My HCP ELISA Says 40 ppm and My LC-MS Says 180 ppm. Which Is Right?
Probably both of them. The two assays do not measure the same quantity, so there is no reason to expect the same answer, and “which is right” is the wrong question. A host cell protein (HCP) ELISA measures total immunoreactivity: the summed binding signal when everything in your sample meets a polyclonal anti-HCP reagent, read against a curve built from a single lysate standard. An LC-MS HCP assay measures mass: it sums the inferred amounts of the individual proteins it identified from their peptides, above a defined threshold, within the sequence database it was allowed to search. A 40 ppm ELISA and a 180 ppm LC-MS result can both be correct, and the gap between them is data, not error. This page explains the five mechanisms behind the gap, gives a protocol for diagnosing which is operating in your case, and answers the regulatory question: which number goes in the filing.
Key facts
- HCP ELISA and LC-MS measure different analytes, so disagreement is not by itself a failure of either assay. USP General Chapter <1132.1> says so directly in Section 7.3: the two are not expected to give the same result, because the methods measure by different mechanisms. The same section cautions specifically against summing individual HCP amounts from LC-MS/MS and comparing that total to a total HCP ELISA value, and Section 7.4 calls the two complementary: each can detect HCPs the other misses, both suit process trending, and agreement between them raises confidence in purity.
- An HCP ELISA reports signal only from HCPs the polyclonal antibody recognizes. USP General Chapter <1132> states the value “can give greater weight to HCPs for which high-affinity antibodies are present in the reagent(s)”.
- LC-MS totals include only proteins identified above threshold and present in the FASTA database searched.
- Parts per million (ppm) in HCP analysis means nanograms of HCP per milligram of product protein (ng/mg).
- USP General Chapter <1132.1>, on LC-MS HCP measurement, was published in USP-NF 2025 Issue 1 on 1 November 2024 with an official date of 1 May 2025. Errata published 31 January 2025 are effective from the same date.
- ICH Q6B (adopted 10 March 1999) names a sensitive immunoassay as the general approach; it does not displace orthogonal methods. USP <1132.1> concludes the same way: a properly designed ELISA remains the workhorse for process development and often for drug substance release, with LC-MS/MS as the orthogonal approach that enables better risk assessment.
What each method actually measures
An HCP ELISA measures total immunoreactivity and an LC-MS HCP assay measures mass. One number is a weighted immunological response summed across an undefined mixture. The other is an arithmetic total over a named list of proteins. The table below sets the two out side by side, attribute by attribute, and every row is a place where the two can part company.
The industry runs both, and the survey numbers say so
Neither method is displacing the other. The BioPhorum Development Group HCP Workstream surveyed its 26 member companies and 18 responded. For identifying and quantifying individual HCPs, 13 of 18 used mass spectrometry and 11 of 18 used an HCP-specific ELISA, with 4 of 18 using enzyme or activity assays and 1 of 18 using gel excision followed by LC-MS [26]. Those counts sum to well over 18, which is the point: most companies run more than one. For total HCP at drug substance release the picture is still dominated by immunoassay, with 13 of 18 using commercial or generic kits, 13 of 18 using a process specific assay and 10 of 18 using platform ELISAs [26].
So the question on this page is not which method to adopt. It is what to do when the two you already run give different answers.
What USP <1132.1> says about comparing the two, and the one comparison it endorses
USP General Chapter <1132.1> says the two results are not expected to be the same, warns against summing individual LC-MS amounts and comparing that total to a total HCP ELISA, and endorses exactly one comparison: a single HCP ELISA against a targeted LC-MS assay for the same protein. All of that sits in Section 7, titled Best Practices for Reporting Data from ELISA Versus LC-MS/MS, which is four subsections devoted to the question at the top of this page. Almost nobody cites it.
Do not expect agreement. Section 7.3 states plainly that ELISA and LC-MS/MS results are not expected to be the same, because the two have “different mechanisms of measurement” (USP <1132.1>, Section 7.3) [25]. That is the anchor. Disagreement is the predicted outcome of measuring two different quantities, not a signal that one assay has failed.
Do not sum LC-MS values and compare the total to an ELISA. This is the specific practice the chapter warns against, and it gives two reasons.
- The ELISA number is not a plain mass total. It weights HCPs by the affinity of the antibodies present in the reagent. It can underreport an abundant HCP through a lack of dilutional linearity. It can miss HCPs that are not immunogenic in the host species used to raise the reagent. And it can overestimate when a single highly immunogenic HCP dominates the response, the effect usually called the jackpot effect.
- The LC-MS number is not a fixed property of the sample either. Which HCPs appear on the list, and what amount is reported for each, depend on how the sample was prepared, how sensitive the instrument is, and which quantitation approach was used.
