arachis oil, olive oil, sesame oil or coconut oil, or in a mineral oil such as liquid paraffin; or a mixture of these. See e.g., U.S. Patent No. 5,716,928 describing using essential oils or essential oil components for increasing bioavailability and reducing inter- and intra-individual variability of orally administered hydrophobic pharmaceutical compounds (see also U.S. Patent No. 5,858,401). The oil suspensions can contain a thickening agent, such as beeswax, hard paraffin or cetyl alcohol. Sweetening agents can be added to provide a palatable oral preparation, such as glycerol, sorbitol orAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0069] sucrose. These formulations can be preserved by the addition of an antioxidant such as ascorbic acid. As an example of an injectable oil vehicle, see Minto (1997) J. Pharmacol. Exp. Ther. 281:93-102.

[0070] Pharmaceutical formulations can also be in the form of oil-in-water emulsions. The oily phase can be a vegetable oil or a mineral oil, described above, or a mixture of these. Suitable emulsifying agents include naturally-occurring gums, such as gum acacia and gum tragacanth, naturally occurring phosphatides, such as soybean lecithin, esters or partial esters derived from fatty acids and hexitol anhydrides, such as sorbitan mono-oleate, and condensation products of these partial esters with ethylene oxide, such as polyoxyethylene sorbitan mono-oleate. The emulsion can also contain sweetening agents and flavoring agents, as in the formulation of syrups and elixirs. Such formulations can also contain a demulcent, a preservative, or a coloring agent. In alternative embodiments, these injectable oil-in-water emulsions comprise a paraffin oil, a sorbitan monooleate, an ethoxylated sorbitan monooleate and / or an ethoxylated sorbitan trioleate.

[0071] In some embodiments, the pharmaceutical compounds can also be delivered as microspheres for slow release in the body. For example, microspheres can be administered via intradermal injection of drug which slowly release subcutaneously; see Rao (1995) J. Biomater Sci. Polym. Ed. 7:623-645; as biodegradable and injectable gel formulations, see, e.g., Gao (1995) Pharm. Res. 12:857-863 (1995); or, as microspheres for oral administration, see, e.g., Eyles (1997) J. Pharm. Pharmacol. 49:669-674.

[0072] In some embodiments, the pharmaceutical compounds can be parenterally administered, such as by intravenous (IV) administration or administration into a body cavity or lumen of an organ. These formulations can comprise a solution of active agent dissolved in a pharmaceutically acceptable carrier. Acceptable vehicles and solvents that can be employed are water and Ringer’s solution, an isotonic sodium chloride. In addition, sterile fixed oils can be employed as a solvent or suspending medium. For this purpose any bland fixed oil can be employed including synthetic mono- or diglycerides. In addition, fatty acids such as oleic acid can likewise be used in the preparation of injectables. These solutions are sterile and generally free of undesirable matter. These formulations may be sterilized by conventional, well known sterilization techniques. The formulations may contain pharmaceutically acceptable auxiliary substances as required to approximate physiological conditions such as pH adjusting and buffering agents, toxicity adjusting agents, e.g., sodium acetate, sodium chloride, potassium chloride, calcium chloride, sodium lactate and the like. The concentration of active agent in these formulations can vary widely, and will be selected primarily based onAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0073] fluid volumes, viscosities, body weight, and the like, in accordance with the particular mode of administration selected and the patient's needs. For IV administration, the formulation can be a sterile injectable preparation, such as a sterile injectable aqueous or oleaginous suspension. This suspension can be formulated using those suitable dispersing or wetting agents and suspending agents. The sterile injectable preparation can also be a suspension in a nontoxic parenterally-acceptable diluent or solvent, such as a solution of 1,3-butanediol. The administration can be by bolus or continuous infusion (e.g., substantially uninterrupted introduction into a blood vessel for a specified period of time).

[0074] In some embodiments, the pharmaceutical compounds and formulations can be lyophilized. Stable lyophilized formulations comprising an inhibitory nucleic acid can be made by lyophilizing a solution comprising a pharmaceutical as described herein and a bulking agent, e.g., mannitol, trehalose, raffinose, and sucrose or mixtures thereof. A process for preparing a stable lyophilized formulation can include lyophilizing a solution about 2.5 mg / mL protein, about 15 mg / mL sucrose, about 19 mg / mL NaCl, and a sodium citrate buffer having a pH greater than 5.5 but less than 6.5. See, e.g., U.S. 20040028670.

[0075] The compositions and formulations can be delivered by the use of liposomes. By using liposomes, particularly where the liposome surface carries ligands specific for target cells, or are otherwise preferentially directed to a specific organ, one can focus the delivery of the active agent into target cells in vivo. See, e.g., U.S. Patent Nos. 6,063,400; 6,007,839; Al-Muhammed (1996) J. Microencapsul. 13:293-306; Chonn (1995) Curr. Opin. Biotechnol. 6:698-708; Ostro (1989) Am. J. Hosp. Pharm. 46:1576-1587. Liposomes are unilamellar or multilamellar vesicles that have a membrane formed from a lipophilic material and an aqueous interior that contains the composition to be delivered. Cationic liposomes are positively charged liposomes that are believed to interact with negatively charged DNA molecules to form a stable complex. Liposomes that are pH-sensitive or negatively-charged are believed to entrap DNA rather than complex with it. Both cationic and noncationic liposomes have been used to deliver DNA to cells.

[0076] Liposomes can also include “sterically stabilized” liposomes, i.e., liposomes comprising one or more specialized lipids. When incorporated into liposomes, these specialized lipids result in liposomes with enhanced circulation lifetimes relative to liposomes lacking such specialized lipids. Examples of sterically stabilized liposomes are those in which part of the vesicle-forming lipid portion of the liposome comprises one or more glycolipids or is derivatized with one or more hydrophilic polymers, such as aAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0077] polyethylene glycol (PEG) moiety. Liposomes and their uses are further described in U.S. Pat. No. 6,287,860.

[0078] The amount of pharmaceutical composition adequate to accomplish this is a therapeutically effective dose. The dosage schedule and amounts effective for this use, i.e., the dosing regimen, will depend upon a variety of factors, including the stage of the disease or condition, the severity of the disease or condition, the general state of the patient's health, the patient’s physical status, age and the like. In calculating the dosage regimen for a patient, the mode of administration also is taken into consideration.

[0079] The dosage regimen also takes into consideration pharmacokinetics parameters well known in the art, i.e., the active agents’ rate of absorption, bioavailability, metabolism, clearance, and the like (see, e.g., Hidalgo- Aragones (1996) J. Steroid Biochem. Mol. Biol.

[0080] 58:611-617; Groning (1996) Pharmazie 51:337-341; Fotherby (1996) Contraception 54:59-69; Johnson (1995) J. Pharm. Sci. 84:1144-1146; Rohatagi (1995) Pharmazie 50:610-613; Brophy (1983) Eur. J. Clin. Pharmacol. 24:103-108; Remington: The Science and Practice of Pharmacy, 21st ed., 2005). The state of the art allows the clinician to determine the dosage regimen for each individual patient, active agent and disease or condition treated. Guidelines provided for similar compositions used as pharmaceuticals can be used as guidance to determine the dosage regiment, i.e., dose schedule and dosage levels, administered practicing the methods as described herein are correct and appropriate.

[0081] Single or multiple administrations of formulations can be given depending on for example: the dosage and frequency as required and tolerated by the patient, the degree and amount of therapeutic effect generated after each administration (e.g., effect on tumor size or growth), and the like. The formulations should provide a sufficient quantity of active agent to effectively treat, prevent or ameliorate conditions, diseases or symptoms.

[0082] Combination Therapies

[0083] In some embodiments, the complement inhibitor is administered alone as a monotherapy. In some embodiments, the complement inhibitor is administered in combination with one or more standard-of-care treatments for TNBC.

