Mammalian sperm selection in vivo and in vitro: an evolutionary perspective on advanced ICSI techniques.

Fertilisation in mammals is preceded by a multilayered selection process shaped by more than 600 million years of evolution. The female reproductive tract acts as a sequential filter-through vaginal pH, cervical mucus, the uterotubal junction, the isthmic sperm reservoir, the cumulus matrix and the zona pellucida-such that only a few hundred capacitated, structurally and molecularly competent spermatozoa reach the ampulla. Intracytoplasmic sperm injection (ICSI), introduced in 1992, bypasses these filters. Advanced techniques-including IMSI, physiological ICSI (PICSI), magnetic-activated cell sorting (MACS), microfluidic sorting and deep-learning image classification-aim to restore part of this discriminatory capacity in the laboratory. Systematic reviews and current professional guidance indicate that none has produced a robust improvement in cumulative live birth in unselected assisted-reproduction populations, although ESHRE considers microfluidic selection an option. This review maps each technique to the in vivo filter it purports to mimic, evaluates the clinical evidence, considers long-term follow-up of ICSI-conceived offspring and imprinting-disorder surveillance, and asks whether targeted, mechanism-based selection in defined subgroups could succeed where unselected use has not. Evidence from non-human mammals-colloid and microfluidic selection in horses and cattle, sex-sorted semen, species differences in ICSI efficiency and mouse ICSI with DNA-damaged spermatozoa-is used to test the framework where human data are observational only. The evolutionary perspective is used as a design framework: for each technique, it identifies the mechanism to be recapitulated and the maternal, pregnancy and offspring outcomes by which benefit and safety should be judged.