Lipopolysaccharides (LPS) are highly toxic compounds, even at a trace amount. When recombinant proteins are produced in E. coli, it is inevitable that LPS contaminates. However, LPS removal is still technically challenging and costly due to the high degree of solubility in a wide range of solvents. In this study, we explored the possibility of using the N-terminal region containing cysteine-rich, EGF-like, and sushi1–3 domains (CES3) of Factor C from the horseshoe crab Carcinoscorpius rotundicauda to develop a platform to remove LPS from recombinant proteins. We expressed CES3 as part of a recombinant protein, BiP:NT:CBM3:SUMO:CES3:His:HDEL, in Nicotiana benthamiana and found that purified or microcrystalline cellulose (MCC) bead-immobilised CES3 showed strong binding to LPS-containing E. coli. To produce CES3:CBM3 in an LPS-free environment, we generated Arabidopsis transgenic plants harbouring a recombinant gene, BiP:NT:SUMO:CES3:CBM3:HDEL, and found that transgenic plants mainly produce CES3:CBM3:His:HDEL, a truncated version of BiP:NT:SUMO:CES3:CBM3:HDEL via endogenous protease-mediated proteolytic processing in vivo. CES3:CBM3:HDEL purified from Arabidopsis plant extracts and immobilised onto MCC beads removed LPS contamination from protein samples. We propose that the CES3:CBM3 fusion protein produced in plants and immobilised on MCC beads can be a robust and easy platform for LPS removal from recombinant proteins.