Fine-Tuning Amyloid Precursor Protein Expression through Nonsense-Mediated mRNA Decay

Studies on genetic robustness recently revealed transcriptional adaptation (TA) as a mechanism by which an organism can compensate for genetic mutations through activation of homologous genes. Here, we discovered that genetic mutations, introducing a premature termination codon (PTC) in the amyloid precursor protein-b (appb) gene, activated TA of two other app family members, appa and amyloid precursor-like protein-2 (aplp2), in zebrafish. The observed transcriptional response of appa and aplp2 required degradation of mutant mRNA and did not depend on Appb protein level. Furthermore, TA between amyloid precursor protein (APP) family members was observed in human neuronal progenitor cells; however, compensation was only present during early neuronal differentiation and could not be detected in a more differentiated neuronal stage or adult zebrafish brain. Using knockdown and chemical inhibition, we showed that nonsense-mediated mRNA decay (NMD) is involved in degradation of mutant mRNA and that Upf1 and Upf2, key proteins in the NMD pathway, regulate the endogenous transcript levels of appa, appb, aplp1, and aplp2. In conclusion, our results suggest that the expression level of App family members is regulated by the NMD pathway and that mutations destabilizing app/APP mRNA can induce genetic compensation by other family members through TA in both zebrafish and human neuronal progenitors.


Introduction
Understanding gene function has fascinated scientists since DNA and the transfer of genes between generations were described.Inbreeding, or forward genetics, by which a phenotype of interest was searched among individuals with random genome-wide mutations or by generating targeted genetic mutations, has been used to elucidate gene function.While the role of some genes has been easy to assess, others have been challenging due to redundant gene function and compensatory mechanisms.One such example is the amyloid beta (A4) precursor protein (App) and the amyloid precursor-like protein (Aplp)-1 and Aplp2 in mice, whose single mutant gene knockouts show only minor phenotypes, while combined knock-out mutations are lethal (von Koch et al., 1997;Heber et al., 2000).While redundancy allows one protein to functionally replace another protein if lost, genetic compensation ensures adjusted expression levels.This type of genetic robustness is an essential system enhancing survival of organisms despite harmful mutations (Kitano, 2004;Felix and Barkoulas, 2015).Recent studies support the existence of a new mechanism governing genetic robustness, described as transcriptional adaptation (TA) or genetic compensation response (GCR; Kimmel et al., 1995;Zhu et al., 2017;El-Brolosy et al., 2019;Ma et al., 2019).These findings came from studies dissecting the phenotypic discrepancies observed between morpholino-antisense oligomer (MO) knockdowns and genetic mutants, originally thought of as MO-induced toxicity or off-target effects (Kok et al., 2015;Rossi et al., 2015).Interestingly, while both TA and GCR were reported as activated by premature termination codon (PTC) mutations, mRNA degradation is important to trigger TA (El-Brolosy et al., 2019).Most PTC-containing transcripts are degraded through nonsense-mediated mRNA decay (NMD), providing an efficient mechanism by which aberrant and toxic protein synthesis in cells are removed.Interestingly, NMD not only serves to scavenge nonfunctional mRNA but also fine-tunes physiologically normal mRNAs (Mendell et al., 2004;Wittmann et al., 2006;McIlwain et al., 2010;Hurt et al., 2013;Kurosaki et al., 2019;Yi et al., 2021).Thus, dysfunction in the NMD proteins is related to pathological conditions including neurodegeneration, neurodevelopmental disorders, and cancer (Giorgi et al., 2007;Bruno et al., 2011;Colak et al., 2013;Lou et al., 2014).
The NMD core protein complex consists of the up-frameshift proteins Upf1, Upf2, and Upf3 (Wang et al., 2001;Lejeune and Maquat, 2005;Amrani et al., 2006).In eukaryotes, there are two Upf3 paralogs, Upf3a and Upf3b, with both overlapping and distinct functions (Bushman et al., 2015;Shum et al., 2016).The Upfs are conserved throughout eukaryotes and serve to detect premature translation events upstream of the normal termination site.
Although not fully understood, the classic complex consists of Upf1, the main degradation factor that binds randomly to mRNA and becomes activated by suppressor of morphogenesis in genitalia-1 (SMG1); Yamashita, 2013).This process is promoted by the interaction with Upf2 and Upf3.However, data now support that the outcome of NMD depends on the assembling subunits, which in part relies on whether the complexes interact or not with the exon junction complex (EJC; Bushman et al., 2015;Yi et al., 2021).The EJC, deposited at exon-exon boundaries of spliced mRNAs, are removed by the ribosomes during the first round of translation.In the event of a PTC, the downstream EJC remains and binds the Upf3b-Upf2 complex by recruiting SMGs to promote Upf1 interaction and RNA decay (Melero et al., 2014;Bufton et al., 2022).Alternatively, Upf1 can bind directly to long 3 ′ -UTRs, without interacting with Upf3b-EJC, which increases the likelihood of NMD (Hurt et al., 2013;Gao and Wilkinson, 2017).
