Acute exposure to organic and inorganic sources of copper: Differential response in intestinal cell lines

Abstract Scope Copper supplementation in nutrition has evolved from using inorganic mineral salts to organically chelated minerals but with limited knowledge of the impact at the cellular level. Methods Here, the impact of inorganic and organic nutrient forms (glycinate, organic acid, and proteinate) of copper on the cellular level is investigated on intestinal cell lines, HT29 and Caco‐2, after a 2‐hr acute exposure to copper compounds and following a 10‐hr recovery. Results Following the 10‐hr recovery, increases were observed in proteins involved in metal binding (metallothioneins) and antioxidant response (sulfiredoxin 1 and heme oxygenase 1), and global proteomic analysis suggested recruitment of the unfolded protein response and proteosomal overloading. Copper organic acid chelate, the only treatment to show striking and sustained reactive oxygen species generation, had the greatest impact on ubiquitinated proteins, reduced autophagy, and increased aggresome formation, reducing growth in both cell lines. The least effect was noted in copper proteinate with negligible impact on aggresome formation or extended growth for either cell line. Conclusion The type and source of copper can impact significantly at the cellular level.

popular, increasing iron and energy levels in humans, and for growth promotion and reducing infection in farm animals.
At the cellular level, the electron transfer potential of copper is important in redox reactions (Nevitt, Ohrvik, & Thiele, 2012).
Copper cellular toxicity is mostly related to the capacity to generate reactive oxygen species (ROS) resulting in oxidative damage to proteins, lipids, and DNA (Gaetke, Chow-Johnson, & Chow, 2014). For this reason, cellular copper levels are strictly maintained (Nevitt et al., 2012). ROS in cells, normally a result of electron leakage from mitochondria, are involved in many cellular signaling pathways, but increased levels result in oxidative stress and inflammatory responses (Shao et al., 2018;Uriu-Adams & Keen, 2005) with significant cross talk in transcription factors controlling these responses, namely nrf2 and NF-κB (Wardyn, Ponsford, & Sanderson, 2015).
Over the last two decades, organic mineral chelates have been increasingly utilized as a trace mineral supply to improve poultry and livestock productivity (Mondal, Paul, Bairagi, Pakhira, & Biswas, 2008;Nollet, Van der Klis, Lensing, & Spring, 2007). Replacing inorganic trace mineral sources with lower levels of more bioavailable organic trace mineral forms not only minimizes the amount excreted but also optimizes trace mineral-associated performance traits (Abdallah, El-Husseiny, & Abdel-Latif, 2009). Typically, organic mineral chelates contain trace metals complexed to either organic acid, amino acid, or peptide ligands, of which several different types of mineral complexes are commercially available. Despite extensive use in agriculture and use as a supplement for humans, a basic understanding of the differences at the cellular level resultant from their use is very limited.
In this study, we analyze the effects of four different sources of copper on proliferating Caco-2 and HT29 cells to evaluate their effect on cell growth, ROS production, cellular uptake, and effects on protein expression. Three organically complexed sources of copper (copper glycinate-Cu Gly, copper organic acid chelate-Cu OAC, and copper proteinate-Cu Pro) were compared to copper sulfate (CuSO 4 ).

| Chemicals
All chemicals (unless otherwise stated) were obtained from Sigma (Poole, UK). FBS was obtained from Invitrogen. All copper compounds were commercially sourced from independent suppliers.

| Cell lines
The human colon carcinoma cell line Caco-2 (cat. HTB37) was obtained from the American Type Culture Collection and HT29 (cat.

