FAM46C and FNDC3A Are Multiple Myeloma Tumor Suppressors That Act in Concert to Impair Clearing of Protein Aggregates and Autophagy

Multiple myeloma, the second most common hematologic malignancy, is caused by the accumulation of abnormal plasma cells. Multiple myeloma cells retain the plasma cell capability to synthesize and secrete immunoglobulins (Ig; ref. 1). Ig mRNAs are translated by endoplasmic reticulum (ER) resident ribosomes and undergo conventional secretion. Nascent Ig chains translocate to the ER lumen where they are folded. During this process, fractions of Igs remain unfolded. The accumulation of unfolded proteins triggers the unfolded protein response (UPR), a three-branch mechanism that maintains ER homeostasis (2). As part of the UPR process, unfolded proteins are retro-translocated from the ER to the cytoplasm and degraded by the proteasome (3). Indeed, proteasome inhibitors are highly effective for multiple myeloma treatment (4, 5), before clinical resistance develops (6).

Accumulation of cytoplasmic proteins that escape proteasome digestion can trigger the formation of intracellular aggregates, also known as aggresomes. Aggresomes can be degraded by autophagy, an intricated pathway of cellular events that results in the clearing of double membrane vesicles by the lysosomal degradative pathway. Nowadays, most studies converge on the concept that autophagy and the ubiquitin-proteasome system are integrated (7) and cooperate to clear ubiquitinylated targets. In addition, autophagy relevant proteins possess activities that intervene with cellular functions linked to membrane biology, such as endocytosis, intracellular vesicular trafficking, and conventional and nonconventional secretion (8).

Genetic analysis has shown that multiple myeloma cells have frequent loss-of-function mutations in genes that are rarely mutated in other cancers (9–11). One of these genes is FAM46C that is mutated in more than 10% of patients with multiple myeloma (9–11). FAM46C induces apoptosis in multiple myeloma cell lines (12). FAM46C is a member of a gene family composed of four highly similar proteins, FAM46A, FAM46B, FAM46C, and FAM46D. With the exception of FAM46D, which is lost in 3% of patients with gastric cancer (13), no mutations of other FAM46 family members have been observed in cancer. Understanding the reason that underlies the specific loss of FAM46C in multiple myeloma may open the avenue for specific therapies.

Unbiased high-throughput screening picked up FAM46C as an IFN-regulated modulator of viral production. In some cases, FAM46C overexpression mildly increased viral production, as for yellow fever virus, or had no effect, as with hepatitis C virus (14). In other cases, FAM46C strongly inhibited viral propagation, as in the case of the influenza virus, H1N1 (15). These observations suggest that the proviral or antiviral effects of FAM46C may depend on specific differences in the way viral particles replicate and egress, rather than on a common process. In this context, autophagy modulation plays important roles in viral intracellular amplification (16).

An early in silico analysis predicted that FAM46 proteins constituted a noncanonical terminal transferase (NT) family containing PAP/OAS1 SBD domains (17). Structural resolution of FAM46B, a FAM46C paralog, did not confirm the existence of PAP/OAS1 domains (18) and suggested structural homology to bacterial nucleotidyl transferases. Interestingly, motif analysis of FAM46B scores the presence of VHS (19), GAT (20), and GAE domains. These domains regulate trafficking pathways for cargo retrieval and degradation (21). In short, structural studies suggest that FAM46 family members may also interact with the trafficking machinery.

The NT activity of FAM46C was reported to add short A-tails to the 3′ untranslated region of ER bound mRNAs encoding for Igs (22, 23). These data favored the model that FAM46C increases mRNA stability and translation of Ig mRNAs at the ER, increasing IgG secretion (22–24). These studies, largely based on the comparison of wild-type (wt) FAM46C with mock controls, did not show whether all ER bound mRNAs increased their translational efficiency, and left the unresolved question on whether FAM46C directly binds mRNA. In addition, an unbiased high-throughput screening for secretion modifiers revealed that FAM46A, a close paralog of FAM46C, represses conventional secretion rather than increasing it (25). Finally, the recent observations that FAM46C may repress oncogenic Akt signaling (26, 27) seem consistent both with its tumor suppressor role and with repression of protein synthesis, given the well-known stimulatory role of the PI3K pathway on translation (28). Thus, alternative models for FAM46C function may explain how the tumor suppressor function of FAM46C is specifically linked to the environment of multiple myeloma.

To answer why FAM46C is lost in multiple myeloma, we focused on the differences between the phenotypes induced by the wt and mutant forms of the protein and on the molecular partners of FAM46C. We conclude that FAM46C functions at the cross-roads between secretion and autophagy. Our data indicate that FAM46C requires the presence of FNDC3A for its tumor suppressor function and its association with the ER. The FAM46C/FNDC3A complex is cytotoxic in multiple myeloma cells, where it impairs autophagy, causes accumulation of aggregates, and apoptosis.

All the cell lines used in this study were obtained from the ATCC. Cell line authentication was performed through qRT-PCR on specific target genes and testing for Mycoplasma contamination was performed monthly through PCR.

Cell lines at passages greater than 10 were not used for the experiments described in this study.

Detailed description of cell lines is provided in Supplementary Materials and Methods.