[0084] In some embodiments, the methods described herein further include co-administering an anti-cancer treatment. In some embodiments, the complement inhibitors can be coadministered with an anti-cancer treatment, wherein the anti-cancer treatment includes platinum-based chemotherapy (e.g., carboplatin, cisplatin), taxane-based chemotherapy (e.g.,Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0085] paclitaxel, nab-paclitaxel), immune checkpoint inhibitors (e.g., anti-PD-1, anti-PD-Ll agents), PARP inhibitors, and antibody-drug conjugates. In some embodiments, the anticancer treatment includes a therapeutic a-lactalbumin vaccine, a platinum-based chemotherapy, a taxane-based chemotherapy, an immune checkpoint inhibitor, a PARP inhibitor, an antibody-drug conjugate, adoptive cell therapy (ACT), or any combinations thereof. In some embodiments, the anti-cancer treatment can include chemotherapy, radiation therapy, surgery, or immunotherapy. In some embodiments, the anti-cancer treatment can include ionizing radiation, a chemotherapeutic agent, a therapeutic antibody, a checkpoint inhibitor, or any combination thereof. In some embodiments, the anti-cancer treatment comprises a chemotherapeutic agent, wherein the chemotherapeutic agent can refer to one or more pro-apoptotic, cytostatic and / or cytotoxic agents, for example specifically including agents utilized and / or recommended for use in treating one or more diseases, disorders or conditions associated with undesirable cell proliferation. In some embodiments, chemotherapeutic agents are useful in the treatment of cancer. In some embodiments, a chemotherapeutic agent may be or comprise one or more alkylating agents, one or more anthracyclines, one or more cytoskeletal disruptors (e.g. microtubule targeting agents such as taxanes, maytansine and analogs thereof, of), one or more epothilones, one or more histone deacetylase inhibitors HDACs), one or more topoisomerase inhibitors (e.g., inhibitors of topoisomerase I and / or topoisomerase II), one or more kinase inhibitors, one or more nucleotide analogs or nucleotide precursor analogs, one or more peptide antibiotics, one or more platinum-based agents, one or more retinoids, one or more vinca alkaloids, and / or one or more analogs of one or more of the following (i.e., that share a relevant anti-proliferative activity). In some embodiments, a chemotherapeutic agent may be utilized in the context of an antibody-drug conjugate. In some embodiments, the chemotherapeutic agent comprises vincristine, prednisone, dexamethasone, busulfan, cisplatin, carboplatin, paclitaxel, docetaxel, nab-paclitaxel, altretamine, capecitabine, cyclophosphamide, etoposide (vp-16), gemcitabine, ifosfamide, irinotecan (cpt-11), liposomal doxorubicin, melphalan, pemetrexed, topotecan, vinorelbine, goserelin, leuprolide, tamoxifen, letrozole, anastrozole, exemestane, bevacizumab, olaparib, rucaparib, niraparib, nivolumab, pembrolizumab, durvalumab, atezolizumab, radioisotopes, monomethyl auristatin E (MMAE; e.g., vedotin), calicheamicins, deruxtecan, DM1, and any combinations thereof.

[0086] Therapeutic ApplicationsAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0087] Provided herein are methods of improving chemotherapeutic response in a triple negative breast cancer (TNBC) patient with residual disease, the method including (a) administering a therapeutically effective amount of a composition comprising a complement inhibitor to the patient, wherein the complement inhibitor inhibits a complement cascade component, and wherein the complement cascade component comprises C1QA, Cl QB, C1QC, C1R, CIS, C2, C3, or C4B, and (b) administering a therapeutic effective amount of an immunotherapy or chemotherapy.

[0088] Also provided herein are methods of identifying a TNBC patient as a candidate for complement C3 inhibitor therapy, the method including (a) obtaining a biological sample from the patient, wherein the biological sample comprises an atypical pre-invasive epithelial region, (b) detecting C3 protein activation in the atypical pre-invasive epithelial region, (c) comparing the C3 protein activation to a reference value derived from a good-prognosis TNBC patient, (d) identifying the patient as a candidate for complement C3 inhibitor therapy based on the C3 protein activation in the pre-invasive epithelial region being elevated relative to the reference value, and (e) administering a complement C3 inhibitor to the identified TNBC patient. In some embodiments, the detecting C3 protein activation comprises immunohistochemical staining the biological sample with an anti-C3 or anti-C3b antibody. In some embodiments, the detecting step (b) further comprises quantifying a C3-positive cell fraction by digital pathology image analysis.

[0089] In some embodiments, a biological sample includes tissue samples (e.g., breast tissue, biopsy specimens), fluid samples (e.g., blood, serum, plasma, urine, saliva, sputum, and cerebrospinal fluid), and solid samples (e.g. stool) can be obtained from a subject. In some embodiments, the biological sample includes a breast tissue sample. In some embodiments, the biological sample includes a breast tissue sample from a TNBC patient. In some embodiments, the biological sample includes a non-cancerous breast tissue sample.

[0090] In some embodiments, a subject is a human. In some embodiments, a subject is a TNBC patient. In some embodiments, the TNBC patient is categorized as a good-prognosis (GPx) TNBC patient, wherein good prognosis in cancer indicates a high likelihood of successful treatment, remission, or long-term survival. In some embodiments, the TNBC patient is categorized as a prognosis-poor (PPx) TNBC patient, wherein a poor cancer prognosis indicates that the disease is advanced, aggressive, or difficult to control, suggesting a lower likelihood of long-term survival. In some embodiments, the TNBC patient is identified as a candidate for a complement C3 inhibitor therapy, wherein the complement C3 inhibitor includes (a) a monoclonal antibody that specifically binds to C3 or a C3 cleavageAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0091] fragment, (b) a compstatin or compstatin analog peptide that inhibits C3 cleavage, (c) a small molecule inhibitor of C3 or C3 convertase, (d) an RNA-based therapeutic targeting C1QA, C1QB, C1QC, C1R, CIS, C3, or C2, or (e) a C3a receptor (C3aR) inhibitor.

[0092] Assays and Methods

[0093] The present disclosure provides sophisticated spatio-molecular profiling assays designed to identify aggressive hallmarks in triple-negative breast cancer (TNBC) and pre-cancerous lesions. These methods leverage high-throughput spatial transcriptomics and precision immunohistochemistry to resolve cellular interactions that are physically unobservable via standard bulk profiling. By integrating spatial context with molecular state, these assays enable a granular understanding of the tumor ecosystem that can accurately inform prognosis and clinical management.

[0094] In one embodiment, the method comprises providing a sample of atypical breast tissue from a subject having TNBC and assaying that sample to determine a level of complement C3 protein activation. C3 activation serves as a critical, early hallmark of aggressiveness in poor-prognosis epithelia, often surfacing before a lesion has fully transitioned to an invasive state. The discovery of differential C3 signaling suggests that certain precancerous lesions are “primed” for rapid progression, a state that cannot be detected by standard morphological review alone. The determination of C3 activation can be performed via specialized immunohistochemical (IHC) staining of the biological sample. Formalin-fixed paraffin-embedded (FFPE) tissue sections, cut at a thickness of approximately 5 pm, can be subjected to heat-induced antigen retrieval using a standard decloaking solution at approximately 95°C. Following treatment with hydrogen peroxide to quench endogenous activity, the sections can be incubated with a highly specific primary antibody, such as an anti-C3 or anti-C3b antibody. Visualization of the staining can be typically achieved using a DAB chromogen kit and counterstaining with hematoxylin to provide cellular context. The fraction of positive cells is then quantified across regions of interest (ROIs), allowing for the statistical identification of aggressive disease states relative to good-prognosis controls.

[0095] The disclosure further provides methods for risk-stratifying pre-cancerous breast lesions by detecting anaphylatoxin-mediated priming in atypical epithelia. This process involves identifying unique ligand-receptor interactions, specifically C3-CD81 and C3-LRP1, which have been found to correlate with poor-prognosis transition and aggressive disease behavior. The C3-CD81 and C3-LRP1 interactions are indicative of a mesenchymalAtorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0096] transition, where precancerous epithelial cells gain increased migratory capacity. These high-risk interactions can be identified using spatial transcriptomic profiling to determine their precise localization within specific spatial origins of the tissue, categorized as the tumor center region or the tumor edge region. The spatial origin of these markers is clinically significant; in poor-prognosis tumors, the tumor center often exhibits distinct metabolic stress, increased immune exhaustion (e.g., PD-1 expression), and localized immunosuppression compared to the tumor periphery. Upon identification of these spatially localized interactions at the tumor center or edge, a C3 -directed therapeutic agent can be administered to the subject to treat high-risk pre-cancerous lesions and potentially halt their transition to aggressive disease. In one embodiment, the spatial profiling utilized herein is Digital Spatial Profiling (DSP), which enables targeted, high-plex interrogation of distinct tissue compartments. This method comprises several integrated technical steps to ensure structural precision and biological accuracy, including hybridizing the FFPE tissue sample with a broad atlas of profiling probes, such as the Whole Transcriptome Atlas comprising over 18,000 probes. Each probe consists of an indexing oligonucleotide attached to a targetbinding moiety via a UV-photocleavable linker. ROIs of approximately 300 pm in diameter are identified by a pathologist using high-resolution immunofluorescent (IF) imaging.