91072201) was obtained from Public Health England Culture
Collection. HT29 and Caco-2 were maintained in MEM supplemented with 1% l-glutamine and 5% or 10% FBS, respectively (growth medium), under normal conditions (37°C, 5% CO 2 ). Both cell lines were mycoplasma negative. These two cell lines were chosen as Caco-2 is a colon-derived cell line that, when differentiates, resembles intestinal enterocytes. HT29, also a colon-derived cell line, resembles goblet cells.

| Treatments
To investigate the effect of copper sources, a slightly toxic concentration of IC20 (to reduce viable cell counts by 20%) over a 2-hr exposure (to reflect the transit in the small intestine, Worsøe et al., 2011) and 10-hr recovery was chosen. An average IC20 value of 0.4 and 0.5 mM copper source, respectively, for Caco-2 and HT29 was measured by viable cell counts with trypan blue following the 10-hr recovery. For all treatments, cells were exposed to freshly prepared 0.4 and 0.5 mM copper solutions in 1% FCS, respectively, for Caco-2 and HT29 for 2 hr and either assayed directly or allowed to recover in growth medium without added copper for specified periods before analysis.
Reactive oxygen species production was monitored with DCFDA (2′,7′-dichlorodihydrofluorescein diacetate, Sigma cat 28850). After overnight attachment, cells loaded with DCFDA were exposed to copper solutions for 2 hr. Following washing, fluorescence was measured on a BioTek plate reader (Synergy HT) with excitation at 485 nm and emission at 528 nm. For the 10-hr recovery, the cells were allowed to recover for just under 8 hr, loaded with DCFDA, and incubated until the 10-hr recovery point, before measuring fluorescence. Cells were trypsinized and viable cell counts made to obtain ROS generation per cell.
For proteomic analysis and validation by Western blotting, exponentially growing cells were exposed as above to copper compounds for 2 hr and then allowed to recover for 10 hr in growth medium.
For recovery assays, cells were set up overnight and treated for 2 hr as above, washed twice in fresh medium (without added copper), and allowed to recover for 24, 48, and 72 hr in growth medium, at which point cells were trypsinized and counted.

| Measurement of copper uptake by ICP-MS
Cells were set up as in treatments with two flasks per condition (one for ICP and the other for cell counts) for 2 hr and washed 2 × in ice-cold PBS. The contents of one flask was lysed in 0.4 ml 2D lysis buffer, the copper content was determined by LC-ICP-MS (liquid chromatography-inductively coupled plasma mass spectrometry), and the other flask was trypsinized for cell counts. Samples for LC-ICP-MS were digested in HNO 3 and analyzed using Agilent Technologies 7700 Series ICP-MS system. Helium gas flow rate used was 4.5 L/min, while the carrier gas was set to 0.81 L/min. Quantitative analysis results were obtained from the MassHunter Workstation Software version A.01.02.

| Proteomic analysis
Label-free LC-MS/MS was carried out on an Ultimate 3000 na-noLC system coupled to an LTQ Orbitrap XL mass spectrometer as previously described (Meleady et al., 2012). Peptide identification through Proteome Discoverer 2.1 using the search algorithms MASCOT (version 2.3) and SEQUEST HT was used to search against the UniProtKB-SwissProt fasta database with 20 274 proteins (taxonomy: Homo sapien) using the following parameters: (a) MS/MS mass tolerance set at 0.5 Da, (b) peptide mass tolerance set to 20 ppm, (c) carbamidomethylation set as a fixed modification, (d) up to two missed cleavages were allowed, and (e) methionine oxidation set as a variable modification. For further analysis, only peptides with ion scores of 40 and above or XCorr of >1.9 for +1, >2.2 for +2, and >3.75 for +3 charged molecules were chosen (Linge et al., 2014). The following criteria were applied to assign proteins as confidently identified: (a) an ANOVA score between the experimental group of ≤0.05 and (b) a fold change ≥1.5. Venn diagrams were obtained using the Venny software (Oliveros, 2007(Oliveros, -2015.

| Western blotting
Cell lysates were separated on a 12% or 4%-12% SDS gels and transferred to PVDF, blocked for 2 hr, and then exposed to primary antibodies (Supporting Information Table S1) overnight at 4°C.
Secondary antibodies conjugated to horseradish peroxidase were detected by enhanced chemiluminescence (Luminol, Santa Cruz, CA, USA). Densitometry was carried out using TotalLab software, TL100.

