*3.4. GO Term Analysis of DEGs*

To clarify the functions of the DEGs (|fold change| > 1.5, FDR < 0.05), GO function enrichment analysis was performed in Metascape (File S2 (Supplementary Materials)) [18]. The top 20 GO terms of four groups (HF: HL60-FeNPs, HP: HL60-PBNPs, KF: KG1a-FeNPs, and KP: KG1a-PBNPs) are displayed in Figures 4 and 5, respectively. Because PBNPs can effectively scavenge ROS via multienzyme-like activity including peroxidase (POD), catalase (CAT), and superoxide dismutase (SOD) activity while FeNPs produced hydroxyl radicals (·OH) through the Fenton reaction and peroxidized lipids [15,37], the difference of GO terms between these two kinds of iron nanoparticles in one cell line (HL60 or KG1a) was firstly characterized. The most significant GO term in HF, HP, KF, and KP was the regulation of the lipid metabolic process, myeloid leukocyte activation, the HFE–transferrin receptor complex, and negative regulation of megakaryocyte, respectively. Interestingly, 25% of GO terms of HF was mainly enriched in "lipids", including regulation of lipid metabolic process, intracellular lipid transport, lipid biosynthetic process, cytoplasmic vesicle membrane, long-chain fatty acid metabolic process, and lipase activity (Figure 4A,B). In comparison, only 10% of GO terms of HP were related to "lipids", including regulation of lipid metabolic process and plasma membrane repair (Figure 4C,D). The results showed that the capability of FeNPs to produce ROS made it easier for FeNPs to regulate lipid metabolism than PBNPs. To further show the roles of genes in lipid metabolism, the DEGs in these GO terms are listed in Table 1. There were 28 (20 up- and 8 downregulated) genes in HF and 13 (9 up- and 4 downregulated) genes in HP involved in lipids regulation, of which 6 shared genes (ABCA1, FPR2, KIT, FADS1, ME1, and AHNAK2) were found. Particularly, the expression of ABCA1 was most significantly upregulated in lipid metabolism of HF and HP. ABCA1 was an important membrane-associated protein and actively participated to phosphatidylcholine, phosphatidylserine, and sphingomyelin transfer [38]. However, ABCA1 and 5 other shared genes were not found in the lipid-associated GO terms of KG1a. Besides these shared genes, ACACA was the unique gene that frequently appeared in lipid metabolism-related GO terms of HF. ACACA, acetyl-CoA carboxylase, was the first and rate-limiting step of de novo fatty acid biosynthesis [39]. This gene was also upregulated in HF and HP. This kind of significant effect of FeNPs on lipid metabolism was also found in the HepG2 cell treated by Fe3O4 nanoparticles in a previous study [36]. This previous study and this study all revealed that the treatment of Fe3O4 nanoparticles induced lipid accumulation in cells. In consistence with the previous study [36], this study also found that several lipid synthesis-related genes were upregulated in HF, including FDFT1, ACAT2, and HMGCS1. In contrast, the most significant biological process was myeloid leukocyte activation and cation homeostasis in the PBNP-treated HL60 cells (Figure 4C,D).

**Figure 4.** Top 20 Gene Ontology (GO) terms of the HL60 cells treated with two kinds of iron nanoparticles: (**A**,**B**) heatmap (**A**) and enrichment network (**B**) colored by the same cluster of GO terms in the FeNP-treated HL60 cells, and (**C**,**D**) heatmap (**C**) and enrichment network (**D**) colored by the same cluster of GO terms in the PBNP-treated HL60 cells. The colored labels followed the order of top 20 GO terms. (**E**) The Venn analysis of top 20 GO terms in the HL60 cells treated with PBNPs and FeNPs. The detailed information of all GO terms is shown in File S2 (Supplementary Materials).

**Figure 5.** Top 20 GO terms of the KG1a cells treated with two kinds of iron nanoparticles: (**A**,**B**) heatmap (**A**) and enrichment network (**B**) colored by the same cluster of GO terms in the FeNP-treated KG1a cells, and (**C**,**D**) heatmap (**C**) and enrichment network (**D**) colored by the same cluster of GO terms in the PBNP-treated KG1a cells. The colored labels followed the order of top 20 GO terms. (**E**) Venn analysis of top 20 GO terms in the KG1a cells treated with PBNPs and FeNPs. The detailed information of all GO terms is shown in File S2 (Supplementary Materials).

