Thank you both. The code is below, simplified a bit for simplicity. ATpoint, your suggestion re: the top 20% of genes by baseMean makes me wonder if the shift in LFC=0 is an artifact of the experiment, which should include an enrichment for certain RNA types in the treated group (grey points in the volcano plot). If these are robustly increased in most cells whereas most other genes in the single-cell data are non-uniformly expressed, perhaps they are driving the normalization?

# Standard Seurat workflow
merged_obj <- merge(x = obj_1, y = c(obj_2, obj_3, obj_4, obj_5, obj_6, obj_7, obj_8, obj_9, obj_10), project = "projectID")
merged_obj$percent_mt <- PercentageFeatureSet(merged_obj, pattern="^mt-")
filtered_obj <- subset(merged_ob, subset = nFeature_RNA >= 200 & nFeature_RNA <= 2500 & percent_mt <= 5)
filtered_obj <- NormalizeData(filtered_obj)
filtered_obj <- FindVariableFeatures(filtered_obj)
filtered_obj <- ScaleData(filtered_obj)
filtered_obj <- RunPCA(filtered_obj)
int_obj <- IntegrateLayers(object = filtered_obj, method = CCAIntegration, orig.reduction = "pca", new.reduction = "integrated.cca", verbose = FALSE)
int_obj[["RNA"]] <- JoinLayers(int_obj[["RNA"]])
int_obj <- FindNeighbors(int_obj, reduction = "integrated.cca", dims = 1:30)
int_obj <- FindClusters(int_obj, resolution = 0.8, cluster.name = "integrated_clusters")
int_obj <- RunUMAP(int_obj, dims = 1:30, reduction = "integrated.cca")

# Annotated cells manually with RenameIdents()
ann_obj$cell_type <- Idents(ann_obj)
ann_obj$cell_type[Cells(ann_obj)] <- paste(Idents(ann_obj))

# Pseudobulk DEG analysis
pseudo_obj <- AggregateExpression(ann_obj, assays = "RNA", return.seurat = TRUE, group.by = c("celltype", "group", "sample"))
pseudo_obj$pseudo_celltype.group <- paste(pseudo_obj$cell_type, pseudo_obj$group, sep = "_")
Idents(pseudo_obj) <- "pseudo_celltype.group"
pseudobulk_DEG <- FindMarkers(object = pseudo_obj, ident.1 = "celltype_group2", ident.2 = "celltype_group1", test.use = "DESeq2")