| Autophagy detection assay
The Cyto-ID Autophagy detection kit (Enzo ENZ-51031) was used to evaluate autophagy following manufacturer's instructions. Caco-2 cells and HT29 cells were exposed to copper solutions for 2 hr followed by a 10-hr recovery in growth medium. Exposure to 30 μM chloroquine in 1% FCS was used as a positive control. The CYTO-ID detection reagent (detects autolysosomes and early autophagic vacuoles) was measured at 480 nm excitation and 530 nm emission, and was normalized to the nuclear Hoechst 33342 (340 nm excitation and 480 nm emission) to give a mean fluorescent intensity per cell that was then compared to control cells not exposed to copper compounds.

| Aggresome detection
The ProteoStat Aggresome detection kit (Cat ENZ-51035) was used to detect denatured protein cargo found within aggresomes following manufacturer's instructions. Caco-2 and HT29 were set up on glass slides and were exposed to copper solutions for 2 hr followed by a 10-hr recovery and a 24-hr recovery. MG132 was used as a positive control incubated for 10 hr or 24 hr. Hoechst 33342 and Aggresome detection reagent were imaged with DAPI and Texas Red filter sets at 20 × magnification and images superimposed using MetaMorph (version 7.7.70).

| Aggregate detection
Protein aggregates were determined by fractioning cell lysates into soluble (Triton X-100) and insoluble fractions that contain aggregates (SDS) as previously described (Rapino, Jung, & Fulda, 2014). Cell lysates taken 10 and 24 hr postrecovery for Caco-2 and HT29 cells, respectively, were lysed in Triton X-100 (2% in PBS with protease and phosphatase inhibitors) and incubated on ice for 30 min, before centrifuging at 18,800 g for 30 min at 4°C. The soluble fraction (supernatant) and the insoluble fraction (resuspended in 1% SDS) were analyzed for ubiquitin-positive proteins by Western blotting.

| Statistics
Statistics were carried out using the Student's t test assuming a twotailed distribution and two samples of unequal variance. Values of p < 0.05 were considered as statistically significant.
F I G U R E 1 ROS production in HT29 (a) and Caco-2 (b) after 2-hr exposure and 10-hr recovery. Results are the average of at least three separate repeats. * denotes statistical significance of p < 0.05 compared to the control for either 2-hr exposure or 10-hr recovery. ROS, reactive oxygen species

| ROS production
Reactive oxygen species generation in Caco-2 and HT29 cells was measured after a 2-hr exposure and also after a 10-hr recovery phase. The most striking result was with Cu OAC whereby ROS increased almost twofold in HT29 and Caco-2 cells (p = 0.036 and p = 0.001, respectively) compared to control after 2-hr exposure ( Figure 1) and was maintained following the 10-hr recovery phase with a 2.4-fold increase noted in HT29 (p = 0.001) and 2.1-fold increase in Caco-2 (p = 0.003) cells noted relative to the control. CuSO 4 and Cu Pro caused transient ROS generation in Caco-2, being reduced and not significant following the 10-hr recovery. The ligands used in the manufacture of the organic copper complexes did not affect ROS production (data not shown).

| Copper uptake on 2-hr exposure
As the bioavailability of the organic and inorganic copper sources may result in different uptake rates, the intracellular copper levels were measured by LC-ICP-MS. Both cell lines took up copper from all compounds ( Figure 2) with Caco-2 displaying a much higher capacity for uptake (2.8-fold higher than HT29).
For Caco-2, uptake levels were quite similar for all copper compounds tested, perhaps indicating saturation; the lowest was CuSO 4 with a 22.7-fold increase (p = 0.004), and the highest was with the Cu Pro with a 28.7-fold increase (p = 0.02). For HT29, Cu Gly resulted in the smallest increase in copper uptake (10fold relative to control) while Cu OAC had the highest at 26-fold, p = 0.036.
F I G U R E 2 Copper uptake after 2-hr exposure to copper compounds as measured by LC-ICP-MS. Results are the average of three separate repeats. * and **denotes significant differences for HT29 and Caco-2 cells, respectively, exposed to coppers relative to control (p < 0.05)