Two numbers, each conditioned on its own method in a different way. Dividing one by the other gives a ratio with no defined meaning.
There is one comparison the chapter does endorse. Where a single HCP is being monitored, it says comparison is appropriate between a single HCP ELISA using a relevant standard and a targeted LC-MS measurement by multiple reaction monitoring (MRM) or parallel reaction monitoring (PRM) with heavy labeled surrogate peptides, and that the correlation between the two can be good. That is a practical instruction, and it is the one to follow when a named high-risk HCP matters: build a protein-specific ELISA and a targeted assay for the same protein, and compare those. Do not compare a targeted number to a total.
Section 7.4 is titled Complementarity and it is the reason to run both. Each method may detect HCPs the other misses; both are suitable for trending a process over time; and where the two do agree, that agreement is worth something, because it is evidence from two independent measurement principles.
Why the two numbers diverge: five mechanisms
Five mechanisms pull the two totals apart, usually more than one at a time. They are ordered by how often they dominate, and between them they are the machinery behind the chapter’s two reasons above.
1. Antibody coverage
HCP antibody coverage is the proportion of the HCP population in your sample that the anti-HCP polyclonal reagent actually binds. It is the largest source of divergence, and the reason an ELISA number most often looks low next to an LC-MS number.
The polyclonal reagent was raised by immunizing an animal with a lysate from a null cell line, a production cell line carrying no product-coding gene. The animal responds to what it is shown, in proportion to how immunogenic each protein is. Proteins scarce in the immunogen, or poorly immunogenic, produce little or no antibody. USP General Chapter <1132> is explicit: the value “can give greater weight to HCPs for which high-affinity antibodies are present in the reagent(s)”. A protein with no antibody against it contributes nothing, however much is present.
Coverage is measured two ways. The long-established method is the 2D western blot: HCPs are separated by isoelectric point and molecular weight, blotted, probed with the anti-HCP reagent, and immunostained spots expressed as a percentage of total spots. A commonly cited expectation is that more than 50% of total HCP should be reactive, spread across the gel rather than clustered [10]. The second is immunocapture coupled to mass spectrometry, which avoids denaturing the antigen. Pilely and colleagues described ELISA-MS, which immobilizes the antibody in the ELISA plate, captures HCPs natively, then identifies them by LC-MS/MS [4]. Its output is not a percentage but “a list of individual HCPs covered by each HCP antibody” [4]. Waldera-Lupa and colleagues published an equivalent immunoaffinity approach (qIAC-MS) naming both detected and missed HCPs [5].
Scoring the gel is its own piece of work, and it is what SpotMap 2D is for: TotalLab’s HCP antibody coverage software for 2D gel electrophoresis, which matches the immunostained image against the total protein image and reports coverage per spot rather than as a single headline figure.
What matters for reconciliation is that coverage can be reported per protein, not only as a headline percentage. If your LC-MS run named the proteins making up most of the 180 ppm, coverage data tells you whether the ELISA could have seen them. USP <1132> cautions that “numerical coverage comparisons should be used with caution because of the many method variables”, so treat a percentage as a qualitative indicator, not a correction factor [9].
A 2D western blot coverage number is not a clean measure of what the antibody can bind, and the strongest evidence for that comes from a group with no commercial stake in the alternative. Seisenberger and colleagues at Roche used affinity-based mass spectrometry and indirect ELISA as orthogonal checks on conventional 2D western blots and traced the apparent detection gaps to two causes: “(i) low amounts of proteins or antibodies being unable to overcome the detection limit and (ii) western blot artifacts due to the loss of conformational epitopes through protein denaturation hindering HCP-antibody recognition” [17]. The explanation usually offered, that no antibody exists against certain low molecular weight HCPs, “seems to play only a minor role” in their data, and they conclude that “CHO-HCP ELISA antibodies are better than qualification studies by 2D-WBs indicate” [17]. If that holds for your reagent, a 2D western blot percentage understates what your ELISA can see, and the coverage term in your reconciliation is smaller than the headline number implies.
There is a considered counterargument, and it matters because it points at a different method. Gillespie and colleagues at Merck treat the denaturing conditions as the feature, not the bug. Writing on capillary western HCP analysis of a Vero cell line vaccine, they note that the conventional pairing of a denaturing 2D western blot with a native ELISA “provides a rather weak link that is currently accepted”, whereas with capillary western “the reagent coverage can be directly linked between the 2D methodology and Simple Western™, as they are both run under denatured and reduced conditions” [18]. On that reading you do not correct for denaturation, you match it on both sides. Pearson and colleagues describe related capillary western HCP work [19]. Both positions can be acted on: Seisenberger tells you not to over-interpret a low 2D western blot percentage, Gillespie tells you that if you want the coverage number to predict the assay, run both under the same conditions.