[0097] Standard fluorescent markers used for this selection include DAPI for nuclei, CD3 for T cells, CD45 for pan-immune cells, and Pan-Cytokeratin (PanCK) for identifying epithelial cells. A digital micromirror device (DMD) is utilized to segment each ROI into specific areas of illumination (AOIs). This allows for the physical isolation of atypical epithelia (PanCK+ segments) from the surrounding tumor microenvironment (PanCK- segments) with approximately 1 pm precision by directing UV illumination specifically onto the desired compartments. The indexing oligonucleotides are released from the atypical epithelia segment via the directed UV illumination. These released tags are collected into a multi-well plate, pooled, and subjected to Next-Generation Sequencing (NGS). The resulting sequencing data provides a spatially resolved transcriptomic profile of the atypical epithelia, enabling the detection of specific signatures such as the C3-CD81 and C3-LRP1 interactions.

[0098] Machine Learning and Deep Learning Frameworks

[0099] The present disclosure employs advanced machine learning and deep learning methodologies to decode the high-dimensional spatio-molecular data generated from triplenegative breast cancer (TNBC) samples. These frameworks are designed to identify subtleAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0100] molecular shifts and establish robust prognostic signatures that are often obscured in standard differential expression analyses.

[0101] To identify a unified epithelial signature predictive of patient prognosis, a deeplearning approach utilizing convolutional neural networks (CNN) can be implemented. The CNN architecture can include parallel convolutional towers, each utilizing a ID convolutional layer followed by LeakyReLU activation and max-pooling (kernel size 2). The resulting outputs can be flattened, concatenated, and processed through fully connected layers, culminating in a dense layer with 128 units and a final two-node output for prognostic classification (GPx vs. PPx). The model can be trained using a binary cross-entropy loss function with an Adagrad optimizer. To ensure generalizability, 100 independent training iterations can be performed using random train-test splits (72% training, 18% validation, 10% testing). Only models achieving an average test accuracy and precision of greater than 90% are retained for downstream feature extraction. Saliency maps are generated for each qualifying model to extract features with high predictive weights (>0.5). These features are then consolidated into ranked gene lists for GPx and PPx using Borda count aggregation. This ranking-based approach captures coordinated shifts in molecular programs rather than relying on the mere presence or absence of unique genes.

[0102] To discretize the functional states of individual cell types within the tumor microenvironment (TME), a machine learning framework called EcoTyper can be utilized and adapted for spatial data. Measurements from segmented PanCK- segments typically represent multiple cell types and hundreds of nuclei. EcoTyper overcomes this by defining distinct functional cell states based on gene expression patterns. The cell state abundances determined by EcoTyper are multiplied by the cell type abundances calculated via spatial deconvolution to obtain a final cell state abundance for each area of illumination (AOI). AOIs are assigned to a specific state if the abundance exceeds the 80th quantile for that cell type. This framework successfully identifies prognostic cell states, such as the CD8+ T-exhausted state (S03) and myofibroblast-like fibroblast states (SOI, S03, S08), which are preferentially increased in poor-prognosis TMEs.

[0103] To characterize the physical and paracrine signaling between the tumor epithelia and the TME, ligand-receptor (L-R) interactions can be inferred using the BulkSignalR package. BulkSignalR utilizes a curated database of 3,249 ligands and receptors and their associated pathways. Interactions are inferred by calculating Spearman correlation coefficients between ligands and receptors (L-R) and receptors and their target genes (R-T). Significant interactions are identified as those with an L-R correlation greater than 0.25 associated with aAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0104] pathway at an FDR (q-value) less than 0.05. This methodology identifies unique interactions in PPx-Atypia, specifically C3-CD81 and C3-LRP1, which indicate the early onset of anaphylatoxin-mediated priming in aggressive disease transition.

[0105] EXAMPLES

[0106] The disclosure is further described in the following examples, which do not limit the scope of the disclosure.

[0107] Materials & Methods

[0108] GeoMX Digital Spatial Profiler

[0109] Spatial transcriptomic profiling was performed using GeoMx-DSPTM (DSP; Nanostring Technologies). Pre-treatment formalin-fixed paraffin-embedded (FFPE) were cut into 10 concurrent 5-mm sections. The first 5 mm section was stained with hematoxylin and eosin (H&E). The second slice was hybridized with the GeoMx Probe Mix for NGS readout (human whole transcriptome Atlas (GeoMx Hu WTA panel, Cat. #: 121401102, Nanostring Technologies) panel as per manufacturer’s protocol. The GeoMx Hu WTA panel includes 18000+ genes with several key genes involved in cancer cell biology and the tumor microenvironment. Slides were stained with GeoMx Solid Tumor TME Morphology Kit (Cat. #: 121300310, Nanostring Technologies) and GeoMx Nuclear Stain Morphology kit (Cat. #: 121300303, Nanostring Technologies) according to the manufacturer’s protocol. Regions of interest (ROI) were selected (diameter of 300 microns, minimum of 3 replicates representative of the tumor edge (including adjacent immune cells), tumor interior, and adjacent non-cancerous tissue. Pan cytokeratin (PanCK) was used to segment each ROI defining specific areas of illumination (AOI) and distinguishing the tumor (PanCK+) from the stromal (PanCK-) component. Stromal areas were further analyzed using the pan-immune marker CD45 and the pan-T-cell marker CD3. Segmentation was conducted on two classes of AOIs within each ROI: PanCK- and PanCK+. For each AOI, probes were collected by the machine and transferred to a PCR plate for library prep using Seq Code primers (GeoMx Seq Code Pack, Cat. #: 121400205-121400206, Nanostring Technologies). Libraries from each AOI were pooled according to their dimension, purified by AMPure XP beads (A63880 Beckman Coulter) clean up, and resuspended in a volume of elution buffer proportional to the number of pooled AOIs. Libraries were assessed using an Agilent Bioanalyzer, then diluted to 1.6 pmol / L and sequenced (paired-end 2 x 27) on Illumina NextSeq2000, with a coverageAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0110] of 100 reads. FastQ files were uploaded to BaseSpace Illumina and converted into DCC files by GeoMxNGSPipeline software and utilized for all downstream processing.

[0111] GeoMX Data QC and Processing

[0112] GeomxTools was used for quality control (QC) and downstream analysis of the DCC files in R. All samples were pooled prior to QC. QC was performed in accordance with GeomxTools developer recommendations (min Nuclei = 40, min Area = 1000, LOQ= 2). Gene and sample filtering were chosen to be at 5% at which a reasonable number of AOIs and probes were retained. Samples that were marked as PTS, were extracted and re-processed independently to allow for time series analysis. In this case, the segment filtering threshold was set at 2%, while gene filtering thresholds were still maintained at 5%. The counts were normalized using Q3 normalization as recommended by GeoMx after gene filtering and transformed into log values for downstream analysis, when necessary(log q). The comparison of interest dictated whether p-values or adjusted p-values were chosen for thresholding and identifying differentially expressed genes. Differential analysis was performed using a linear mixed model. Comparisons of interest once again dictated whether an LMM was performed with a random slope (across slides) or with a random slope and intercept (within slide).