Cell lines expressing inducible constructs of wt and mutant FAM46C alleles were induced by administration of 2 μg/mL doxycycline hydrochloride (Sigma, catalog no., D3447) for up to 8 days. Lentiviral vectors expressing short hairpin RNAs (shRNA) for FAM46C and FNDC3A were obtained by calcium phosphate transient transfection of HEK293T cells. FAM46C and FNDC3A downmodulation was obtained by infection of multiple myeloma cell lines with the respective shRNA lentiviral vectors. Plasmids, lentiviral vectors, transfection, and infection procedures are all detailed in Supplementary Materials and Methods.

Immunofluorescence stainings were performed on multiple myeloma LP-1 cells and HEK cells, using anti-calnexin, anti-FLAG, anti-HA, and anti-LC3-B antibodies. FACS analysis was performed on multiple myeloma and HEK cells following propidium iodide, aggresome, autophagosome, or LAMP1 staining.

Electron microscopy was performed on LP-1 cells expressing either wt FAM46C or the D90G-mutant allele. The regulatory T cell (Treg) in vitro assay was performed using conditioned media from multiple myeloma cell lines expressing either wt FAM46C or the D90G-mutant allele and Treg suppression inspector beads as described in Supplementary Materials and Methods.

Detailed procedures for each technique can be found in the Supplementary Materials and Methods section.

OPM-2 xenografts were obtained by injecting NOD scid gamma male mice. All experimental procedures performed complied with national regulations and ethical approval was obtained by IACUC688. Methods and IHC are detailed in Supplementary Materials and Methods.

Samples derived from LP-1 multiple myeloma cells were prepared for gel filtration analysis as detailed in Supplementary Materials and Methods. Nucleus, cytoplasm, and ER isolation was performed using the NE-PER Nuclear Cytoplasmic Extraction Reagent Kit (Pierce) or the Endoplasmic Reticulum Isolation Kit (Sigma, catalog no. ER0100), respectively, as described in Supplementary Materials and Methods. Western blotting and the full list of antibodies used is listed in Supplementary Materials and Methods. The poly(A) tail-length assay was performed using the Poly(A) Tail-Length Assay Kit (Thermo Fisher Scientific, catalog no. 764551KT), using the oligonucleotides listed in Supplementary Materials and Methods. Polysome profiles were prepared by ultracentrifugation of purified polyribosomes loaded on 15%–50% sucrose gradients. The subsequent analysis of translation through polysomal profiling is fully described in the Supplementary Materials and Methods section.

The double immunopurification, mass spectrometry, secretomics, and SWATH analyses were performed on LP-1 multiple myeloma or HEK cells. Details are described in Supplementary Materials and Methods.

RNA sequencing (RNA-seq) was performed on total RNA samples or on RNA extracted from ER fractions.

Microarray analysis was performed on total RNAs. RNA extraction techniques, RNA-seq analysis, datamining procedures, and microarray analysis are all detailed in Supplementary Materials and Methods.

All quantitations are expressed as means ± SD. Statistical P values were calculated by t tests, NS, P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001. The exact type of t test used can be found in the legend corresponding to each figure.

We screened for multiple myeloma cells lacking functional FAM46C to reexpress near-physiologic levels of the wt allele. For the screening of endogenous levels of FAM46C, we had to rely on mRNA data and genetic analysis, because we could not specifically detect the endogenous protein, using both in-house produced and three commercially available antibodies, including those from published sources. Examples of these observations are in Supplementary Fig. S1A and S1B that show (i) the lack of correlation between FAM46C protein staining and mRNA levels and (ii) the lack of loss of FAM46C staining upon shRNA treatment.

After the genetic screening, two cell lines were selected: the LP-1 cell line, which harbors a homozygous deletion of FAM46C and the OPM-2 cell line, which expresses a mutated FAM46C allele accompanied by LOH (Supplementary Fig. S1C). For analysis of FAM46C-induced effects, we produced a doxycycline-inducible FLAG-tagged version of the gene. We made sure that the expression levels of the mRNA of C-TERM-FLAG-FAM46C were comparable with those of endogenous FAM46C, at least at the mRNA level (see Supplementary Fig. S1D), and then opted for confronting wt FAM46C with selected mutant alleles of FAM46C, as found in tumors (Supplementary Fig. S1E). Given the fact that all mutants gave similar effects on apoptosis and cell-cycle progression, the point mutant variant, D90G, one of the most frequently found in patients with multiple myeloma (Supplementary Fig. S1E; ref. 11), was consequently used throughout the project. This approach resulted in a set of observations that pinpoint the differences between tumor suppressor FAM46C and a FAM46Callele that has lost tumor suppressor functions.

Expression of the wt form of FAM46C slowed down cell duplication (Fig. 1A), caused a modest reduction in G₁–S-phase cell-cycle progression (Supplementary Fig. S1F), inhibited the colony-forming capacity of multiple myeloma cells plated on soft agar up to 65% (Fig. 1B), and robustly induced apoptosis (Fig. 1C). Expression of neither the D90G-mutant allele nor other alleles, frequently found in patients with multiple myeloma (Supplementary Fig. S1E), had apoptotic effects (Fig. 1C; Supplementary Fig. S1G and S1H). Given the fact that FAM46C is mutated only in multiple myeloma, we asked whether the proapoptotic effect of FAM46C could be extended to other cell lines or not. Expression of FAM46C in HEK293 cells did not trigger apoptosis (Fig. 1D), suggesting that FAM46C-induced apoptosis is not due to a general effect.