Considering these new findings, we set out to address if mutations in APP induce TA of other APP family members as a mechanism to activate compensation.We first investigated if TA caused the phenotypic difference observed in the partial knockdown of appb using MO (Joshi et al., 2009;Abramsson et al., 2013;Banote et al., 2016) compared with the CRISPR/Cas9-generated genetic appb mutant (shown as appb -/-; Banote et al., 2020).Like mice and humans, zebrafish harbor genes coding for aplp1 and aplp2, while the partial third genome duplication gave rise to two APP orthologs, appa and appb (Nicolas and Hassan, 2014).To confirm that degradation of faulty mRNA is important to trigger TA, we generated the RNA-less appb mutant (shown as appb P-/-) lacking appb.We then investigated if TA within the APP gene family is conserved in human neuronal cells.Finally, to reveal the mechanism behind the TA between App family members, we inhibited or knocked down NMD core proteins and measured expression level of App family members.

Materials and Methods
Animal care and ethics statement.Zebrafish (Danio rerio) were maintained in Aquatic Housing Systems (Aquaneering) on a 14/10 h light/dark cycle at 28.5°C, at the facility of the Institute of Neuroscience and Physiology, University of Gothenburg.NaHCO 3 and coral sand were used to keep the system water at a pH between 7.2 and 7.4 and Instant Ocean salt to sustain the conductivity at 900 μS.Larva were raised in embryo medium (EM; in mM: 1.0 MgSO 4 , 0.15 KH 2 PO 4 , 0.042 Na 2 HPO 4 , 1 CaCl 2 , 0.5 KCl, 15 NaCl, 0.7 NaHCO 3 ) at 28.5°C in a dark incubator.Fish were fed Gemma granular fish food (Skretting) twice a day and live brine shrimp (Artemia) or marine L-type rotifers (Brachionus plicatilis, ZM Systems).Staging of fish embryos was carried out according to hours postfertilization (hpf) or days postfertilization (dpf; Kimmel et al., 1995).For all the experiments, 0.02% tricaine methanesulfonate (MS-222, Sigma Aldrich) was used to anesthetize larva before experiment.All animal procedures were performed in accordance with the University of Gothenburg animal care committee's regulations.All studies involving zebrafish were performed in accordance with local guidelines and approved by regional authorities.The experiments used AB wild-type fish and the previously published appb 26_2 mutants abbreviated as appb −/− (Banote et al., 2020).The appb P−/− (appb P17 ) fish line was generated in this study.Fish of either gender was used in the experiments.
Mutagenesis using the CRISPR/Cas9 system.Generation of RNA-less mutants was performed using the CRISPR/Cas9 system, as previously described (Varshney et al., 2016).Guide RNAs (gRNAs) were designed to delete approximately 500-1,000 base pairs (bp) of the appb gene, depending on the position of gRNAs, including the ATG site.A 1,129 base-pair sequence, located 127 bp downstream and 1,002 bp upstream of the ATG, was used for blasting with CRISPOR online tool (www.crispor.tefor.net) to design gRNAs, including the 5 ′ -UTR region and exon 1 of appb.Twelve different targetspecific DNA oligos were selected based on their GC content, specificity score, and the number of possible off-targets.A modified version of target DNA was ordered with additional GG motif in the beginning of target sequence to facilitate RNA polymerase binding and more efficient in vitro transcription (Extended Data Table 4-1).Guide RNAs were named based on the location of the PAM sequence with respect to the contig used during the research where a low number is upstream of the ATG site and the highest numbers are downstream of the ATG site.gRNAs were synthesized by assembling each target-specific DNA oligomer with a "generic" DNA oligomer (5 ′ -AAAAGCACCGACTCGGTGCCACTTTTTCAA GTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAAC-3 ′ ) coding for the guide RNA.The two oligos were annealed and extended with Pfu Ultra High-Fidelity DNA Polymerase (Agilent) to produce a double-stranded DNA fragment.The resulting product serves as a template for in vitro transcription using T7 Quick High Yield RNA Synthesis Kit (New England Biolabs).Embryos were coinjected with 50 pg gRNA and 300 pg Cas9 protein (Integrated DNA Technologies), into the yolk at the one-cell stage.Injected embryos were screened for gRNA activity using T7 Endonuclease I assay (Mashal et al., 1995) using specific reverse and forward primers to amplify the target site (Extended Data Table 4-2).The confirmed two active gRNAs were coinjected into wild-type zebrafish, and the embryos were raised to adulthood.Each founder (F0) was outcrossed with wild-type zebrafish of AB background.Ten embryos were selected; a 1,212 bp region surrounding the target site was amplified by PCR using a forward primer (5 ′ -TGTCA TGCGTTTTCCCTTCAC-3 ′ ) and a reverse primer (5 ′ -CTTATCCAGCCCTTCCAGTCG-3 ′ ).Gel electrophoresis and Sanger sequencing were performed to identify the founders carrying a deletion between the two gRNAs used.The remaining embryos of the confirmed founders carrying promoter deletions were raised to adulthood (F1).Fin clipping was performed on F1 generation to extract DNA and identify heterozygotic individuals.Identified heterozygotes carrying the mutant allele were outcrossed with wild-type AB zebrafish for two generations and then inbred to generate homozygotes.Sanger sequencing of homozygous mutants and wild-type zebrafish was performed on genomic DNA from fin clips using DNeasy Blood & Tissue Kits (Qiagen).In short, genomic DNA was amplified using the mentioned forward primer and reverse primer.The amplified product was cleaned with FastAP and exonuclease I (Thermo Fisher Scientific) according to the manufacturer's instructions.Sanger sequencing with BigDye Terminator v1.1 Cycle Sequencing Kit (Applied Biosystems) on an ABI3130xl sequencer (SeqGen) revealed a large deletion of 979 bp between the sites of used gRNA s with 7 bp added, resulting in a total deletion of 972 bp between position −918 and +61 if considering A of ATG as +1.