[0113] Differential analysis using Linear Mixed Models

[0114] The differential analysis was performed using the linear mixed models (LMM) as prescribed by the GeomxTools vignette. LMM allows users to account for the subsampling per tissue; in other words, adjusting for the fact that the multiple regions of interest captured in a slide / tissue section are not independent observations. As indicated in the vignette of the GeomxTools two LMM models are possible when performing differential analysis, (i) within-slide comparison, with a random slope: a within-slide comparison calls for comparisons of regions of interest within a slide, with each slide present in the sample set serving as a replicate. For example, in instances where it was assessed if there exists a difference in invasive epithelia identified at the center of the tumor to the edge of the tumors, a within-slide analysis was used. In this example, only slides that have captured invasive epithelia from both center and edge were considered. Similarly, this approach was utilized to also compare Invasive vs normal-adjacent epithelia and invasive vs atypia in both prognostic groups (ii) across slide comparison, without a random slope. For instance, when all invasive epithelia was compared coming from two distinct population cohorts say GPx and PPx.Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0115] Differential analysis was performed using the entire feature space (10150) or the NMF reduced feature space.

[0116] Image intensity and spatial entropy analysis

[0117] Images of each ROI and each PanCK segmented AOI region were obtained from the GeoMx DSP instrument using the “ROI Report” extraction. All images were first segmented to obtain only the portion inside the ROI region. The four-color channels were extracted, and an intensity filter was applied to remove the pixels in the lowest 10th percentile for each color channel. For the immune cell intensity analysis, the color channels within each ROI region were analyzed for the PanCK- AOI only. For each color channel, the intensity signal was computed by dividing the fraction of pass-filter pixels for the specified color by the total pass-filter pixels of all colors in the PanCK- AOI. Spatial entropy analysis was performed using the full ROI image. Using the intensity-filtered color pixels, both univariate and bivariate entropy measures were computed. For univariate entropy, Batty’s spatial entropy was performed to assess the region heterogeneity of each color channel in the ROI. For bivariate entropy, different pairs of color channels were used and a marked point pattern dataset was constructed for each color pair to assess the spatial relationship between the immune cells (yellow and red channels) vs epithelial cells (green channel). Spatial entropy was assessed using Altieri’s spatial entropy which was implemented in R using the SpatEntropy package. First, Leibovici's implementation of Shannon’s relative spatial entropy was used to filter out ROIs where color distributions were not explained by their spatial cooccurrences (ROIs with Hreiativ Z) > 0.5 were removed). For the ROIs that passed the spatial entropy filter, Altieri’s spatial entropy was the used with distances of 5, 10, 50, 100, 250, and 500 pixels which were chosen so that the distance steps would span from cell-cell interactions to half the diameter of the ROI. The spatial partial information calculated on each distance step showed a maximum value between 5-50 pixels which corresponds to the distance of ~1 cell diameter or a cell’s nearest neighbors. The primary measure used to quantify the spatial dependence of the color pairs was the spatial mutual information, MKZNVy

[0118] Spatial entropy was assessed using the immunofluorescent image of each ROI by calculating the entropy for two bivariate distributions: CD3 vs PanCK staining and CD45 vs PanCK staining. For each bivariate distribution, both global and scale-dependent measures of entropy were calculated. The global measures, which assessed the entire ROI, included Shannon’s relative Z entropy and spatial mutual information; however, these measures are often insufficient to detect dependencies that only occur at specific length scales. Scale measures,Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0119] which are assessed at different length scales within the spatial distribution, were calculated at pixel lengths of 5, 10, 50, 100, 250, and 500 and were chosen to span distances from cell neighbors (~15pm) to the radius of the ROIs (~150pm). Altieri’s spatial entropy was used to produce both spatial mutual information and scale metrics from which the spatial partial information was used to assess the dependence of the two channels. For each immunofluorescent ROI image, the relevant channels were extracted, and an intensity filter was applied as previously described. The pixel locations from the two channels being assessed were converted to a point pattern dataset and the entropy measures were calculated using the relevant functions from the R SpatEntropy package. To compare the spatial entropy metrics between different sample classifications, a two-sided T-test was performed, and the FDR p-value was assessed for significance at the 0.05 level.

[0120] Non-negative matrix factorization (NMF) analysis

[0121] Non-negative matrix factorization (NMF) is a powerful dimensionality reduction and clustering technique and has proven particularly useful for identifying underlying patterns and structures in high-dimensional non-negative gene expression data. In this study, NMF was applied using the NMF library in R, on log q data generated from PanCK+ segments in GPx and PPx (to mitigate the influence of extreme values and provide a more stable representation of the underlying biology). To ensure the robustness of the NMF model, its stability was assessed using a consensus matrix approach. Specifically, the number of runs (nrun) were set to 5 and tested ranks ranging from 2 to 10. This process was iterated 20 times with different random seeds. The best rank for each seed was determined based on metrics such as cophenetic correlation, dispersion, and silhouette score, ensuring a comprehensive evaluation of model performance. Subsequently, the selected best rank for each random seed was utilized to run the NMF model for 100 iterations (nrun = 100), allowing for further refinement and convergence of the factorization process. The metagenes were extracted using the "extractFeature" function with the method parameter set to "combine" within the NMF library. These metagenes represent a compact and interpretable representation of the original features, capturing the consensus patterns across multiple runs and random seeds. Finally, the union of all 20 seed runs was utilized as the reduced feature space, referred to as “consensus features” for downstream analysis. These consensus features encapsulate the most relevant and stable biological signals present in the dataset, enabling effective interpretation and subsequent analyses.Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0122] Spatial cell-type deconvolution using SpatialDecon, and a TNBC-specific reference matrix The reference matrix readily available for deconvolution of the TME within SpatialDecon vl.l(called SafeTME) is largely derived from the immune and stromal cells derived from PBMCs and flow-sorted tumor cells, a TNBC-specific deconvolution matrix was developed using the “create_profile_matrix” function of the SpatialDecon package. This function requires a cell * gene count matrix, along with the cell annotations as data input. To achieve this, the scRNAseq data was downloaded from a recently published study specific to breast cancers (downloaded from singlecell.broadinstitute.org / single_cell / study / SCP1039 / a-single-cell-and-spatially-resolved-atlas-of-human-breast-cancers#study-download) containing 100064 cells. From this, all single cells were extracted from all TNBC (42512 cells). Since there was specific interest in generating a reference matrix for deconvoluting the PanCK-segments of the ROIs, epithelial cells (cancer and normal) were excluded bringing down the # of single cells used from this study to 30719. As provided in the supplementary of the original study, the cellType major and cellType minor information was retained across 30719 cells. Focus was placed on 7 major types as captured by the cellType major information, including B -cells, T-cells, Endothelial, Plasmablasts, perivascular-like (PVL) (with increased expression of MCAM / CD146, ACTA2 and PDGFRB), Myeloid, and CAFs (renamed as Fibroblast / Mesenchymal)) to drive the generation of the reference matrix. As previously described, these ~30K cells were further processed and filtered using Seurat v4.3.0.1, resulting in a filtered data matrix with 11951 features x 26451 cells. The TNBC specific reference profile matrix using the “create_profile_matrix” command (minCellNum=10, scalingFactor=l and minGenes=100) in SpatialDecon vl.l, with the expression of 1066 / 11951 genes identified as breast cancer-specific serving as input. This reduced gene list was obtained by amalgamating the genes in the safeTME matrix, the genes expressed in breast TME, and 135 breast-specific genes, belonging to the 7 major cell types. SpatialDecon vl.l was subsequently also utilized to generate the spatial deconvolution of the TME using the “RunSpatialDecon” function and the reference matrix generated above as profile matrix. Proportions of cell counts as well as the abundance estimates were extracted and plotted, as outlined in the vignette.

[0123] EcoTyper

[0124] TPM transformed filtered probe data (before Q3 normalization) for each segment and served as input to EcoTyper. The cell states were computed using the ecotyper recovery bulk.R script, following their vignettes. Given the nature of the GeoMX data, the cell stateAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0125] abundances determined by EcoTyper were further multiplied, with the cell type abundances for each of the overlapping cell types to get a final cell state abundance for each AOI. AOIs were assigned to the cell state if the state exceeded the 80th quantile value of the final cell state abundance computed for each cell type. The description of the cell states was obtained, and the set of all cell state markers used for generating the violin plots, was obtained from the authors.

[0126] BulkSignalR

[0127] BulkSignalR was utilized to infer the ligand-receptor interactions (LRI) in the TME of the Tenr subset. BulkSignalR utilizes a curated database of 3249 ligands, receptors, their associated pathways and downstream target genes as the background from which to make the inferences. For the L-R interactions captured in the TME, Q3-normalized data from all 10150 features across 61 GPx Tenr AOIs and 40 PPx Tenr was utilized in this analysis (set.