FAM46C is proposed to polyadenylate and stabilize mRNAs (23). We, therefore, compared the effects of wt FAM46C with those of the D90G mutant and of a mock control, on gene expression and on the capability to add a poly(A) tail to an acceptor substrate (23). First, we checked whether specific mRNAs were more abundant in cells expressing wt FAM46C versus those expressing the D90G mutant or the mock. Strikingly, reexpression of wt FAM46C in OPM-2 cells induced a minimal modulation of mRNAs at the single gene level, compared with D90G-expressing cells (Supplementary Table S1). To detect coordinated changes of gene expression that may give a hint on the proapoptotic changes driven by FAM46C, we performed gene set enrichment analysis (GSEA). GSEA revealed that wt FAM46C expression triggered, when compared with the expression of the D90G mutant and the mock, an ER stress and UPR signature (Fig. 1E), in agreement with (12). Many of the induced genes were ATF6 targets (Fig. 1E; annotated Supplementary Table S1, sheets FAM vs. MOCK and UPR). Chop (DDIT3) was mildly, but always, significantly induced (Supplementary Table S1, sheet CHOP increase).

We tested whether FAM46C was able to add poly(A) to (A)20 RNA substrates when compared with the D90G mutant. We found that, the polyadenylation activity of wt FAM46C was (i) just above the background level when compared with mock controls and (ii) identical to the activity of the loss-of-function D90G mutant. In the same assay, with as little as 1/10 of exposure time, the activity of E. Coli Poly(A) polymerase was very robust (Supplementary Fig. S1I). Next, we tested whether FAM46C bound mRNAs, by performing CLIP analysis. We cross-linked radioactively labeled RNA to the immunoprecipitated FAM46C, and then analyzed the presence of RNA by autoradiography, either naïve or after urea washes. Urea washes allow to discriminate whether FAM46C directly and specifically binds mRNAs. FAM46C bound mRNAs, however, urea washes completely abolished the binding, indicating that FAM46C binding is not direct (Supplementary Fig. S1J).

The combination of (i) the strong proapoptotic role of the wt versus the mutant, with the lack of a robust effect on the expression levels of specific mRNAs, with (ii) the lack of difference between the transferase activity of the wt versus the mutant, suggests that polyadenylation cannot account for the tumor suppressor phenotypes induced by wt FAM46C.

However, in agreement with (12, 23), our results demonstrate that FAM46C is a tumor suppressor in multiple myeloma cells, where it induces a peculiar ER stress response through unknown mechanisms.

We characterized the relationship between FAM46C and ER stress by defining its cellular localization and the UPR. First, we addressed FAM46C cellular localization. By cell fractionation, we found that FAM46C is cytoplasmic (Fig. 2A; Supplementary Fig. S2A), and barely detectable in the nucleus. Nuclear FAM46C could be ascribed to contamination because: (i) it correlated with residual cytoplasmic markers and (ii) it is never visible by immunofluorescence (Fig. 2B; Supplementary Fig. S2B). Immunofluorescence experiments in multiple myeloma cells and epithelial cells demonstrated that FAM46C localizes in the cytoplasm, in an external ring, partly overlapping with ER marker, calnexin (Fig. 2B; Supplementary Fig. S2B). The localization of FAM46C in close apposition with the ER is consistent with the ER stress signature and leads to two questions, (i) why and (ii) how FAM46C is sorted to the ER in the absence of ER localization signals.

We addressed the first question. Classically, the UPR is described by three connected branches, PERK-eIF2α, XBP1 splicing, and ATF6 translocation, that converge on CHOP (DDIT3) transcription. DDIT3 mRNA was increased in cells expressing wt FAM46C versus the D90G mutant (Supplementary Fig. S2C). We did not detect clear changes in XBP1 splicing (Supplementary Fig. S2D). As described previously, ATF6 targets were upregulated (Fig. 1E; Supplementary Tables S1 and S2–S5). SESN2 is an important target, downstream of ATF6 activation (29). Indeed, we confirmed in multiple conditions that SESN2 mRNA is highly upregulated by FAM46C expression (Supplementary Fig. S2E). These data suggest that the ER stress signature driven by FAM46C occurs mainly through ATF6.

Next, we wanted to define the closest cellular event connected to ER stress. Analytically, an increase in ER stress can be caused by increased proteasomal load, which results from increased protein synthesis (Supplementary Fig. S2F). Analysis of ER polysomes showed that FAM46C expression did not stimulate initiation of translation (Supplementary Fig. S2G), moreover, expression of FAM46C had no detectable effect on eIF2α phosphorylation (Fig. 2C), a marker of unfolded protein accumulation in the ER. A second possibility is that, FAM46C affects proteasomal efficiency (Supplementary Fig. S2F). We found that polyubiquitin accumulation did not increase in wt FAM46C–expressing cells versus D90G-expressing cells (Fig. 2D), moreover, the sensitivity to proteasome inhibitors was not altered in wt FAM46C–expressing cells (Supplementary Fig. S2H). Taken together, these data rule out increased protein synthesis load in the ER.

Next, we asked whether FAM46C could induce accumulation of protein aggregates that escaped the proteasome system. By using the ProteoStat reagent (30), which detects hydrophobic stretches of aggregated cytoplasmic proteins, we found that wt FAM46C induces a steady-state accumulation of protein aggregates that peaks 4 days after induction (Fig. 2E and F). The formation of aggregates induced by FAM46C expression was confirmed in several cell lines and, consistently, FAM46C downregulation reduced aggregate accumulation (Supplementary Fig. S2I–S2K).