All the experiments in this study have been performed on the F2 and F3 generation.
Capped and uncapped RNA transcription.Since we were not able to clone the appb mutant cDNA from appb −/− larvae due to extensive RNA degradation, we instead introduced the mutation in a previously cloned appb wild-type cDNA using PfuUltra High-Fidelity DNA Polymerase (Agilent) and the QuikChange Primer Design software to design primers for mutagenesis.The following primers were used to introduce the appb −/− mutation; Fwd 5 ′ -TCGGTGGGCTCAGGA GCCTCAGGTGGCCATG-3 ′ and Rev 5 ′ -CTGAGGCTCCTGAGCCCACCGAGTCATCCG-3 ′ .The wild-type appb cDNA in pcDNA3 (Abramsson et al., 2013) vector was amplified, original plasmid removed by DpnI (New England Biolabs) digestion for 3 h, and then transformed into One Shot Top10 Chemically Competent E. coli (Thermo Fisher Scientific).The mutation was verified by Sanger sequencing.The pSYC-97 plasmid (pCS4+-NLS-EGFP-P2A-mCherry-CAAX) was a gift from Seok-Yong Choi (Addgene plasmid #31565) and was used as the control plasmid (Kim et al., 2011).pcDNA3 and pSYC-97 plasmids were linearized using NaeI (Biolabs) and NotI (Roche) and then in vitro transcribed using mMESSA GE mMACHINET7 Ultra Transcription kit (Invitrogen) and the mMESSAGE mMACHINE SP6 promoter kit (Invitrogen), respectively.To transcribe uncapped RNAs, 20 µl reaction containing 1 µg linearized DNA, 1× transcription buffer, 1× DIG-dNTP, 20 U T7 polymerase, and 20 U RNase inhibitor (all from Roche) was made and incubated for 2 h at 37°C.After incubation, RNA was treated with 4U DNase TURBO (Ambion) and incubated for 15 min at 37°C.The reaction was stopped with 0.5 M EDTA.All RNAs were purified using RNA Clean & Concentrator-5 from Zymolab and diluted to 25-100 ng/μl for injection.
Pharmacological treatments.To inhibit NMD, we treated 24 hpf appb −/− mutant larvae with 10 µM NMDi14 (MedChemExpress) or with 0.1% DMSO (Thermo Fisher Scientific), and 24 h later they were collected and snap frozen for RNA extraction.To block translation, we treated 22-24 hpf appb −/− mutants with EM with or without 15 µg/ml cycloheximide (CHX; Sigma Aldrich).After 5 h, embryos were collected and snap frozen for RNA extraction.For each sample, 10 larvae per well were treated in a 24-well plate.The experiments were done in three biologically independent replicates and each round included five technical replicates.
Western blot.Protein was extracted at 3 dpf from wild-type (WT), appbMO translation blocker, appbMO splicing blocker, and appb P−/− mutant larvae (60 larvae per n; n = 3) to confirm the loss of protein in mutants.Larvae were killed, deyolked in cold EM and snap frozen in liquid nitrogen and stored in −80°C before use.Samples were thawed on ice and homogenized in an ice-cold lysis buffer (10 mM Tris-HCl, pH 8.0, 2% sodium deoxycholate, 2% SDS, 1 mM EDTA, 0.5 M NaCl, 15% glycerol) mixed with protease inhibitors cocktail (Roche) using a 23 G syringe.Homogenized samples were incubated for 20 min on ice and sonicated at maximum intensity for 10 min.After sonication, samples were incubated for 30 min on ice and centrifuged at 10,000 × g at 4°C.The supernatant was transferred into a new 1.5 ml Eppendorf tube and protein concentration measured with a BCA Protein Assay Kit (Thermo Fisher Scientific).Proteins were separated on a NuPAGE NOVEX Bis-TRIS pre-cast gel (Thermo Fisher Scientific) and transferred onto a transfer blot-turbo membrane (Bio-Rad).The membrane was incubated in a 5% milk as a blocking solution for 2 h at RT and then immunoblotted with the polyclonal appb-specific antibody (EER15; 1:10,000), an inhouse-generated polyclonal rabbit anti-Appb antibody against N-terminal half of Aβ peptides (Banote et al., 2020), overnight at 4°C.To detect Appb, we then washed the membrane in TBS-Tween 3 × 10 min at RT and incubated with the secondary antibody IRDye 800CW Goat anti-rabbit (1:10,000) IgG (LI-Cor Biotechnology).The membrane was stripped with stripping buffer for 10 min (Thermo Fisher Scientific), washed in TBS-Tween 3 × 10 min, and then blocked for 1 h in 5% milk as the blocking solution.The membrane was then incubated with a loading concentration control mouse anti-GAPDH-HRP conjugated (1:10,000; Novus Biologicals) for 1 h followed by 3 × 10 min washes in TBS-Tween.The signal was developed using the SuperSignal West Dura Extended Duration Substrate Kit (Thermo Fisher Scientific) and imaged using ChemiDoc Imaging Lab program (Bio-Rad).Quantifications were performed by ImageJ Fiji software (NIH).Band intensities were normalized to loading control and shown as relative controls.