[0128] Seed(123), min. count = 0 and prop = 0). For the L-R interactions described in atypia, Q3-normalized data was used from all 10150 features, in the ROIs with atypic epithelia. LRI pathways were included with a minimum of 4 downstream target genes present in the gene list (min. positive = 4). BulkSignalR requires null distributions of Spearman correlation coefficients between a ligand and receptor (L-R) and receptor and its target genes (R-T). The Gaussian kernel-based empirical model was identified as the appropriate statistical model for L-R and R-T correlations in both GPx and PPx. When inferring significant LRI, those with L-R correlation > 0.25 and association with a pathway with q < 0.05 were included in the final list for GPx and PPx. Pathway significance was calculated as a combination of L-R and R-T correlation significance, and the most significant pathway for each unique L-R was used.

[0129] Enrichment and TF target analysis

[0130] ssGSEA was applied to expression data from each AOI, separately for the PanCK+ and PanCK- segments, using the Hallmark gene set from MSigDB using the GenePattern software. Additionally, all enrichment presented in this manuscript were performed and plotted using either clusterProfiler or enrichR. TF target analysis was performed using DecoupleR . String-db vl2 was utilized to identify protein-protein interactions (medium confidence) in gene sets of interest. All the networks were then exported to Cystoscape and visualized in Cystoscape. Enrichment of the String PPI was performed within Cystoscape using GO BP, Wiki pathways, and Reactome pathways categories. The enrichment reported was identified with a fdr<10-4 or better.Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0131] Convolution neural networks

[0132] NMF-reduced gene expression data from pooled GPx and PPx AOIs was utilized to identify a gene signature defining the epithelial state for each prognostic group, using convolution neural networks. The data was split into distinct subsets, consisting of training, validation, and testing sets using a stratified splitting strategy. The use of such a strategy guaranteed a balanced representation of GPx and PPx classes in each group. Specifically, 10% of the data was allocated for testing, 18% for validation, and reserved the remaining 72% for training. Furthermore, to ensure reproducibility, a random seed mechanism was incorporated during the allocation of data, ranging from 1 to 500, facilitating consistent data partitioning across iterations. Once the data was appropriately partitioned, each sample’s original ID structure was reshaped to 2D. The CNN was instantiated with this restructured data to allow for the use of traditional 2D CNN layers. The Borda count method was utilized for ranking genes / features as obtained from the saliency maps. Borda count is a ranked polling method that takes the position of each element and assigns a value for its corresponding position. Elements lower in ranking receive fewer “points” and are ranked less highly in the aggregate ranking. The concordance between the two ranked lists generated was assessed, utilizing the aggregateRanks function (method= “Stuart”) available via the RobustRankAggreg vl.2.1 package in R. Genes with a score of 0.05 were identified as similarly ranked, while genes with a score >0.5 were identified as being most dissimilar between the two gene lists.

[0133] Example 1 - Study design and development of a ROI annotation strategy

[0134] Utilizing GeoMX® DSP the cellular heterogeneity was spatially profiled in 32 TNBC samples with documented differential prognoses (retrospective). TNBC samples were defined by <1% protein expression of ER and PR by immunohistochemistry (IHC) and no evidence of HER2 / neu gene overexpression or amplification by combining IHC and FISH. All samples were reviewed by a single dedicated pathologist (D.S.) FFPE breast tissues of 17 treatment-naive women GPx, and 15 treatment-naive PPx women were chosen from the Louisiana Tumor Registry (IRB#1169). In GPx, 8 women self-identified race as Black, and 9 as White, while in PPx 5 women self-identified as Black and 10 as White (FIG. 1A). The samples were H&E stained as well as subject to immunofluorescence (IF) imaging using 4 IF antibodies to identify various cell types including DAPI (blue) to identify nuclei, CD3 (yellow) for T-cells, CD45 (red) pan-immune cells, and pan-cytokeratin, CK15 (green) for epithelial cells (FIG. IB). Regions of illumination / interest (ROIs, diameter=300 microns) were selected byAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0135] pathologist / trained clinician at the tumor interface, within the center of the tumor (with or without immune cells), and when available, regions distant from the center and edge (isolated foci). Regions with morphologically normal and / or precancerous changes in the epithelial cells were also selected as ROIs. GeoMX additionally offers the ability to segment areas of illumination / interest (AOIs) within ROIs using fluorescently labeled antibodies to independently transcriptionally profile each segmented area. Each ROI was segmented on the cytokeratin antibody (CK15) to identify epithelial cells within each ROI, henceforth referred to as the PanCK+ AOI / segment, and all the negative space captured with the ROI as PanCK-AOI / segment representing the surrounding tumor milieu. Exemplar ROIs with varying immune infiltration, and pathological epithelial states, taken from different regions of the excisions are shown in FIG. 1C.

[0136] A total of 717 regions of interest (ROIs, 345 PPx and 372 GPx) were identified and a total of 1424 AOIs (737 GPx and 687 PPx, FIG. ID) were captured from the 32 samples, with uniform distribution of segments across samples (FIG. 6A). To meaningfully, consistently and systematically document and categorize the ROIs within the slides and across the samples, a unique annotation schema was developed and adopted to be used in the downstream analysis. Each of the 717 ROIs were annotated using the S-E-R-P-L-D-N annotation schema. Briefly, each ROI was annotated by a pathologist / clinician from IF images, with the following descriptors: Sample (S), Region (R), Epithelial type (E), Immune cell localization (L), Immune cell population (P), Immune cell distribution (D) and Immune cell number (N). These descriptors were used to uniquely classify each ROI for downstream algorithmic processing. The 717 ROIs / 1424 AOIs were probed using GeoMX Whole Transcriptomic Atlas (GeoMX Hu WTA™), containing 18815 probes. The 1424 DCC files generated were subsequently processed using the GeomxTools. Probe and gene QC of the 1424 AOIs resulted in a filtered data object containing 10150 genes across 1151 AOIs (604 GPx and 547 PPx, FIGs. IE and 6B). FIGs. 6C-6D summarize the distribution of ROIs pre-and post- QC, from different regions of the slide (center, edge and isolated foci), as well as AOIs with differing epithelial morphology (Invasive I, and others such as A, D, J, N, and O). Specifically, given their distribution, only ROIs that contain epithelial cells that exhibit atypia (A); invasiveness / transformed (I) and appear normal (adjoining areas of atypia / invasiveness (J)) to delineate the state of the tumor ecosystem and interactions with the surrounding TME were emphasized.

[0137] Most of the ROIs from GPx and PPx samples, post-processing, showed separation between the segments (AOIs), indicative of differences in the underlying biology of theAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0138] cellular landscape captured by segmenting, (FIG. IF). Clustering of the pooled signal from each ROI (average expression across PanCK+ and PanCK- AOIs) captured the inter-patient and intra-tumoral heterogeneity of the TNBC set (FIG. 6E). To better dissect this heterogeneity, the molecular underpinnings and state of the tumor and its TME were characterized as captured by the PanCK- and PanCK+ segments and the interactions that exist across them described in the following sections.

[0139] Example 2 - Deconvoluting the cell-type repertoire of the tumor microenvironment in good and poor prognosis TNBC

[0140] Spatial segmentation of each ROI on PanCK captured the non-epithelial cell milieu in the PanCK- region, including stromal cells and the tumor-infiltrating lymphocytes (TILs), representative of the TME in the TNBC samples. Specifically, the 313 GPx and 272 PPx PanCK- AOIs captured showed varying levels of immune infiltration as detected using the CD3+ (yellow channel) and CD45+ (red channel) via immunofluorescence imaging.

[0141] Intensity-based immune quantification (red channel) showed a significant difference between GPx and PPx(p<0.05, albeit the yellow channel being statistically non-significant) (FIGs. 2A-2B). Recent research has suggested a differential immune cell distribution between tumor edge and center may influence tumor aggressiveness and prognosis in other cancers.