Given the role of autophagy in clearing protein aggregates (31), we checked whether FAM46C expression affected autophagy. The analysis was performed on different cell lines, with similar results. It is known that p62, the classical autophagy receptor, accumulates during autophagic inhibition. We found that FAM46C-expressing cells had an increase in p62 protein abundance, as compared with the mock or to the D90G mutant, as shown in the FAM46C reconstituted cell line, LP-1 (Fig. 3A). The number of LC3B puncta is a marker of autophagy. So, next, we double-stained nonsynchronized multiple myeloma cells for LC3B and FAM46C, and quantitated LC3B puncta. The expression of wt FAM46C led to a reduction in LC3B puncta when compared with expression of the D90G mutant (Fig. 3B). Next, we measured the rates of lysosomal degradation by monitoring the ratio of LC3B-II/LC3B-I, which is a specific indicator of autophagic flux, in the presence or absence of autophagic inhibitor, bafilomycin A1, which blocks autophagosome–lysosome fusion causing accumulation of autophagosomes. We found that LP-1 cells that reexpressed wt FAM46C had less LC3BII accumulation when compared with cells expressing the D90G-mutant allele (Fig. 3C) both in the presence and absence of bafilomycin A1, indicating a reduction in the autophagic flux. We also showed the inverse correlation by performing shRNA experiments on the RPMI cell line, which naturally expresses normal levels of wt FAM46C. We found that 50% FAM46C downregulation (Fig. 3D, right) caused an increase in the LC3B-II/LC3B-I ratio (Fig. 3D, left). Reconstitution of FAM46C in OPM-2 cells decreased the LC3BII/LC3BI ratio in the presence of bafilomycin A1 (Supplementary Fig. S3A). To further understand the dynamics of autophagic inhibition driven by wt FAM46C, we performed FACS analysis using a monodansylcadaverine derivate, which labels autophagic vesicles. We found that FAM46C-expressing cells had a reduction in the number of autophagosomes (Fig. 3E). Depletion of FAM46C with two shRNAs increased autophagosome formation in the U266 multiple myeloma cell line (Supplementary Fig. S3B). Taken together, all these data show that the increase of FAM46C expression causes a decrease of autophagic fluxes in all the analyzed cells, and vice versa. Finally, we tested whether FAM46C could physically interact with the autophagic machinery by immunoprecipitation of FAM46C, followed by LC3B detection in the pulldowns. We found that FAM46C does not interact with LC3B (Supplementary Fig. S3C).

To check whether the in vitro results could be confirmed in vivo, we performed in vivo experiments in NOD scid gamma mice, which are more supportive to xenograft growth due to their complete immunodeficiency. Multiple myeloma xenograft models were established with the human multiple myeloma OPM-2 cell line harboring either wt FAM46C or the D90G-mutant doxycycline-inducible constructs (Fig. 4A). We evaluated the effects of FAM46C expression over an 8-day period after doxycycline administration. As expected, cells expressing wt FAM46C produced smaller tumors compared with cells expressing the D90G mutant (Fig. 4B), confirming the onco-suppressant role of FAM46C in vivo. Next, we characterized by IHC the tumor biopsies derived from our xenograft mice models. As predicted, FAM46C-expressing tumors had reduced proliferation, as shown by Ki67 staining, and increased apoptosis, as shown by detection of cleaved caspase-3 (Fig. 4C; Supplementary Fig. S3D). Moreover, we found a drastic increase in p62 abundance only in tumors expressing the wt form of the protein (Fig. 4C; Supplementary Fig. S3D). Taken together, our data indicate that the expression of FAM46C reduces autophagy, leads to the accumulation of intracellular aggregates in vitro, and disfavors tumor growth with concurrent accumulation of p62 in vivo. The lack of a direct interaction of FAM46C with the LC3B suggests that FAM46C effects are due to a cross-talk with the autophagic flux through unknown elements.

We reasoned that the analysis of FAM46C proteomic network could help us to develop testable hypotheses regarding its mechanism. We first addressed whether FAM46C was monomeric or part of a complex. By performing gel filtration chromatography followed by quantitative analysis, we found that 40% of FAM46C elutes with large molecular weight (MW) particles (Fig. 5A). As a positive control for gel filtration, we tested eIF6, which associates with ribosomes (Fig. 5A; ref. 32), and as expected, it eluted with both soluble and large MW particles. To find partners of FAM46C that are relevant for multiple myeloma progression, we combined (i) datamining of available databases, (ii) direct mass spectrometry studies using the inducible FAM46C construct, and (iii) mutational analysis of multiple myeloma databases. This stringent approach was meant to help pin down FAM46C interactors that have a role in multiple myeloma. We found that FNDC3A, an ER-associated protein, fulfilled all three criteria.

In short, we demonstrated the interaction between FNC3A and FAM46C in multiple cell lines and through several independent approaches (Fig. 5A–D; Supplementary Fig. S4A–S4H). Mass spectrometry identified FNDC3A as a FAM46C interactor (Supplementary Table S3). We showed that recombinant FAM46C is able to bind FNDC3A by in vitro pulldown assays (Fig. 5B). Immunoprecipitation of FAM46C leads to the coprecipitation of FNDC3A (Fig. 5C). Immunoprecipitation of FNDC3A leads to the coprecipitation of FAM46C (Supplementary Fig. S4A). The interaction between FAM46C and FNDC3A was not affected by RNase treatment (Supplementary Fig. S4B). Colocalization of FAM46C and FNDC3A was visible by immunofluorescence (Fig. 5D). The FAM46C–FNDC3A interaction was also seen in epithelial cells (Supplementary Fig. S4C) and in several other multiple myeloma cell lines (Supplementary Fig. S4D–S4F). These data indicate the unequivocal existence of a FAM46C/FNDC3A complex.