| Transporter expression on 2-hr exposure to copper compounds
As the organic copper sources have been reported to be more efficiently absorbed than inorganic copper in vivo (European Medicines Agency, 2007), the impact on transporters related to metal or peptide uptake (copper transporter 1-CTR1, zinc efflux transporter 1-ZnT1, di-metal transporter 1-DMT1, and peptide transporter 1-PepT1) was investigated. In Caco-2 cells, PepT1, ZnT1, and CTR1 were increased particularly for CuSO 4 and Cu Pro, and no change in DMT1 ( Figure 3

| Proteomic analysis of copper-treated Caco-2 and HT29 cells after a 10-hr recovery
Both cell lines were subjected to proteomic analysis following a 10-hr recovery phase after the 2-hr copper treatment (Tables 1-4). Comparing the organic copper sources to the inorganic form in HT29, six proteins were shared between the organic compounds that were significantly altered compared to the control, and were not altered significantly in CuSO 4 -treated cells (Figure 4a and Table 1). These included metal binding proteins (MT1E and MT1G), redox-related proteins (aldo-keto reductase family 1 member B15 (AKR1B15) and SRXN1), and autophagy-related proteins (sequestosome-1 (SQSTM1) and phosphatidylinositol-binding clathrin assembly protein (PICALM)).
For Caco-2, comparison of the organic to the inorganic copper sources indicated that only one protein was differentially expressed TA B L E 2 Decreased protein expression in HT29 cells treated with organic or inorganic copper compared to control cells (2-hr exposure, 10-hr recovery) in the organic copper compounds, heat-shock 70 kDa protein 6 (HSPA6) with highest expression in Cu OAC at 6.2-fold compared to the control (Figure 4b).
Biological processes most affected by treatment included protein translation, ER function, and the unfolded protein response ( Figure 5, Tables 1-3). In HT29, organic copper sources showed greater changes in proteins involved in antioxidant response, transcription, autophagy, and metal binding observed than inorganic copper, which only caused a marginal increase in proteins involved in cell cycle. In Caco-2, inorganic copper supplementation produced TA B L E 3 Increased protein expression in Caco-2 cells treated with organic or inorganic copper compared to control cells (2-hr exposure and 10-hr recovery) with ANOVA value of p < 0.05 less changes in cytoskeletal, ER/UPR/proteosomal, and antioxidant responses than the organic copper compounds.

| Validation of proteomic targets
Targets from proteomic analysis were assessed by Western

| Impact of copper exposure on protein misfolding
Proteomic analysis pointed to changes in proteosomal proteins and increased expression of protein chaperones involved in handling misfolded proteins. The presence of high molecular weight ubiquitinated proteins was indicative of overloaded proteosomal degradation ( Figure 8) with Cu OAC causing the greatest effect.
While modest in HT29 cells (1.8-fold), the effect was more pronounced in Caco-2 with a threefold increase compared to the control (p < 0.05).

| Autophagy
Increased SQSTM1 an autophagy adaptor protein, seen in both cell lines, may be due to oxidative stress or stalling in autophagic flux.
Microtubule-associated protein 1 light chain 3 (LC3-I) protein, whose increased expression or conversion to LC3-II are important markers of autophagy, was increased by 2.3-fold compared to control in HT29 without conversion to LC3-II ( Figure 9a). For Caco-2, expression of LC3-I and LC3-II was increased following exposure to Cu OAC by 7.7-fold and 23-fold, respectively ( Figure 9b). These may indicate increased autophagy or stalling of autophagy at the 10-hr recovery point; however, a fluorescent autophagy assay showed no increase in fluorescence in HT29 or Caco-2 cells treated with Cu OAC (Figure 9c) compared to control. In fact, the only significant result was a reduction in autophagy in Caco-2 on treatment with Cu OAC (47% reduction, p = 0.018) when compared to the control cells.

| Aggresome formation
Aggresomes are inclusion bodies of misfolded proteins located in a perinuclear space that serve to protect cells from the damaging effects of aggregates, and LC3 and SQSTM1 expression have been linked with aggresome formation. Using a fluorescent detection kit, aggresomes were detected in Cu OAC-treated HT29 and Caco-2 cells (Figure 10a and b). Cu Pro-treated cells did not display signs of aggresome formation in either cell line, while Cu Gly and CuSO 4 caused some sporadic aggresome formation in HT29 but not in Caco-2. Isolation of aggresomes through sequential solubilization in Triton X-100 and then in SDS demonstrated the presence of ubiquitinated protein aggregates (typical of aggresome formation) following Cu OAC addition and to a lesser extent in Cu Gly-and CuSO 4 -supplemented cells (Figure 10c and d). Cu Pro addition resulted in the lowest level of ubiquitinated proteins in the insoluble fraction of the four copper sources tested.