[0142] Prominent transcriptional dysregulation was observed in the TME between AOIs from center to edge in both GPx (albeit not statistically significant, FIG. 7A) and PPx (FIG. 2C). In PPx, however, the TME of AOIs from the center (vs edge) showed a distinct increase (adj p <0.05) in genes associated with metabolic and cellular stress and remodeling including ENO1, CYC1, ATP5F1C, HSPA1B, and AHSA1. Increased expression of genes associated with tumor survival and immune evasion was seen at the edge (vs center) including genes such as NR3C1, BCL2, and AP3BP1. No statistically (fdr<0.05) significant difference was observed between the TME of GPx and PPx, chosen at different regions of the tumor slide- Center, edge and isolated foci (FIG. 7B). A comparison of the genes expressed in the PanCK-segments irrespective of the spatial region between GPx and PPx (M-A plot, FIG. 2C) however showed a higher relative expression in GPx-TME for genes involved in immune response such as B2M, CD74, immunoglobulins and HLA-DRB1; while PPx-TME showed increased expression of genes associated with abnormal cell morphology and EMT including PTBP1, COL1A1, COL1A2, COL6A2, FLNA, MYH9, and VIM.

[0143] To gain quantitative insights into the differential contribution of various cell types to the TME in GPx and PPx TNBC, SpatialDecon vl.l was utilized. As outlined herein, first, aAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0144] custom reference profile matrix of TNBC was generated, for use in SpatialDecon. Utilizing this, the abundances of 7 major cell types (T-cells, B- cells, plasmablasts, myeloid, mesenchymal cells (perivascular-like (PVLs) and fibroblasts), and endothelial cells) were estimated for both GPx and PPx samples (FIG. 2D). Correlation of the deconvoluted T-cell proportions with the proportion of T-cells captured via IF (CD3+ channel intensities in GPx and PPx (average r2~0.7, FIG. 2E) showed good concordance. A visual inspection of the abundances (FIG. 2D) highlighted: 1. the presence of a richer immune repertoire (T, B, and plasmablasts) within GPx, and a richer myeloid cell presence, in AOIs that showed immune infiltration; 2. higher mesenchymal content (fibroblast and PVL) within PPx. Based on the analysis of quantile distributions (IQR and range) of cell type frequencies, it was found that GPx AOIs exhibit a slightly more varied and richer immune presence (myeloid cells, T, B, and plasmablasts), consistent with the visual summary (FIG. 7C). A similar analysis of PPx AOIs likewise showed a stronger and more varied mesenchymal presence (PVL+Fibroblasts) compared to GPx. 31 of the 32 patients profiled here were previously subtyped into three distinct groups (CC1- increased immune infiltration; CC2- immune desert and CC3-increased fatty acid / nuclear receptor signaling) using bulk-expression measurements Interestingly, a higher proportion of patients in the CC2 and CC3 subgroups were identified within the PPx (-66% of the samples). Furthermore, utilizing the spatial deconvolution calculated above, frequencies were averaged across all AOIs for each patient in GPx (FIG.

[0145] 2G) and PPx (FIG. 2F). It was observed that PPx women subtyped into the CC2 subgroup showed a distinct lack of T-cells in the TME, consistent with the definition of CC2 (FIG. 2H). GPx had a higher proportion of samples belonging to the CC1 subtype (-64%), which was previously shown to correlate with better prognosis. Given the differential T-cell distribution between the prognostic groups, the nature of the T-cell enriched AOIs was examined in the following section.

[0146] Example 3 - GPx and PPx TNBC exhibit distinctly different cell states in the TME with significant immune infiltration

[0147] Recent research suggests that only a small portion of the TILs exist in a tumor-reactive state, implying that a majority of TILs are quiescent and not functional. It was asked if there was a difference in the functional state of the TILs and surrounding stroma, within ROIs with significant T-cell infiltration identified by deconvolution, across GPx and PPx, irrespective of the epithelial state captured within the AOI (A, J, or I). To control for the intra-patient heterogeneity, patient samples were chosen that showed greater than a third ofAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0148] their AOIs with a significant proportion of T-cells (>0.75 quantiles) confirmed via IF, resulting in 61 AOIs from 5GPx women (FIG. 7D) and 40 AOIs from 3 PPx women (FIG.

[0149] 7E) and, henceforth referred to as the Tenr subsets. To investigate the spatial organization of the immune cells in these Tenr subsets across GPx and PPx, a spatial entropy analysis was performed on immunofluorescent images of these regions. The spatial partial information (SPI) was calculated, a measure of how much spatial organization can explain the distribution of staining. In the GPx-Tenr subset, CD3+ staining showed significantly higher SPI with PanCK+ staining at distances of ~28 and 70pm (FIG. 3A). The distance here signifies the average pairwise distance from the distribution profile. A higher SPI implies that the distribution of T-cells is more dependent on tumor cell organization. A higher SPI was observed in GPx-Tenr than PPx-Tenr implying increased interaction of T-cells with tumor cells in the GPx-Tenr subset.

[0150] Differential analysis of genes expressed in the PanCK- segments of GPx-Tenr w.r.t PPx-Tenr (at a restrictive threshold of adj p <0.05), showed significant upregulation of genes such as CD37, MAPK3, MARS1 and GID8 in PPx, that have been previously shown to correlate with poor prognosis (FIG.3B). However, these measurements (and results) provide a unique challenge to discretizing the “transcriptomic state” of the various deconvoluted immune cell types, as the measurements are made on tens to hundreds of nuclei and across cell types within the PanCK- segments. To overcome this challenge, a manifold approach was adopted to characterizing the transcriptional states for the deconvoluted cell types in GPx and PPx. First, a recently published machine learning framework called EcoTyper was leveraged and adapted, which defines distinct functional cell states for various cell types in carcinomas, using publicly available bulk expression data. Using EcoTyper, and the gene expression measured within the PanCK- segments, it was identified that a majority of the cell types exist in similar cell states within both GPx and PPx (FIG. 3C). For instance, most of the fibroblasts exist in state S05, while B cells exist in naive (SOI) and activated (S02 and S05) states. Notably, however, there are certain cell states more pronounced within PPx-Tenr (tabulated in FIG. 3D). For instance, fibroblasts in PPx-Tenr showed increased presence of myofibroblast-like (SOI), CAF1 (S03), and pro-migratory states (S08). A higher proportion of AOIs were found to be in the CD8 T-exhausted state (S03) within PPx- Tenr. Additionally, a significant difference was observed in the average expression of cell-state markers for several of these unique states for instance, CD8 T cells in S03 in PPx had increased expression compared to AOIs in S03 within GPx (FIG. 7F). Next, utilizing BulkSignalR 123 and 69 ligand-receptor (L-R) interactions (at corr>0.25, q<0.05) were identified that areAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0151] uniquely exhibited in the TME of the GPx- Tenr and PPx- Tenr, respectively (FIG. 3E). Cell types that were most correlated with each LR-Interactions based on the deconvoluted abundances were also computed. Specifically, several of the L-R interactions uniquely identified within the PPx-Tenr stroma such as COL1 A1-CD44, MMP9-CD44, and SPP1-ITGAV (FIG. 3E) are known to correlate with poor prognosis, a more mesenchymal tumor characterized by aggressive biology of tumors, consistent with EcoTyper results.

[0152] Example 4 - Characterizing the epithelial cell states in good prognosis and poor prognosis TNBC

[0153] The PanCK+ segments captured via GeoMX DSP allowed unprecedented insights into the state of the epithelial cells within GPx and PPx. Broad characterization of the cell states was first performed using single-sample GSEA (ssGSEA) on 291 GPx and 275 PPx epithelial AOIs (PanCK+) (FIGs. 8A-8B). ssGSEA results highlighted a prominent increase in mechanisms associated with proliferation, DNA repair and metabolism within the invasive epithelia, in both GPx and PPx (FIGs. 8A-8B). The AOIs with non-invasive epithelia in both subsets were more prominently associated with signaling cascades involved in EMT including WNT P-catenin, Notch, and KRAS signaling. It was also observed that a majority of the AOIs with tumor cells (or invasive epithelia (I)) clustered distinctly from AOIs with other epithelial types, in both GPx and PPx, irrespective of the region.

[0154] To improve the signal-to-noise ratio and emphasize molecular mechanisms that are significantly differential between GPx and PPx epithelia, non-negative matrix factorization was utilized to reduce the number of features. 1712 consensus features (across 291 GPx PanCK+ AOIs and 1677 consensus features across 275 PPx PanCK+ AOIs) were identified and used for all further downstream analyses of the epithelial state.