The carboxy-terminus of FNDC3A is similar to the one of ER “tail-anchored” proteins (33). Indeed, we found, by costaining of FNDC3A with the ER luminal marker, calnexin, that FNDC3A localized at the ER (Fig. 5E). These data are in agreement with the in silico topology of FNDC3A, which harbors a C-terminal transmembrane domain (Fig. 5F), and with proteomics, which further confirm its localization at the cytosolic side of the ER. The colocalization between FNDC3A and FAM46C was also confirmed in microsomal preparations and by fluorescence microscopy (Supplementary Fig. S4G and S4H). Given FNDC3A localization, we checked whether FAM46C subcellular localization depended on its interaction with FNDC3A. We exploited L288insG, a FAM46C mutant that does not efficiently bind FNDC3A (Fig. 5G), and is frequently found in patients with multiple myeloma (Supplementary Fig. S1E). FAM46C-L288insG mutant showed a disrupted ER localization (Fig. 5H). In conclusion, FAM46C interacts with FNDC3A and this interaction may be required for its localization at the ER.

Datasets of patients with multiple myeloma (34) show that low mRNA levels of either FNDC3A or FAM46C correlate with decreased overall survival (Fig. 6A; Supplementary Fig. S5A). Loss of FNDC3A is found in the multiple myeloma genomic DNA dataset of (11), encompassing 203 patients. FNDC3A mutations generate loss-of-function proteins at low frequency (4/203), all mutually exclusive with FAM46C mutations (Fig. 6B), suggesting a genetic interaction between FAM46C and FNDC3A. Given that the low frequency of FNDC3A mutations was not sufficient to rule out a random association with the maintenance of wt FAM46C, we tested their genetic interaction in vitro.

Mixing LP-1 multiple myeloma cells expressing wt FAM46C (90%) with cells devoid of FAM46C (10%) leads to the progressive loss of FAM46C expression, possibly because of the reduced fitness of FAM46C-expressing cells. The same experiment, when performed by mixing 90% of D90G FAM46C–expressing cells with 10% of cells devoid of D90G, does not lead to loss of D90G FAM46C expression (Supplementary Fig. S5B). Next, we tested whether depletion of FNDC3A delayed loss of FAM46C. Indeed, depletion of as little as 50% of endogenous FNDC3A, in this context, resulted in the retained expression of wt FAM46C (Fig. 6C and D), indicating that FNDC3A expression mediates the loss of fitness of FAM46C-expressing cells.

To further investigate whether FNDC3A acted as a modulator of FAM46C function and as a putative multiple myeloma tumor suppressor, we screened for multiple myeloma cell lines that had lost both alleles of FNDC3A (Supplementary Fig. S5C and S5D). We found that the Delta47 cell line had lost FNDC3A mRNA expression due to biallelic mutations. Reexpression of FNDC3A in Delta47 cells (Supplementary Fig. S5E) impaired cell growth (Supplementary Fig. S5F) and induced apoptosis (Fig. 6E, top). Analysis of apoptosis in various conditions of FAM46C expression confirmed that reinduction of (i) FAM46C in FNDC3A-mutated Delta47 cells and (ii) FNDC3A in FAM46C-mutated LP-1 cells did not significantly increase apoptosis, confirming the mutual necessity for both proteins (Supplementary Fig. S5G and S5H). Interestingly, neither FNDC3A expression in HEK293 cells trigger apoptosis (Fig. 6E, bottom), nor did overexpression of both FNDC3A and FAM46C (Supplementary Fig. S5I), thus confirming that the tumor suppressor function of the FAM46C/FNDC3A complex is limited to the multiple myeloma scenario.

Next, we analyzed whether the phenotype induced by the reexpression of FNDC3A in cells that had lost it, but retained normal levels of wt FAM46C, triggered similar effects to the one observed by FAM46C expression (Figs. 2–4). We, therefore, analyzed the gene expression signature, aggregate formation, and autophagy impairment. RNA-seq data indicated (Supplementary Table S2) that FNDC3A expression triggered a significant enrichment in gene ontology (GO) terms related to ER stress (Fig. 6F). FNDC3A reexpression also caused protein aggregate accumulation (Fig. 6G). To close the circle, we checked for FNDC3A involvement in autophagy. Basal autophagy in Delta47 cells was higher than in either OPM-2 or LP-1 cells; this said, expression of FNDC3A caused a reduction in the LC3B-II/LC3B-I ratio (Fig. 6H) and reduced the number of autophagosomes (Fig. 6I). Taken together, these data strongly suggest that the FAM46C/FNDC3A complex coordinately induces autophagic impairment, accumulation of intracellular aggregates, and apoptosis.