RNA extraction from whole zebrafish larvae and quantitative PCR.To show the mRNA level of app family members, RNA was extracted from whole larvae at 24 hpf.Embryos of AB or appb +/+ background were used as controls.Each experiment was performed at least three times with four to five technical replicates (considered as "N") of 10 larvae each.Total RNA was extracted from 10 larvae, using TRI Reagent (Sigma Aldrich).Then, RNA samples were treated with RQ1 1× RNase-free DNase reaction buffer and RQ1 RNase-free DNase (Promega kit).cDNA was synthesized using High-Capacity RNA-to-cDNA Kit (Applied Biosystems) and converted in a single-cycle reaction on a 2,720 Thermal Cycler (Applied Biosystems).Quantitative PCR (qPCR) was performed with inventoried TaqMan Gene Expression Assays with FAM reporter dye (Thermo Fisher Scientific) in TaqMan Fast Advanced Master Mix (Applied Biosystems).The assay was carried out on Micro-Amp 96-well optical microtiter plates on a QuantStudio 3 Real-Time PCR System Software (Applied Biosystems).qPCR results were analyzed with the QuantStudio Design and Analysis software v1.5.2 (Applied Biosystems).Briefly, C T from each sample was normalized with average C T :s of eef1a1l1 and actb1, and then the relative quantity was determined using the ΔΔC T method (Livak and Schmittgen, 2001) with a sample of wild-type embryos (6-72 hpf and adult brain) as the calibrator.TaqMan Gene Expression Assays (Applied Biosystems) were used for the following genes: Amyloid Beta (A4) Precursor Protein A (appa, Dr 03144364_m1 and Dr 03144365_m1), Amyloid Beta (A4) Precursor Protein B (appb, Dr 03080308_m1 and Dr 03080304_m1), Amyloid Beta Precursor Like Protein 1 (aplp1, AJCSWD2), Amyloid Beta Precursor Like Protein 2 (aplp2, Dr 03437773_m1), Eukaryotic Translation Elongation Factor 1 Alpha 1, Like 1 (eef1a1l1, Dr 03432748_m1), and Actin Beta 1 (actb1, Dr 03432610_m1).To confirm the deletion of appb in appb P−/− , SYBR green primers were designed to amplify exon 16-17 in appb and eef1a1 primer pairs used as housekeeping gene (Extended Data Table 4-3).
RNA sequencing.RNA extracted from whole zebrafish larvae was performed as described above.The quantity and quality of isolated RNA was determined using a 4200 TapeStation Automated Electrophoresis System (Agilent Technologies).All samples had an RNA integrity number >9.3.Sequencing libraries were prepared using the Illumina Stranded mRNA prep kit (Illumina), following the manufacturer's instructions.Construction of libraries were prepared using the Illumina Stranded mRNA Prep Guide 1000000124518 v02 (Illumina), using 500 ng of total RNA input.The Novaseq 6000 platform was used, and 150 bp paired-end reads were generated by Clinical Genomics.
Bioinformatic analysis.To control the quality of the sequencing data, a multiqc report, version 1.13, was generated.Refseq and Ensembl reference genomes for D. rerio were collected from the NCBI and Ensembl online databases.Indexing of reference genomes as well as alignment was performed using STAR version 2.7.10b (Dobin et al., 2013).To improve the alignment of the novel splice junction, STAR was run with the two-pass Mode argument (Veeneman et al., 2016).BAM files were indexed by SAMtools (Danecek et al., 2021) version 1.9 whereafter BAM and index files were used for visualization in IGV version 2.16.2.
Sashimi plots were generated using Integrative Genomics Viewer 2.16.Minimum Junction coverage was set between 9 and 30.
Manual assembly of the region spanning appa was performed by searching for contigs matching the exon 11-12 of appa (NM_131564.2) in Refseq using Blastn with search parameters set to the whole-genome shotgun contigs (WGS), D. rerio as the organism and searching for highly similar sequences.Exon 11-12 mapped 100% to the contig JALCZS010004005.1.Our RNAseq data and previously cloned appa cDNA indicated the existence of an additional exon between 11 and 12.The new exon 12 (marked in blue in Extended Data Fig. 7-1) showed a 100% sequence alignment with contig LKPD02013461.1 that was included within JALCZS010004005.1.The contig JALCZS010004005.1 aligned to both FP067437.2and FO704780.1.A fully annotated sequence of the region can be shared if requested.
Trinity (Grabherr et al., 2011) version 2.15.1 was run to create a de novo transcriptome using the reads from the three wild-type samples.Alignment of the WT control reads to the de novo transcriptome was performed by running STAR, leaving out -sjdbGTFfile -twopassMode arguments, as this was done in order to estimate the read content of the de novo assembly.The manually created sequence was aligned to the de novo assembly using BLAT version 35.Furthermore, BLAT was run to investigate alignment between an Ensembl reference transcriptome (https://ftp.ensembl.org/pub/release-111/fasta/danio_rerio/cdna/) and the de novo transcriptome generated by trinity.
Cell culture and transfection.Cortical neural progenitor cells (NPCs) and neurons were differentiated from two lines of huma-induced pluripotent stem cells (hiPSCs), Ctrl1 (Sposito et al., 2015) and ChiPSC22 (Cellartis by Takara Bio Europe), using a modified version of the protocol from Shi et al. (2012).The detailed neural differentiation procedure is described before (Bergstrom et al., 2016).NPCs were cultured on human laminin L521 (0.5 µg/cm 2 , BioLamina) until around 35 d after induction, and then further differentiated into neurons for 35 more days on poly-L ornithine (0.01%, Sigma Aldrich) and human laminin L521 (0.5 µg/cm 2 ).All cell cultures were kept in a humidified atmosphere at 5% CO 2 and 37°C.