[0155] Example 5 - A distinct transcriptional profile underlies invasive tumor epithelia with differential prognosis

[0156] Epithelial tumors are characterized by rapid growth and differentiation and exist in varying states across the tumor. It was hypothesized that the trajectory followed by epithelial cells from a non-invasive state to an invasive state differs in groups with differential prognosis, and subsequently proscribe prognosis.

[0157] As a first step, it was assessed if the state of the tumor cell (invasive epithelia) is informed by the spatial location of the chosen ROIs (center or edge) of the tumor (FIGs. 4A and 9A). A comparison of tumor cells (AOIs) from the center of the tumor to those at theAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0158] edge of the tumor showed no statistically significant differences (at fdr<O.O5) in either GPx or PPx patients (FIG. 4B). These indicated that the transcriptional state of the tumor epithelia (I) is independent of the spatial context. Next, transcriptional programs underlying the invasive (tumor) epithelial state was identified by comparing them with normal adjacent epithelia in both GPx and PPx (FIG. 4C). A significant upregulation of genes (FDR <0.05) was observed in invasive epithelia, irrespective of prognosis, and included several histones such as H3C15, H4C15, Hl-5, H2AC19, H2AZ1, H3C7; genes associated with cytoskeletal stability and cell cycle control including TMSB10, STMN1, CDKN2A were also highly differentially regulated (FIG. 4C). A closer inspection of the upregulated DEGs in the invasive epithelia of PPx and GPx (PPx-I and GPx-I), showed a large degree of overlap (191, FIG. 4D), with the overlapping genes being enriched for proliferation (G2M checkpoint, MYC targets, E2F targets, mTORCl signaling, FIG. 9B). This finding is suggestive of a highly dynamic cell state of invasive epithelia compared to its normal-adjacent counterparts.

[0159] Interestingly, however, DEGs uniquely dysregulated within each prognostic group, compared to their normal adjacent showed activation of distinct programs. Notably, uniquely differentially expressed genes in the GPx invasive epithelia (155 / 346) were identified to belong to a tightly connected network of interacting genes (PPI enrichment adj p <10-16, FIGs. 4E and 9C) associated with immune signaling, antigen processing and presentation (APC), and cellular adhesion including genes such as MX1, IFI6, CTSB / D, HSP90AA1, PSMB1, PSMC4, PSMC2, ANXA2 / 5. Transcription factors (TFs) such as API1, FOS, and STAT3 were identified as TFs (using DecoupleR) driving this response of the upregulated DEGs identified uniquely in GPx-I (FIG. 9D). STAT3 has been previously reported to correlate with favorable prognosis. Furthermore, the 54 immune transcripts identified within this module showed significantly higher expression in GPx-I compared to PPx-I epithelia (FIG. 4F). These results are notable in light of the recent research that has indicated immune mimicry of tumor epithelial cells is widespread in tumors and is portrayed as a feature distinct to oncogenic transformation. Utilizing the recently published optimized immune response signature (oIRS) of tumor cells, a preferential increase of oIRS was further observed in the tumor epithelia of GPx-I compared to PPx-I (FIG. 4G). Several components also involved in APC were uniquely increased in GPx-I. APC has been previously suggested to be an independent prognostic factor in BC. A similar analysis of the uniquely upregulated genes in PPx-I showed evidence for extensive metabolic reprogramming and chromatin remodeling including genes such as MKI67, NDUFAB1, NDUFS6, HSPA9, AKT2, PFKP, PRDG2, MDH1 / 2, ATP5F1C, CTPS1, LDHA (mitochondrial and cellular metabolism, fdr<0.001) andAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0160] AURKB, KIF2C, KIF1C, H2As, TPX2 (chromatin remodeling / organization, fdr<O.OOl) (FIGs. 4H and 9E). Transcription Factor (TF)-target enrichment analysis of the uniquely upregulated DEGs showed increased activity of TFs across a majority of the AOIs, such as HIF1 A, MYC and AR that are known to be crucial determinants of oxidative, glycolytic and lipid metabolism (FIG. 9F).

[0161] Example 6 - Pre-invasive epithelia show evidence of aggressive biology in poor prognosis TNBC

[0162] Based on the earlier observation that the tumor epithelial (invasive) cell state is different between groups that display differential prognosis, it was hypothesized that the aggressive biology associated with poor prognosis TNBC is more apparent in the pre-invasive epithelial state of the PPx samples. To test this hypothesis, the transcriptional landscape of the pre-invasive state (A) was compared to invasive epithelia within both groups (PPx and GPx). It was observed that PPx exhibits a more dominant transcriptional regulation in transitioning from atypia to invasive epithelia (567 DEGs, fdr<0.05, FIG. 5A right), compared to GPx. Consistent with the hypothesis, it was observed that the 134 genes upregulated in PPx-atypia (PPx- A) were largely associated with EMT (loss of baso-apical polarity), regulation of cell migration, cell motility and cell-cell adhesion (FIG. 5B).

[0163] Increased expression of chemokines was observed in the atypia of PPx, including CXCL12, CXCL14 and CXCR2. PPx- A also shows an increased presence of KRT15 and KRT34 transcripts that have been suggested to correlate with an invasive phenotype in certain epithelial tumors with poor prognosis. Evidence in the literature points to complement activation in tumor initiation and the role of tumor cells in modulating complement activation within the tumor microenvironment (TME). It was noted that the expression of several upstream regulators in the classical complement cascade including C1QA / B / C, C1R, CIS, C2, C3, and C4B was significantly reduced in the transition from PPx-A to PPx-I (FIG. 5C).

[0164] In contrast, they were notably higher in PPx-A compared to GPx- A (FIGs. 5D and 10). To better characterize the role of complement in the interactions between tumor cells and the TME in GPx-A and PPx-A, L-R analysis was performed on PanCK+ and PanCK- in ROIs with atypic epithelia using BulkSignalR. 115 L-R interactions unique to GPx-A and 140 interactions unique to PPx-A (FIG. 5E) were identified. It was noted that both groups showed distinct C3-associated interactions (correlated in epithelia)- PPx-A showed C3-CD81 and C3-LRP1 binding. Expression of CD81 (on B-cells and exosomes secreted by CAFs) and LRP1 (expressed on epithelial cells and fibroblasts) have both been previously associatedAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0165] with EMT and poor prognosis. C3-CD81 and C3-LRP1 interactions captured here could indicate anaphylatoxin-mediated priming of precancerous epithelia into tumor cells capable of increased migration. In contrast, GPx-A was enriched for C3-ITGAX, likely indicative of an anti-tumorigenic opsonization event.

[0166] Example 7 - Identification of an epithelial gene signature underlying differential prognosis

[0167] Results presented in the sections above outline a distinct transition along the spectrum of epithelial states (J, A, I) in tumorigenesis. It was further sought to identify a universal epithelial signature that is agnostic to cell state but dictates prognosis. To this extent, NMF was re-run on the entire set of PanCK+ AOIs (566 AOIs, 291 GPx, and 275 PPx) and identified 1722 consensus features. A deep-learning approach was employed to identify an epithelial signature containing 235 and 236 genes that are essential to accurately label GPx and PPx epithelia. Briefly, the CNN developed within this study consisted of two distinct towers, with each tower, incorporating a Convolutional Neural Network (CNN) layer, followed by a LeakReLU activation function and a Max Pool layer with a kernel size of (2, 2) (FIG. 5F). The two pooled outputs were then flattened and concatenated before being fed into fully connected layers, initially with 128 nodes and ultimately narrowing down to 2 nodes, representing the model's output “prognosis”. The model was trained using an Adagrad optimizer with a learning rate set at 0.001 and a Binary Cross-Entropy (BCE) loss function for a duration of 200 epochs. Discrete saliency maps were generated for both good prognosis and poor prognosis to discern the most influential features. Next, genes that had a weightage of greater than 0.5 were extracted, to ensure the importance of features in determining prognosis. The genes were then consolidated into a ranked list using Borda counts. This process was repeated 100 times to ensure robustness and reliability. Results were only used from each newly trained model if its accuracy and precision on the test case were greater than an average of 90%. Finally, these ranked lists were compiled from each iteration into a comprehensive aggregated list, again using the Borda counts to establish a definitive ranking of features over these multiple iterations and represent the features that were most often used by CNN to predict the labels. A rank comparison of the two gene lists highlighted features / genes that were discriminatory (more dissimilarly ranked between the two lists) and were associated largely with a MYC-centric cellular stress response (red) and cell cycle control (purple) (FIG. 11) and included genes such as several histones H2BC11 / 17,Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT

[0168] H2AC13 / 19, MYC, YWHAZ, and CALR that have been previously shown to correlate with survival.