Next, we asked whether FAM46C/FNDC3A is part of a larger network of proteins and how it might reduce autophagy. We designed a multi-step purification assay based on a double-immunoprecipitation strategy, performing first, a FAM46C pulldown on total lysates, and then, FNDC3A immunoprecipitation on the FAM46C-purified protein (Supplementary Fig. S6A). Pulldown experiments were done both in cells not undergoing apoptosis, HEK293T, and cells undergoing FAM46-induced apoptosis (Fig. 1D), to discriminate for potential differential partners and common pathways. Data obtained is briefly described (Supplementary Table S3). Copurification was achieved for both FAM46C and FNDC3A (Supplementary Fig. S6B) in both cell lines (Supplementary Table S3). Enrichment analysis confirmed that a fraction of FAM46C is not part of the FAM46C/FNDC3A complex (Supplementary Fig. S6C), in-line with gel filtration experiments (Fig. 5A). Mass spectrometry analysis of the purified complex showed that the more prominent proteins were represented by FAM46C, FNDC3A, ER resident proteins, and trafficking proteins (Supplementary Table S3). We performed GO analysis on the list of FAM46C and FNDC3A interactors, and found a strong enrichment in terms related to vacuole structure and to extracellular exosomes (Fig. 7A). Overall, these data suggest that the function of the FAM46C/FNDC3A complex is related to the regulation of secretion, and as such it may indirectly interfere with autophagic secretion (8). We, therefore, analyzed which secretory pathway is affected by FAM46C.

To provide evidence that the tumor suppression capability of FAM46C was linked to secretory fluxes, we analyzed FAM46C-induced secretome, comparing it with the D90G mutant. To assess whether secretion depended on mRNA levels on ER polysomes, we performed secretomics paired with RNA-seq of ER-associated mRNAs (Fig. 7B; Supplementary Tables S4 and S5). By this approach, we may distinguish whether increased secretion of a protein follows the rate of translation of its mRNA or it is due to a trafficking alteration, namely by scoring whether the amount of secreted proteins trends with the amount of increased mRNA on the ER or not.

First, RNA-seq data of LP-1 cells expressing wt FAM46C, compared with the D90G mutants, fully confirms, as with OPM-2 cells, the presence of an ATF6 signature (Supplementary Table S5, sheet ALL DATA and GO). Unbiased analysis of IgG mRNAs does not confirm the existence of consistent variations in IgG mRNA levels (23) driven by FAM46C (Supplementary Table S5, sheet Igs). The secretome of FAM46C-expressing cells versus that of D90G-expressing cells contained several differentially secreted proteins (Supplementary Table S4). FAM46C expression caused increased secretion of proteins involved in alternative vesicle-mediated secretion pathways (Fig. 7C; Supplementary Table S4). The secretome showed strong association with the lysosome/exosome, containing Cathepsin B, Legumain, and DNase2, and with immune modulation (Supplementary Fig. S6D). The comparison between the mRNA levels present on the ER and the secretome indicated that the lysosomal enzymes, cathepsin B and legumain, were secreted more efficiently in the presence of increased mRNA levels on the ER (Fig. 7D; Supplementary Fig. S6E; Supplementary Table S4), but proteins from the secretory pathway, like SMOC1, were less secreted (Fig. 7D, left) in spite of significantly increased mRNA levels (Fig. 7D, right; Supplementary Table S4, sheetRNAseqprot). It has been reported that FAM46C increases IgG secretion through increased mRNA stability (23). Expression of wt FAM46C, compared with the D90G mutant, did not alter canonical Ig secretion (Supplementary Fig. S6F; Supplementary Table S4). On the same line, we tested by qRT-PCR on polysomes and PAT assays whether LGMN, DNase2, and CATB mRNAs had preferential translation and/or polyadenylation. Within the sensitivity limits of our approach, we confirmed transcriptional changes (Supplementary Fig. S6G, top), but we did not detect preferential loading on polysomes (Supplementary Fig. S6G, bottom) or increased polyadenylation (Supplementary Fig. S6H). In summary, our data confirm a connection between the FAM46C complex and secretion, and a cross-talk with autophagic fluxes by lysosomal rerouting or trafficking. In addition, the absence of a strict correlation between mRNA levels and the amount of the corresponding secreted protein favors the conclusion that wt FAM46C alters secretory fluxes independently from translational efficiency and mRNA stability. We also rule out a relationship between the tumor suppressor capability of FAM46C and the increased secretion of IgG proteins.

To test whether lysosomal exocytosis was increased by FAM46C expression, as suggested by DNas2 in the medium, we measured the levels of membrane bound LAMP1, a lysosomal marker enriched at the plasma membrane during lysosomal exocytosis (35). Plasma membrane bound LAMP1 was higher in FAM46C- versus D90G-expressing cells (Fig. 7E). Finally, we tested whether the immune modulation signature predicted by the bioinformatics analysis was accompanied by changes in the biological activity. Because conditioned media from multiple myeloma cells are capable of favoring differentiation of Tregs and immunosuppression, we tested whether Treg polarization could be inhibited by the FAM46C-driven secretome. Indeed, coculturing of multiple myeloma cells expressing wt FAM46C with differentiating CD4⁺ Tregs (Supplementary Fig. S6I, left), caused an overall reduction of their differentiation efficiency compared with coculturing with cells expressing the D90G-mutant allele (Supplementary Fig. S6I, center). The effect did not require cell-to-cell interaction, as it was also seen using conditioned media (Supplementary Fig. S6I, right). We conclude that the FAM46C/FNDC3A complex alters intracellular trafficking, and suggest that it diverts lysosomes from aggresome–autophagosome fusion to exocytosis.