Illustrations. Figure 1A made by J.C. using Affinity Designer (Serif Europe, Apple).Figures 6A and 10 (illustration of conclusion) were created in Biorender.com.
Statistical analysis.Statistical analysis was performed using GraphPad Prism 9 software (Prism).Continuous data were presented using mean and standard deviation of the mean (±SD).Total number of individual samples is shown as "N" and the number of biological independent replicates as "n." Results were compared statistically using two-tailed Student's t tests unless stated otherwise.Statistical significance was set at *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.0001.

Results
Discrepancy between knockdown and knock-out phenotypes is not due to morpholino off-target effects Functional redundancy between App family members is generally accepted as the underlying mechanism behind the lack of major phenotypes in single-gene-knock-out mice (Shariati and De Strooper, 2013).However, if such redundancy also engages genetic compensation is not yet known.Similar to many other zebrafish mutants (Kok et al., 2015;Rossi et al., 2015;Morgens et al., 2016;Housden et al., 2017), the genetic appb mutants (Banote et al., 2020) show a milder phenotype than the appb knockdowns (Joshi et al., 2009;Abramsson et al., 2013;Banote et al., 2016).To evaluate the phenotypic discrepancy between knockdown and knock-out, we began by addressing the specificity of the appb antisense morpholino (appbMO).As previously reported, knockdown of appb results in defect convergence and extension, curved trunk, and partial or complete loss of Mauthner cells (MC) in zebrafish embryos (Joshi et al., 2009;Banote et al., 2016).We used the previously reported appb −/− , harboring a PTC in exon 2 (Banote et al., 2020), to study the genetic loss of appb.Immunostaining of appb −/− using an antibody against neurofilament (RMO44) showed the presence of two bilaterally positioned MCs like wild-type embryos (Fig. 1A-C).We reasoned that if the loss of MCs in appb knockdown larvae was an off-target effect, then this phenotype should be preserved in the appbMO-injected appb −/− .To this end, we evaluated MC formation in wildtype and appb −/− embryos injected with an appb antisense morpholino (appbMO; Fig. 1D-F).Interestingly, while MCs were absent unilaterally (∼31%) or bilaterally (∼47%) in the appbMO-injected wild-type embryos, two MCs were present in all appbMO-injected appb −/− at 48 h postfertilization (hpf; Fig. 1G).Thus, the lack of MC defects in appb −/− indicates an appbspecific effect of the appbMO and that the appb mutation masks the morpholino phenotype.
Upregulated expression of app family members in genetic appb −/− but not in morpholino-mediated appb knockdown The above results, together with the observed degradation of mRNA transcript in appb −/− zebrafish (Banote et al., 2020), made us ask if the retained MC in appb −/− was the consequence of genetic compensation (Rossi et al., 2015;El-Brolosy et al., 2019;Ma et al., 2019) by other app family members.We therefore analyzed the mRNA levels of all app family members (appa, appb, aplp1, and aplp2) in both appbMO-injected wild-type (Fig. 2A and Extended Data Fig. 2-1A) and appb −/− larvae (Fig. 2D).Downregulation of appb, using either the splice-blocking morpholino or a translation-blocking morpholino, reduced appb mRNA and protein levels (Fig. 2A-C and Extended Data Fig. 2-1A-C) and aplp1 mRNA levels but no change was observed on appa or aplp2 expression levels (Fig. 2A and Extended Data Fig. 2-1A).While appb mRNA levels were reduced in appb −/− larvae, we observed a small but significant increase in the mRNA levels of appa and aplp2 and although not significant, a trend toward upregulation of aplp1 compared with their wild-type siblings (Fig. 2D).This suggests that while appb knockdown only affects aplp1 expression, the genetic loss of function mutation in appb increase the mRNA levels of the more similar appa and aplp2 genes.To evaluate if such compensation persists over time, we analyzed mRNA expression levels in the appb −/− adult brain.While appb expression level was still low, there was no significant change in the other app family members suggesting that TA may be more active during early stages of development (Fig. 2E).Together, these data show that a genetic appb mutation, but not a morpholino-mediated appb knockdown, upregulated transcript levels of appa and aplp2 during early stages of development.

TA depends on appb mRNA degradation and not on protein level
To investigate if the transcriptional response observed in appb −/− is activated by the presence of the PTC-containing appb mRNA, we injected stable, capped appb −/− mRNA into wild-type zebrafish.The use of mutant mRNA prevents translation of a functional protein from the injected mRNA and should thus not change the Appb protein level.Injection of capped appb −/− RNA, carrying a m 7 G(5 ′ )ppp(5 ′ )G at the 5 ′ -end to protect the mRNA from degradation (Furuichi and Shatkin, 1977;Hsu and Stevens, 1993;Sachs, 1993), decreased aplp1 mRNA levels but did not change appa or aplp2 expression compared with the eGFP mRNA control (Fig. 3A).To test the effect of RNA decay, we used in vitro transcribed uncapped mutant appb mRNA, known to rapidly degrade due to 5 ′ -to 3 ′ exonucleases (Mukherjee et al., 2012).Injection of uncapped mutant appb mRNA into wild-type zebrafish upregulated appa and aplp2 and downregulated aplp1 transcript levels at 24 hpf, compared with uncapped eGFP control mRNA (Fig. 3B).Together, this shows that appb mRNA decay, but not the mutation or the Appb protein level, upregulate appa and aplp2 transcription, indicating a similar mechanism as described in TA (El-Brolosy et al., 2019).In contrast, the aplp1 expression suggests a different response mechanism that does not depend on the specific mutation or mRNA degradation alone.