Claims

Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCTWHAT IS CLAIMED IS:

1. A method of treating triple-negative breast cancer in a subject, the method comprising administering a therapeutically effective amount of a composition comprising a complement inhibitor to the subject,wherein the complement inhibitor inhibits a complement cascade component, and wherein the complement cascade component comprises C1QA, Cl QB, C1QC, C1R, CIS, C2, C3, or C4B, thereby treating the triple-negative breast cancer in the subject.

2. The method of claim 1, further comprising identifying the subject as having (i) less than 1% expression of estrogen and progesterone receptors and / or wild-type expression of HER2 / neu gene compared to reference non-cancerous breast tissue, and / or (ii) elevated complement C3 activation in atypical epithelial breast tissue as compared to reference non-cancerous breast tissue.

3. The method of claim 1, wherein the complement inhibitor is a C3 inhibitor.

4. The method of claim 3, wherein the C3 inhibitor comprises:(a) a monoclonal antibody that specifically binds to C3 or a C3 cleavage fragment; (b) a compstatin or compstatin analog peptide that inhibits C3 cleavage;(c) a small molecule inhibitor of C3 or C3 convertase;(d) an RNA-based therapeutic targeting C1QA, C1QB, C1QC, C1R, CIS, C3, or C2; or(e) a C3a receptor (C3aR) inhibitor.

5. The method of claim 3, wherein the C3 inhibitor comprises pegcetacoplan or compstatin.

6. The method of any one of claims 1-5, wherein the subject has been previously administered a neoadjuvant chemotherapy or an adjuvant chemotherapy.

7. The method of any one of claims 1-6, wherein the subject has residual diseases.Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT8. The method of any one of claims 1-7, wherein the administering comprises intravenous administration, subcutaneous administration, intraperitoneal administration, rectal administration, oral administration, or combinations thereof.

9. The method of any one of claims 1-8, further comprising administering an anti-cancer treatment.

10. The method of claim 9, wherein the anti-cancer treatment comprises a therapeutic a-lactalbumin vaccine, a platinum-based chemotherapy, a taxane-based chemotherapy, an immune checkpoint inhibitor, a PARP inhibitor, an antibody-drug conjugate, adoptive cell therapy (ACT), or any combinations thereof.

11. A method of improving chemotherapeutic response in a triple negative breast cancer (TNBC) patient with residual disease, the method comprising:(a) administering a therapeutically effective amount of a composition comprising a complement inhibitor to the patient, wherein the complement inhibitor inhibits a complement cascade component, and wherein the complement cascade component comprises C1QA, C1QB, C1QC, C1R, CIS, C2, C3, or C4B; and(b) administering a therapeutic effective amount of an immunotherapy or chemotherapy, thereby improving chemotherapeutic response in a TNBC patient with residual disease.

12. The method of claim 11, further comprising identifying the subject as having (i) less than 1% expression of estrogen and progesterone receptors and / or wild-type expression of HER2 / neu gene as compared to reference non-cancerous breast tissue, and / or (ii) elevated complement C3 activation in atypical epithelial breast tissue as compared to reference non-cancerous breast tissue.

13. The method of claim 12, wherein the complement inhibitor is a C3 inhibitor.

14. The method of claim 13, wherein the C3 inhibitor comprises:(a) a monoclonal antibody that specifically binds to C3 or a C3 cleavage fragment; (b) a compstatin or compstatin analog peptide that inhibits C3 cleavage;(c) a small molecule inhibitor of C3 or C3 convertase;Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT(d) an RNA-based therapeutic targeting C1QA, C1QB, C1QC, C1R, CIS, C3, or C2; or(e) a C3a receptor (C3aR) inhibitor.

15. The method of claim 14, wherein the C3 inhibitor comprises pegcetacoplan or compstatin.

16. The method of any one of claims 11-15, wherein the subject has been previously administered a neoadjuvant chemotherapy or an adjuvant chemotherapy.

17. The method of any one of claims 11-16, wherein the administering comprises intravenous administration, subcutaneous administration, intraperitoneal administration, rectal administration, oral administration, or combinations thereof.

18. The method of any one of claims 11-17, further comprising administering an anticancer treatment.

19. The method of claim 18, wherein the anti-cancer treatment comprises a therapeutic a-lactalbumin vaccine, a platinum-based chemotherapy, a taxane-based chemotherapy, an immune checkpoint inhibitor, a PARP inhibitor, an antibody-drug conjugate, adoptive cell therapy (ACT), or any combinations thereof.

20. A method of identifying a TNBC patient as a candidate for complement C3 inhibitor therapy, the method comprising:(a) obtaining a biological sample from the patient, wherein the biological sample comprises an atypical pre-invasive epithelial region;(b) detecting C3 protein activation in the atypical pre-invasive epithelial region;(c) comparing the C3 protein activation to a reference value derived from a goodprognosis TNBC patient;(d) identifying the patient as a candidate for complement C3 inhibitor therapy based on the C3 protein activation in the pre-invasive epithelial region being elevated relative to the reference value; and(e) administering a complement C3 inhibitor to the identified TNBC patient.Attorney Docket No: 15670-0458WO1 / SD2025-224-2PCT21. The method of claim 20, wherein the detecting C3 protein activation comprises immunohistochemical staining the biological sample with an anti-C3 or anti-C3b antibody.

22. The method of claim 21, wherein the detecting step (b) further comprises quantifying a C3-positive cell fraction by digital pathology image analysis.

23. The method of any one of claims 20-22, wherein the complement C3 inhibitor comprises:(a) a monoclonal antibody that specifically binds to C3 or a C3 cleavage fragment; (b) a compstatin or compstatin analog peptide that inhibits C3 cleavage;(c) a small molecule inhibitor of C3 or C3 convertase;(d) an RNA-based therapeutic targeting C1QA, C1QB, C1QC, C1R, CIS, C3, or C2; or(e) a C3a receptor (C3aR) inhibitor.

24. A method comprising:(a) providing a sample of atypical breast tissue from a subject having triple-negative breast cancer; and(b) assaying the sample to determine a level of C3 protein activation in the sample.

25. The method of claim 24, wherein the determining of the level of C3 protein activation comprises immunohistochemical staining the biological sample with an anti-C3 or anti-C3b antibody.

26. A method comprising:(a) providing a breast tissue sample from a subject, the sample containing atypical epithelia;(b) assaying the breast tissue sample using spatial transcriptomic profiling to detect a presence of anaphylatoxin-mediated priming in the atypical epithelia by identifying at least one unique ligand-receptor interaction selected from the group consisting of C3-CD81 and C3-LRP1;(c) determining that the identified ligand-receptor interaction is localized within an area of illumination (AOI) derived from a specific spatial origin within the breast tissueAttorney Docket No: 15670-0458WO1 / SD2025-224-2PCTsample, wherein the spatial origin is selected from the group consisting of a tumor center region and a tumor edge region; and(d) administering a C3-directed therapeutic agent to the subject identified as having the C3-CD81 or C3-LRP1 interaction at the tumor center or tumor edge to treat high-risk pre-cancerous breast lesions.

27. The method of claim 26, wherein the spatial transcriptomic profiling of the atypical epithelia comprises:(a) hybridizing the breast tissue sample with a plurality of profiling probes, each profiling probe comprising an indexing oligonucleotide attached to a target-binding moiety via a UV- photocleavable linker;(b) identifying at least one region of interest (ROI) within the hybridized breast tissue sample using immunofluorescent markers;(c) segmenting the ROI into an AOI corresponding to the atypical epithelia using a digital micromirror device to direct UV illumination with ~1 pm precision;(d) releasing the indexing oligonucleotides from the atypical epithelia segment via the directed UV illumination; and(e) sequencing the released indexing oligonucleotides to identify the presence of the C3-CD81 or C3-LRPl interaction.