In the attempt to visualize alterations of intracellular structures driven by trafficking or autophagic inhibition, we performed electron microscopy imaging on multiple myeloma cells expressing either wt FAM46C or the D90G mutant. We found that multiple myeloma cells expressing wt FAM46C had intracellular peculiarities, including dilated ER cisternae, vacuole structures, and small-sized vesicles, such as endosomes (Fig. 7F), which are compatible with altered secretion and/or vesicular transport and simultaneous accumulation of protein aggregates.

The previous experiments underlining changes in autophagic and secretory pathways, suggest that reconstitution of the FAM46C/FNDC3A complex may also alter the pharmacologic sensitivity of multiple myeloma cells. To unveil potential pharmacologic weaknesses of multiple myeloma cells expressing a functional FAM46C/FNDC3A complex, we performed a connectivity map analysis (35) that predicted sensitivity to inhibitors of the sphingosine kinase pathway, DL-PMP and SA-792728 (Supplementary Fig. S7A). Hypothesizing that the complex could affect the sensitivity to sphingosine kinase inhibitors, we tested the effects of sphingosine kinase inhibition on cells reconstituted or depleted for the FAM46C/FNDC3A complex. Treatment with SK1-I, which inhibits both SK1 and SK2, inhibited cell growth of FAM46C-expressing LP-1 cells (Supplementary Fig. S7B) and OPM-2 cells (Supplementary Fig. S7C) compared with D90G-expressing cells. In agreement with these data, the downmodulation of FAM46C in U266 multiple myeloma cells, which express physiologic levels of wt FAM46C, caused a decrease in the sensitivity to sphingosine kinase inhibitors (Supplementary Fig. S7D). Reexpression of FNDC3A in Delta47 cells also increased sensitivity to SK1-I (Supplementary Fig. S7E) treatment, thus suggesting that the FAM46C/FNDC3A complex alters sensitivity to specific inhibitors.

In this work, we have collected data that indicate the existence of the novel FAM46C/FNDC3A complex on the cytoplasmic side of the ER. The FAM46C/FNDC3A complex alters trafficking and secretion and, in multiple myeloma cells, it interferes with the autophagic flux, leading to aggregate formation and apoptosis. Indeed, it is well-known that autophagy relevant genes broadly interact with the trafficking machinery (36). The FAM46C/FNDC3A complex, in multiple myeloma cells, may steer lysosomes from fusion with autophagosomes to the plasma membrane, thus impairing autophagy.

FAM46C is lost or mutated in more than 10% of patients with multiple myeloma, but not in other cancers (9–11), raising the intriguing question of which specific process makes multiple myeloma cells sensitive to FAM46C expression. To answer this question, we characterized the suppressor function of wt FAM46C in comparison with loss-of-function alleles, such as the D90G mutant. We then searched for FAM46C potential role by combining proteomics, genetics, and cell biology data. Our conclusion is that, the proapoptotic role of FAM46C in multiple myeloma cells is accompanied by an accumulation of protein aggregates because of impaired autophagy. Our data sustain the hypothesis that FAM46C is a part of a multi-protein network that regulates secretion, and its interaction with FNDC3A is necessary for its tumor suppressor role. Three results confirm the physiologic relevance of the FNDC3A/FAM46C complex: (i) reexpression of FNDC3A in multiple myeloma cells that had naturally lost the gene, recapitulates the effects of FAM46C reexpression, (ii) depletion of FNDC3A in cells with functional FAM46C leads to increased survival, and (iii) loss of interaction between FAM46C and FNDC3A results in defective FAM46C localization. Finally, despite the low frequency of FNDC3A mutations in patients with multiple myeloma, high FNDC3A levels positively correlate with OS. In summary, FAM46C/FNDC3A is a new multiple myeloma–specific tumor suppressor complex.

The proapoptotic role of FAM46C in multiple myeloma is well-established (12, 23, 24), as well as the high incidence of FAM46C mutations (9–11). All the studies published so far reveal that expression of FAM46C in multiple myeloma cells causes an ER stress signature (12, 26). One possible cause of this signature was an increased proteasome burden because of the increased translation of IgG mRNAs. In general, we and others have confirmed that the sensitivity of multiple myeloma cells to proteasome inhibitors is not affected by FAM46C expression (12, 26). The lack of increased sensitivity to proteasome inhibitors is in-line with the observations that wt FAM46C does not induce a general increase of Ig mRNA translation on ER resident ribosomes. It should be noted that this finding does not challenge the general conclusion that, through other mechanisms, the expression of FAM46C modulates the transition from immature B cells to antibody producing plasma cells (22). The transition from a B cell to an antibody producing plasma cell requires a striking remodeling of the trafficking apparatus, with an increase in the ER and of the endocytic pathway, which is required for membrane homeostasis. FAM46C increases lysosomal secretion, reduces the autophagic flux, causes accumulation of protein aggregates, and triggers programmed cell death. All these phenomena suggest that, by diverting lysosomal trafficking toward secretion, rather than to autophagosome–lysosome fusion, FAM46C may inhibit a prosurvival autophagy route. Finally, the autophagic inhibition driven by FAM46C was proapoptotic only in multiple myeloma cells. This observation indicates that the FAM46C/FNDC3A complex alters secretory pathways that interfere with lysosomal activity, but also that this event becomes rate limiting for survival only in the setup of a cell that is subject to proteotoxic stress.