No TA in an RNA-less appb mutant allele Our results indicated that TA of appa and aplp2 depends on mutant appb mRNA degradation.To test this hypothesis, we deleted the upstream 5 ′ -region and transcription starts in appb exon 1 to generate a mutant without mRNA (shown as appb P−/− in this study).To this end, we used the CRISPR/Cas9 system to delete the 5 ′ -upstream region and exon 1 of appb (Fig. 4).Two (gRNA89 and gRNA1064) out of 12 gRNAs (Extended Data Table 4-1) were efficient and used to remove a 972 bp fragment, including most of exon 1 (2 base pairs remaining) and its immediate 5 ′ -upstream region (Fig. 4A).The generated appb P heterozygous mutants (appb P+/− ) were outcrossed at least two generations and then inbreed to produce appb P homozygous mutants (appb P−/− ).Western blot and qPCR confirm that the appb P homozygous mutants have very low or no Appb protein and appb mRNA production (Fig. 4B,C; Extended Data Fig. 4-1).Interestingly, the appb P−/− showed no change in appa, aplp1, or aplp2 gene expression (Fig. 4D).However, neurofilament immunostaining showed  that appb P−/− larvae, similarly to appb mutants, exhibit two bilaterally positioned M-cells (Fig. 4E) that are maintained after appbMO injections (Fig. 4F).This shows that M-cell development occurs in the absence of appb and suggests the involvement of mechanisms beyond TA.

Nonsense-mediated mRNA decay
PTC containing mRNA generally activates NMD (Czaplinski et al., 1998;Kurosaki et al., 2019).To test the involvement of NMD in appb −/− mRNA decay, we pharmacologically blocked NMD using NMDi14, disrupting the UPF1-SMG7 interaction, or by blocking translation elongation with CHX, also known as a potent NMD inhibitor (Carter et al., 1995).Our data show a significant increase in appb mRNA levels in both NMDi14-and CHX-treated embryos (Fig. 5A,C).Furthermore, inhibition of NMD increased both appa and aplp2 mRNA levels (Fig. 5B,D), with CHX treatment resulting in a more pronounced increase, also including aplp1 mRNA levels.These results were contrary to the expected.Since decreased appb degradation would reduce TA, we predicted reduced appa and aplp2 mRNA levels.Instead, our results suggest that while NMD is involved in appb −/− mRNA decay, the increased transcript levels indicate that the surveillance system regulates normal mRNA turnover of app family members.To test this, we inhibited NMD and found that CHX, but not NMDi14, led to a general increase of appa, appb, aplp1, and aplp2 mRNA levels in wild-type larvae (Fig. 5E,F).Taken together, these data indicate a major role of NMD in the regulation of the physiological turnover of appa, appb, aplp1, and aplp2 mRNAs.
To determine the role of NMD in TA and normal mRNA turnover, we downregulated upf1, upf2, and upf3s in both wildtype and appb −/− larvae and analyzed mRNA levels of all app family members (Fig. 7).Downregulation of upf1 and upf2 increased mRNA levels of appa and aplp2 in wild-type and appb −/− larvae compared with the uninjected controls (Fig. 7A,B).Furthermore, NMD inhibition in appb −/− significantly increased expression of appa and aplp2 compared with the injected controls with a similar change observed in appa expression after upf2 inhibition.Thus, there is a gradual increase in appa and aplp2 mRNA levels where downregulation of upf1 or upf2 in wild types led to a similar upregulation as detected in appb −/− , with even more elevated levels observed in injected appb −/− (Fig. 7A,B).In contrast, the mRNA levels of aplp1 increased with upf2 inhibition in both wild types and appb −/− , while the aplp1 increase was only observed in wild types (not in appb −/− ) with a upf1 knockdown.It is however likely that there is a general increase in aplp1 but that the sample variation obscured this effect (Fig. 7).Together, our data suggest that upregulation of appa and aplp2, when knocking down upf1 and upf2, is due to both increased mRNA stability and TA due to the remaining appb −/− mRNA decay, since knockdown of upf1 and upf2 did not rescue appb mRNA decay in appb −/− .Contrary, knockdown of upf3a/upf3b only changed the mRNA levels in appb −/− indicating that Upf3s mainly are active in the presence of a PTC (Fig. 7C).Knockdown of upf3a/upf3b increased appb mRNA levels in appb −/− and in addition to appa and aplp2, also increased aplp1 mRNA levels compared with uninjected wild-type controls.These results indicate that while Upf3s mediate degradation of mutant appb mRNA, they are not involved in regulation of the normal mRNA turnover.