For KG1a, there was few GO terms related to lipid metabolism. Only three genes (ITPK1, PLCG2, and POU1F1) were involved in inositol triphosphate metabolic process in KF, and four genes (SPTA1, TGFB2, CXCR4, and EHD2) mediated plasma membrane organization in KP. In contrast, GO terms that responded to metal ions were highly enriched in KG1a (Figure 5A–D). As listed in Table 2, 4 GO terms (containing 13 genes) were found in KF and just one term (containing 4 genes) was found in HF. It implied that KG1a was more sensitive to metal ions than HL60 under the FeNP treatment. For example, TFRC and TFR2 genes were over-suppressed in KF, which belong to the transferrin receptor-like family and are necessary for cellular iron uptake [40]. Another receptor, BMPR1B, overexpressed in KF, can specifically bind the bone morphogenetic protein (BMP) to regulate a wide range of biological processes including iron homeostasis, fat and bone development, and ovulation [41].


**Table 1.** GO terms of lipid metabolisms in KG1a and HL60 exposed to FeNPs or PBNPs (Top 20, *p* < 0.01).

Note: HF: HL60-FeNPs, HP: HL60-PBNPs, KF: KG1a-FeNPs, KP: KG1a-PBNPs.



Note: HF: HL60-FeNPs, HP: HL60-PBNPs, KF: KG1a-FeNPs, KP: KG1a-PBNPs.

It was found that the antioxidation-related genes were highly regulated in two cells treated by two kinds of iron nanoparticles. NQO1 and GCLM were commonly upregulated in two cells treated by two kinds of nanoparticles (Figure 2D). Two genes (NQO1 and HMOX1) that was closely associated with antioxidant metabolism were in upregulated by two kinds of nanoparticles in KG1a (File S1 (Supplementary Materials)). NQO1 and HMOX1 are representative NRF2-regulated antioxidation genes [26]. NQO1 reduces quinone to hydroquinone, and HMOX1 catalyzes the degradation of heme to biliverdin, CO, and Fe+. If either one of the two genes was knocked out, it would enhance erastinand sorafenib-induced ferroptosis in hepatocellular carcinoma cells [7]. The two genes may help KG1a cells to scavenge ROS and thus resist iron nanoparticle-induced ferroptosis. Another antioxidation gene, GPX3, was only upregulated in HF. Together with the upregulated GCLM and GCLC in HF, these three typical NRF2-regulated antioxidation genes [26] revealed high oxidative stress in HF, which agrees with the decrease of cell viability of only HF in all treatments (Figure 1C). In addition, iron nanoparticles-induced oxidative stress was also demonstrated by the wide regulation of the expression of as many as 15 (HF), 42 (KF), 50 (HP), and 70 (KP) genes coding oxidase, reductase, peroxidase, dehydrogenase, epoxidase, and oxidoreductase (File S4 (Supplementary Materials)). It was interesting that a recently identified anti-ferroptosis gene, AIFM2 (renamed as ferroptosis suppressor protein 1, FSP1, a CoQ oxidoreductase) [42,43], was also upregulated in KF and HP. Many proteins sequester transition metals or transport them and thus indirectly act as antioxidants by suppressing formation of HO· from H2O2 by Fenton chemistry [44]. These proteins include ferritin (comprising light FTL1 and heavy FTH1 subunits), ferroportin (FPN1/SLC40A1), metallothionein, and ceruloplasmin. FTL was commonly upregulated by two kinds of nanoparticles in two cells (Figure 2D). FTH1 was upregulated in KF, HP, and KP (File S1 (Supplementary Materials)). Another an important antioxidation gene, SLC7A11, was commonly significantly upregulated in two cells treated by two kinds of nanoparticles (Figure 2D). SLC7A11 imported cysteine into cells for glutathione synthesis and thus plays key role in antioxidation of cells. Besides SLC7A11, as many as 13, 40, 48, and 56 soluble carrier (SLC) family genes were differentially regulated in HF, HP, KF, and KP, respectively (File S1 (Supplementary Materials)). These SLC family genes encoding passive transporters, ion coupled transporters, and exchangers were crucial for intracellular ion homeostasis [45,46]. It was worthy to note that metallothioneins (MTs) were generally downregulated in HL60 and KG1a exposed to PBNPs, including MT1E, MT1F, MT1G, MT1X, and MT2A in HL60, and MT1X and MT2A in KG1a. With a high content of cysteine residues, metallothioneins bound various heavy metals for detoxification in cells and acted as antioxidants to protect against hydroxyl free radicals [47]. The inhibited metallothioneins further supported the low metal toxicity and good antioxidant capacity of PBNPs in HL60 and KG1a. When exposed to PBNPs, this iron nanoparticle can effectively scavenge ROS via multienzyme-like activity to reduce the burden of cellular antioxidants.