The physiologic role of the FAM46C/FNDC3A complex is intriguing. Several data point to the existence of multiple FAM46/FNDC3 complexes that regulate trafficking and secretion. FAM46C is part of a family of proteins that includes FAM46A, FAM46B, and FAM46D. FAM46C knockout leads to the disassembly of the manchette, a structure essential for trafficking (37) and FAM46A depletion affects BMP signaling (38). Analysis on the protein sequence of FAM46B homolog, F7E7M3 (http://www.rcsb.org/pdb/protein/F7E7M3), shows the presence of VHS (19), GAT (20), and GAE domains (39), which are typical of proteins necessary for transport and lysosomal targeting. Concerning FAM46 family partners, high-throughput proteomic studies have shown that FAM46A interacts with both FNDC3A and a closely related protein FNDC3B (40). Secretion of surfactant-associated proteins that are associated with lamellar bodies, lysosome-related organelles, is decreased in the lung of FNDC3B knockouts (41). FNDC3A depletion alters the deposition of the extracellular matrix (42), a phenomenon requiring remodeling by secreted proteases. Finally, genetic studies suggest the existence of specific complexes in specific organs. Deletion of either FNDC3A (33) or FAM46C causes male sterility and abnormal sperm maturation (37), whereas a bone defect is observed with either FAM46A or FNDC3B knockouts (43, 44). In the context of B cells, lysosomal exocytosis has been suggested to regulate antigen extraction at the immunologic synapse (45). In conclusion, each cell may have a unique combination of FAM46 and FNDC3 isoforms that may become rate limiting for secretion-related phenotypes.

An unbiased search for potential sensitivities of FAM46C-expressing cells, by connectivity map analysis (46), also predicted a potential sensitivity to sphingosine kinase inhibitors. Experimental data confirmed the prediction, because we provide evidence that the expression of physiologic levels of the FAM46C/FNDC3A complex increases sensitivity to SK1-I. The connection between multiple myeloma, FAM46C, lysosome biology, and sphingosine kinases is new, but not totally surprising. Sensitivity of multiple myeloma cells to sphingosine kinase inhibitors has been recently proposed (47, 48). ATF6, a major mammalian UPR sensor, is activated by specific sphingolipids (49), and is also robustly induced by reexpression of either wt FAM46C or FNDC3A. Gaucher disease, which affects lysosomal processing of sphingolipids has an increased and unexplained risk of development of multiple myeloma (50). We hypothesize that clinical studies of their efficacy may take in account genetic data regarding the components of the FAM46C/FNDC3A network.

The final exciting question relates to the enzymatic activity of FAM46C. Our work shows a lack of correlation between FAM46C nucleotidyl transferase activity and its tumor suppressor capability when compared with the D90G loss-of-function mutant. Is there a possibility that the enzymatic activity of FAM46C is different? Structural and biochemical studies have shown that FAM46B, a FAM46C paralog, has a much higher nucleotidyl transferase activity compared with FAM46C and similarities to bacterial nucleotidyl transferases (18). One possibility is that the D90G mutation alters the specificity of FAM46C to unknown substrates. In this context, the known superfamily of transferases capable to bind ATP, the substrate of FAM46B (18), is composed of more than 40 different subfamilies including protein kinases, lipid kinases, and others. Up to now, in spite of the lack of homology of FAM46 members to eukaryotic nucleotidyl transferases, we have only tested the possibility that FAM46C acts as a nucleotidyl transferase. The weak activity of FAM46C (18) may be, therefore, due to the fact that its correct substrate still needs to be identified.

G. Tonon reports personal fees from Janssen (invited presentation) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

N. Manfrini: Conceptualization, formal analysis, validation, investigation, methodology, writing-original draft, writing-review and editing. M. Mancino: Conceptualization, formal analysis, investigation, writing-original draft, writing-review and editing. A. Miluzio: Formal analysis, investigation, visualization, writing-review and editing. S. Oliveto: Investigation, visualization. M. Balestra: Investigation. P. Calamita: Investigation. R. Alfieri: Data curation, formal analysis. R.L. Rossi: Data curation, formal analysis. M. Sassoè-Pognetto: Formal analysis, investigation, visualization. C. Salio: Formal analysis, investigation, visualization. A. Cuomo: Data curation, formal analysis. T. Bonaldi: Formal analysis, investigation, writing-review and editing. M. Manfredi: Resources, formal analysis, investigation. E. Marengo: Project administration, writing-review and editing. E. Ranzato: Formal analysis, investigation. S. Martinotti: Formal analysis, investigation. D. Cittaro: Formal analysis, investigation. G. Tonon: Conceptualization, formal analysis, writing-review and editing. S. Biffo: Conceptualization, resources, formal analysis, supervision, funding acquisition, writing-original draft, writing-review and editing.

We would like to thank Ilaria Mariani and Deborah Salvi Mesa for their initial help with FNDC3A downmodulation experiments and fluorescence microscopy experiments; Alessia Tommasini and Eugenio Graceffo for technical support in mass spectrometry analysis of FAM46C and FAM46C/FNDC3A interactomics; Mariacristina Crosti for her help with FACS experiments; Chiara Cordiglieri for supervision with fluorescence microscopy; and Simone Cenci and Enrico Milan for help with early autophagic analysis. This article was supported by grant AIRC IG 19973 to S. Biffo, by AIRC 9965 5‰ to G. Tonon, by MIUR project “Dipartimenti di Eccellenza 2018–2022” to Department of Neuroscience “Rita Levi Montalcini” to M. Sassoè-Pognetto and by unrestricted grant from “Fondazione Romeo ed Enrica Invernizzi.”

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.