High-throughput single-cell RNA sequencing (scRNA-Seq) has become a routine platform for the dissection of solid tumours into their cellular components. The development of methods to incorporate detection of cellular protein epitopes via barcoded antibodies has enabled advances in immunophenotyping but its application to investigating tissue immunology is limited. Here we present an optimised experimental and analytical framework for the phenotyping of normal and malignant human tissues by integrated scRNA-Seq and barcoded antibody analysis of single cell suspensions and tissue sections. We show that the integration of RNA and protein resolves immune subsets at higher resolution than either modality alone. By estimating the contribution of individual antibodies to cell clustering, and their optimal concentration, we guide the development of antibody panels for the analysis of solid tissues. Additionally, we describe Spatial Indexing of Transcriptomes and Epitopes (SITE-Seq), a method for spatially-resolved joint transcriptome and epitope analysis of tissues. We then applied a joint single cell RNA and epitope analysis to a cohort of 6 breast cancer samples to reveal the improved resolution of the tumour microenvironment and novel markers for resting and activated tissue infiltrating lymphocytes (TILs). We reveal how activated CD4+ T follicular helper cells in breast cancers acquire two distinct states, one associated with markers of tissue residency and exhaustion, and a previously undescribed phenotype marked by the expression of IGFL2 and NMB, found to be associated with poor prognosis and response to checkpoint immunotherapy. This work highlights the importance of multi-omic methods for the phenotyping of cell states and the emergence of novel cellular states in the tumour microenvironment.