| Impact of copper treatment on cellular survival
With the presence of many ubiquitinated high molecular weight proteins, the impact of the 2-hr copper treatment on cells beyond the 10-hr recovery was assessed ( Figure 11). In HT29, neither CuSO 4 nor Cu Pro had a lasting effect on growth over the 72 hrs monitored.
F I G U R E 6 Validation of HT29 proteomic target proteins HSPA6, SQSTM1, SRXN1, and MT1. Results are the average of at least three separate repeats. Representative blots are shown. Asterisks indicate p < 0.05 when compared to the control Cu Gly caused a 33% reduction in growth (p = 0.04) over the initial 24 hrs but did not affect growth thereafter. Cu OAC significantly reduced growth over the three time points monitored, effecting a 33% reduction even after 72-hr recovery (p = 0.01). In Caco-2, the impact on cell growth was more severe. Both Cu Gly and Cu OAC showed similar reductions in growth to CuSO 4 that were compounded with time reaching 46%, 31%, and 32% of the control, respectively. Cu Pro was the only copper treatment that did not negatively impact on the growth of the cells.

| D ISCUSS I ON
Copper is an essential micronutrient, being involved in many biological processes from photosynthesis to aerobic respiration (Nevitt et al., 2012). In humans, copper is the most tightly controlled micronutrient (Gaetke et al., 2014) (Solomon, 2009). As copper sediments in water after interacting with organic matter and settles on the seabed, crustaceans and shellfish such as shrimp, crabs, prawns, and oysters are being and even within studies may reflect variations in sensitivity to copper within normal populations (Elliot, Frio, & Jarman, 2017).
Of the copper compounds tested, Cu OAC had the most striking effect on both cell lines, causing significant sustained ROS, while F I G U R E 1 0 Fluorescent aggresome images of HT29 (a) and Caco-2 (b) after a 2-hr exposure and 24-and 10-hr recovery, respectively. The aggresome detection reagent shown in red is present close to nucleus (stained with Hoechst). Images were combined with MetaMorph and are representative of at least five replicates for each cell line. MG132 was used as a positive control that inhibits proteosomal activity.  Proteomic expression profiling after 10-hr copper recovery showed further differences between the copper sources. As expected, there were shared changes in proteins involved in oxidative stress and copper storage but there were also copper source-specific and uniquely altered protein expression patterns noted. It is perhaps no coincidence that Cu OAC caused the greatest proteomic changes given the sustained ROS generation. It should be noted that as only one time point was analyzed, it is possible that proteins involved in very early response may have been missed. Indeed, the changes in the transporters detected at the end of the two-hour exposure were not detected after the 10-hour recovery.
Shared protein changes involved in oxidative stress were sorbitol dehydrogenase (SORD) in HT29 and HMOX1 in Caco-2. SORD is a component of the polyol pathway and increased activity resulted in increased peroxynitrite and reduced glutathione (GSH) (Tang et al., 2010). Heme oxygenase 1 (HMOX1) is essential in heme catabolism, and increased expression is induced through nrf2 signaling during oxidative stress (Wang et al., 2012) and has also been implicated in lipid peroxidation (Peña, Colbenz, & Kiselyov, 2015). While HMOX1 expression was increased in all copper sources in Caco-2 cells, this did not reflect the ROS levels as detected by DCFDA. Metal storage or detoxification by metallothionein (MT) expression was observed in both cell lines, being the most increased protein both cell lines.
Metallothionein is a small protein (6-7 kDa) with up to nine copper binding sites that allows it to mop up excess metals especially zinc and copper.