The surveillance mechanism of the NMD pathway is preventing faulty mRNAs from being translated.An alternative explanation to the increased mRNA levels observed in wild types upon NMD inhibition would be the presence of alternative splice variants including poisonous exons or other faulty splice events.To determine if such events were present, we analyzed transcript variants present in wild-type and Upf1 knockdowns by RNA sequencing.Sashimi plots of mRNA sequencing reads showed no alternative splicing of appa, appb, aplp1, and aplp2 in upf1 knockdown embryos compared with wild-type controls (Fig. 8A-D).Interestingly exon 11-12 in appa lacked splice arches (marked with red box in Fig. 8A) suggesting that no RNA reads spanned these exons.This region is differently annotated in Refseq and Ensembl, and by manual curation, we found that insertion of the contig LKPD02013461.1 between exons 11 and 12 gave a better sequence alignment that matched with de novo assembled RNAseq data (data not shown; illustrated in Extended Data Fig. 8-1).Furthermore, based on our data, evaluation of percent spliced-in (PSI) values did show a significantly increased number of alternative splicing of other genes in the Upf1 knockdowns samples; however none of the app family members were included among those (Extended Data Table 8-1).These results suggest that the increased transcripts level observed after NMD inhibition corresponds to normal protein coding mRNA transcripts.

TA in human neuronal progenitor cells but not in differentiated human neuronal cells in vitro
To address if the compensation found in zebrafish could be translated to human neurons, we used human neuronal progenitor cells (hNPCs) and neurons differentiated from hiPSCs.Transfection of both cell types with uncapped human APP (hAPP) mRNA resulted in increased APP levels (Fig. 9A,C).However, upregulation of APLP2 was only observed in multipotent hNPCs (Fig. 9B) and not in terminally differentiated neurons (Fig. 9D).No change was observed in APLP1 expression (Fig. 9B,D).This suggests that the TA response is lower in terminally differentiated neurons compared with neuronal progenitors.

Discussion
Genetic approaches aiming to understand the function of genes or mutations involved in disease frequently confront the absence of phenotypes.In zebrafish, the mismatch commonly observed between morpholino knockdown and genetic mutations further confound the question on how the techniques used to introduce genetic modifications affect the phenotypic outcome (Kok et al., 2015).Thus, understanding the underlying mechanisms would not only facilitate genetic approaches for functional gene studies but could also serve in the search for disease-modifying therapies.The partly unclear physiological functions of APP and the APLPs have been attributed to redundancy between family members (reviewed in Muller et al., 2017).Although APP and APLP2 functionally can substitute for each other, the question regarding if and how the loss of one gene activates the expression of other homologous genes remains.Here we show that mutations in the appb gene in zebrafish induce TA of other family members through mRNA degradation.Furthermore, our data shows that the NMD surveillance pathway regulate the physiological transcript levels of all app family members (Fig. 10).The involvement of NMD in regulation of physiological transcript levels has gained increasing interest and is now known to be an important part of biological processes such as spermatogenesis, cell stress, development, and neurogenesis (reviewed in Nasif et al., 2018).Thus, impairments in NMD proteins are associated with intellectual disability but also seem to be linked to neurodegenerative disorders including amyotrophic lateral sclerosis (ALS; Z.Sun et al., 2011;Jackson et al., 2015).Here we found that the expression of app family genes is regulated by the NMD pathway, and our data suggest Upf1 and Upf2 to be central factors in both wild type and mutants.Contrary, Upf3 only seems to be active in PTC-carrying mutants, which is in accordance with previous reports suggesting that Upf3 predominantly triggers NMD as an EJC in PTC-carrying mRNAs (Nguyen et al., 2012(Nguyen et al., , 2014;;Bushman et al., 2015;Yi et al., 2021).In contrast to a previous report showing changed pre-mRNA splicing reported in a genetic upf1 mutant (Lawir et al., 2020), our data support a role of the NMD system in regulating the level of normal protein coding mRNA transcripts.Thus, we propose here that the NMD pathway act to fine-tune the physiological expression level of members of the App family.
Through its function in scavenging faulty mRNAs, the NMD system is also part of activating TA, either through mRNA degradation via Upf1, Upf2, and Upf3b or through the GCR-activating complex in which Upf3a and Wdr5 act to upregulated homologous genes without mRNA decay (Ma et al., 2019).Our data suggest that activation of TA by lesions in an app-gene does not depend on protein level but mRNA degradation since injection of a mutant transcript in wild types but not the RNA-less mutants changed transcription levels of other homologs.However, it is important to note that all   lesions do not induce the same response.Here, downregulation of appb using either the PTC introducing splice-blocking appb morpholino (Abramsson et al., 2013) or the translation-blocking morpholino both decreased appb mRNA and protein levels but did not change transcript levels of appa and aplp2 compared with controls.These findings are in accordance with a previous study by El-Brolosy et al., suggesting that some lesions result in strong NMD while others do not, mainly due differences in mRNA decay (El-Brolosy et al., 2019).Similarly, here the difference observed in TA could be due to the stronger appb mRNA decay in the genetic appb mutant compared with the splice blocking morpholino.The remaining mRNA decay could also explain why TA was still detected in mutants with blocked NMD.The mechanisms by which Upf1 mediates mRNA decay depends on the interacting proteins available, which is determined by the cells' proliferative and differentiation state (Z.Sun et al., 2011;L. Sun et al., 2023).Our data indicate that Upf1-SMG7 interaction is present in PTC-containing mutants but that the interaction is not required for physiological app family expression regulation.The moderate increase in PTC-containing appb mRNA level after both genetic and pharmacological inhibitions of the NMD pathway indicate that knockdown of NMD is not enough to fully block appb mRNA decay.Furthermore, the regulatory function of NMD on normal transcript levels complicated our analysis, since inhibition of the NMD system led to elevated transcript levels in the absence of mutations and even higher levels in the presence of a PTC-carrying mutation.Taken together, the increased transcript level in NMD-inhibited appb −/− is therefore likely an accumulated effect composed of an increase in normal transcript and TA driven by the remaining appb mRNA decay.
Furthermore, we noted that the app family genes responded slightly different to the appb −/− mutation.While appb mRNA degradation increased appa and aplp2 expression, the often-inverse response of aplp1 indicates a potentially different regulatory mechanism.Previous reports have suggested that TA is most active between genes with a high degree of conservation.The nucleotide sequence of aplp1 is more distantly related with appb compared with appa and aplp2 which may be one reason for the observed response.We also note that downregulation of aplp1 correlates with the presence of wildtype appb mRNA while upregulation is observed in the absence of wild-type appb mRNA.It is therefore plausible that the aplp1 mRNA level is linked to Appb protein levels.Finally, while we have shown TA between app family members, the lack of Mauthner cell phenotype in the mRNA-less mutants argues for additional mechanisms, used to sense loss of App family members.While the mRNA defect introduced by the splice-blocking morpholino is present in both wild types and appb −/− , the influence of a faulty mRNA is absent in mRNA-less mutants.Different mutations have varying effect on the transcriptome, but the extent to which such variations affect cell fate and differentiation remains to be analyzed.This furthermore shed light on the complex regulation of this protein family.
Although the presence of TA in early development is clear, our data indicate that this mechanism may be restricted to nascent neurons or neuronal progenitors as we could not detect any changes in transcript level in adult zebrafish brain or in terminally differentiated human neuronal cells cultured in vitro.This is consistent with results indicating that part of the NMD proteins are downregulated during differentiation of neuronal stem cells (Alrahbeni et al., 2015) and that the activity of the NMD machinery differs between cell types (Zetoune et al., 2008).In addition, it is possible that a strong TA activation in less differentiated cells is overshadowed by the nonresponding cells.That would also explain why analysis of App, Aplp1, and Aplp2 single mutant mice did not support elevated expression of other family members, neither on protein (von Koch et al., 1997) nor on mRNA level (Aydin et al., 2011).Nevertheless, it is likely that TA may complicate functional studies of APP and APLPs when using genetic mutations that enhance mRNA decay.
The involvement of TA in disease is a challenging question.In humans, APP dosage is critical in the pathogenesis of Alzheimer's disease (AD) as individuals carrying three copies of APP due to a partial or complete duplication of chromosome 21 (Down syndrome) relentlessly develop AD (Zigman et al., 2002;McCarron et al., 2014).In sporadic AD, reports on increased APP levels also suggest that the expression correlates with amyloid levels and may be crucial for the disease progression (Beyreuther et al., 1993;Preece et al., 2004;Matsui et al., 2007).Thus, based on our study, genetic alterations in APLP2 or genes in the NMD pathway could contribute to changed APP levels in the adult brain.Such events would inevitably promote amyloid formation and AD progression.
In summary, we describe a mechanism by which mutations in one App family member activate TA to induce the expression of other App family members to compensate for its loss and that this mechanism is present both in zebrafish and human NPC.Such compensation is, according to our data, an early event in neurogenesis as TA only was observed in neuronal progenitor cells and not in the adult brain or in vitro cultured human terminally differentiated neuronal cells.We also propose that the NMD pathway is involved in regulating the physiological transcript levels of all App family members.However, the extent to which the NMD-mediated regulation of APP family members contribute to brain homeostasis and disease remains to be investigated.

Figure 4 .
Figure 4.The RNA-less appb P mutation does not induce TA of other App family members.A, Deletion of exon 1 and the 5 ′ upstream sequence of appb was performed using two gRNAs binding −934 bp 5 ′ to and 64 bp 3 ′ of the ATG start, where A is considered as +1.gRNA target sequences are underlined, and PAM sequences are marked in red.B, Western blot analysis of Appb and GAPDH levels in wild type and appb P−/− at 3 dpf.C, Quantification of Western blot data.D, Relative expression levels of appa and appb (and Extended Data Fig. 4-1), aplp1, and aplp2 in appb P−/− (N = 15) and wild-type appbP +/+ siblings (N = 14).Ct values of appb in appb P−/− were above the detection threshold set to 40.E, RMO44 staining in hindbrain of wild-type and appb P−/− mutants at 48 hpf.F, Quantification of MC number in appbMO injected and noninjected wild type (WT) and appb P−/− ."N" indicates number of brains.MC, Mauthner cell.Scale bar, 50 μm.B, D, n = 3 biologically independent samples.F, n = 2-3 biologically independent samples.Wild-type expression levels were set at 1. Data shown as mean + SD.Student's two-tailed t test was used to calculate p values.***p < 0.005 and ****p < 0.0001.ns, nonsignificant.Additional data relating to these analyses are provided in Extended Data Tables 4-1-4-3.

Figure 8 .
Figure 8. Sashimi plots of app family members in wild-type controls and upf1 morpholino injected larvae at 24 hpf.A-D, Splicing of appa (A), appb (B), aplp1 (C), and aplp2 (D) are shown in wild-type controls (black) and upf1 morpholino injected (green) larvae.Numbers indicate number of RNAseq reads and reference gene is outlined in blue.Red star indicates lack of splice junction between exons in appa.Additional data relating to these analyses are provided in Extended Data Figure 8-1 and Extended DataTable 8-1.

Figure 10 .
Figure 10.NMD pathway.A, B, Illustration of the mediated mRNA regulation of app family members under physiological conditions (A) and in appb mutants (B).