Reactive oxygen species generation and oxidative stress are also implicated in pro-inflammatory signaling cascades with extensive cross talk between NF-ĸB and the antioxidant response transcription factor nrf2 reported (Wardyn et al., 2015). In this study, increased expression of HMOX1 and Sequestosome-1 suggest activated nrf2 signaling. Activation of NF-κB results in the secretion of inflammatory signals (IL-6, IL-8, and PGE2), as well as activation of ERK/JNK signaling and changes in expression of a variety of proteins (Morgan & Liu, 2011). In this study, while whole cell lysates at a 10-hour recovery would not be suited to detecting changes in secreted proteins or detecting changes in phosphorylation of ERK/ JNK, changes in levels of antioxidant targets of NF-κB including F I G U R E 11 Long-term cellular growth response after copper treatment. Cells exposed to copper compounds for 2 hr, were allowed to recover, and trypsinized after 24-, 48-and 72-hr recovery.  (Alcaraz et al., 2004).
Protein changes specific to organic copper sources included six proteins that were significantly differentially expressed in HT29, with only one protein, HSPA6, noted in Caco-2 and previously detected in response to proteosomal stress (Noonan, Gregory Fournier, & Hightower, 2008). In HT29, SQSTM1, AKR1B15, PICALM, SRXN1, and MT1E and MT1G were only significantly altered following treatment with the organic copper compounds.
However, given the small differences between the organic copper compounds and CuSO 4 and the very variable response with the inorganic copper product, these may not be suitable markers to monitor organic copper overload.
Some of the greatest changes in proteins involved endoplasmic reticulum function, unfolded protein response (UPR), and proteosomal activity, and this was further noted by increases in high molecular weight ubiquitinated proteins particularly in Cu OAC (Figure 8).
Proteosomal overload leads to increased amounts of ubiquitinated misfolded proteins that interact with each other to form aggregates causing proteotoxicity. To alleviate this, aggregates can be degraded by autophagy (Lamark & Johansen, 2012) or concentrated into aggresomes if excessive. Aggresomes are depots of aggregated misfolded proteins stored in a perinuclear location as a protection mechanism from misfolded proteins (Johnston, Ward, & Kopito, 1998) and have been reported in metal-treated cells (Sahni, Bae, Jansson, & Richardson, 2017). The aggresome-autophagy system is induced as a protein quality control mechanism. When cells are no longer stressed, the aggresome is broken down into microaggregates which can be digested by autophagy (Gao et al., 2016).
The increase in the autophagy adaptor proteins SQSTM1 and LC3-I/LC3-II did not correlate with a block in autophagy or with increased autophagy as there was no increase in autophagy structures (autolysosomes or early autophagic compartments). Indeed, SQSTM1, LC3-II, and BAG3 have also been implicated in aggresome formation (Calderilla-Barbosa et al., 2014) and were increased particularly in Cu OAC suggesting aggresome formation. In HT29, dynein, a retrograde transporter carrying aggregates to an aggresome, was increased on Cu OAC treatment. In both cell lines, further changes in heat-shock proteins involved in the UPR (DNAJA1, HSPA1B, DNAJB1, HSPA6, and HSPH1) suggest a common mechanism of Cu OAC action.
While aggresomes are observed to have a protective role in the cell, aggresomes can interfere with proteosomal activity (Höhn & Grune, 2013) and long-term exposure have led to DNA damage and cell cycle arrest (Lu, Boschetti, & Tunnacliffe, 2015).
Evaluation of the longer-term effect on cell growth revealed Cu OAC treatment to have a negative long-term impact on cellular growth of Caco-2 and may be related to aggresomes detected Cu OAC-treated cells.
In summary, detection by DCFDA indicated differences in the ability of the organic and inorganic copper sources to stimulate ROS. Unraveling changes at the proteomic level proved most interesting with not only differential responses being noted between the two different cell lines, but also in a product-specific fashion.
Divergences were noted in terms of the cell-specific copper uptake and in the expression of copper-related transporters, autophagy activity, and aggresome formation. Specifically, Cu OAC appeared to induce undesirable reactions due to sustained ROS and the genera-

E TH I C A L S TATEM ENT
Human and animal testing is not applicable to this study.