An HCP that the polyclonal antibody was never raised against
contributes nothing to the ELISA result, whatever its true
concentration.
Coverage percentage is method-dependent, and the field has agreed no threshold
There is no pharmacopeial coverage target. USP <1132> defines the term, “Coverage: Describes the assessment of how completely a population of polyclonal antibodies recognize the population of HCPs”, and stops there. It names no percentage. The numbers in circulation, most often 50% and 70%, are industry practice, not compendial requirements, and they came from different methods that are not interchangeable.
Cygnus Technologies, the largest supplier of HCP ELISA reagents, states in its published FAQ that “In our experience, a well generated and affinity purified antibody will react to more than 70% of individual HCPs as demonstrated by traditional 2D Western blot correlated to silver stain” [20]. Cygnus also makes the point that the number itself is not portable: Alla Zilberman writes in BioPharm International that “coverage percentage depends on the assessment method and can significantly differ between various methods” [10]. Their preferred approach is antibody affinity extraction (AAE), in which the anti-HCP antibody is immobilized on a column and used to pull out the HCPs it binds, so that the bound and unbound fractions can be compared. Cygnus reports that AAE is “over 100 times higher” in sensitivity than 2D western blot [21].
The best available demonstration of method dependence is Cygnus’s own data, because it measures one antibody three ways. In their AAE whitepaper, the same goat anti-CHO reagent returns three different coverage figures depending only on how coverage was assessed [22].
Fifty-five percent and ninety-two percent, one antibody, no change to the reagent. That single comparison is the most useful thing on this page for anyone trying to interpret a coverage number in a supplier datasheet, and it comes from the supplier with the largest commercial interest in coverage numbers, which is why it is worth citing rather than arguing against.
One methodological observation explains the spread, and it is an observation rather than a criticism. An immunoaffinity method such as AAE defines part of its own denominator with the antibody under test, because the HCP population it resolves on the gel is the population that procedure recovers. A total-protein stain on a 2D gel defines the denominator non-selectively, by everything that stains. Those are different questions about the same reagent, so they give different answers, and neither is wrong. It does mean a coverage percentage is uninterpretable without the method that produced it, which is exactly what Zilberman says [10].
USP makes the same point in one line, and it is worth having on the record because it comes from the pharmacopeia rather than from a competing supplier. Section 4.1 of USP General Chapter <1132.1> describes immunocapturing HCPs with the immobilized polyclonal anti-HCP antibodies from the ELISA, and notes that one use of the technique is assessing the coverage of those reagents by comparing the HCPs identified in the eluate against those identified in the load. In the caveats it attaches to that sample preparation option, the chapter records two limits: the approach can only capture HCPs the polyclonal mixture actually holds antibodies against, and it is “not orthogonal to ELISA” (USP <1132.1>, Section 4.1) [25]. Both follow from the method using the antibody as the capture reagent. That is not an argument against immunoaffinity coverage work, which is informative and which the chapter describes as differing from the traditional 2D SDS-PAGE and 2D western blot route. It is an argument for reading an immunoaffinity coverage figure as a statement about the antibody made using the antibody, and for keeping a non-selective method in the package alongside it.
The Rockland figures bound the top of the range. Their AccuSignal E. coli HCP ELISA datafile reports DIBE coverage of 94% for DH5-alpha, 90% for Origami2 and 94% for Rosetta, describing DIBE as “an enhanced version of 2D gel electrophoresis that demonstrates how well a polyclonal antibody can interact with antigen targets in a lysate (i.e., its coverage)” [23]. A method that routinely returns 90% and a method that routinely returns 55% are not disagreeing about the antibody. They are answering different questions.
For what has actually been accepted, the CASSS WCBP 2020 roundtable notes are the most useful public record. Industry participants recorded that “Generally, >50% coverage is needed”, that in one case an “In-house assay (improved plate washing) was developed to achieve 65% coverage and accepted by the agency”, and that “Some like to see 60% or higher but depends on the cell line. Can justify if between 50-60%” [24]. That is the real expectation: a defensible number with a stated method and a rationale, not a fixed threshold.
2. Standard mismatch
The ELISA standard is a null-cell-line lysate. Your sample is purified drug substance. These two protein populations have almost nothing in common. The calibration curve is built from a lysate representing the upstream HCP population: thousands of proteins in roughly cytosolic proportions. By the end of the purification train, that population has been filtered through Protein A, ion exchange and polishing. What remains is a small, heavily skewed set: proteins that co-purify because they bind the product, bind the resin, or resist the wash. Your sample’s signal is read against a curve from a different mixture entirely, so the conversion assumes equivalent average immunoreactivity, which is false by construction.
Two failure modes follow. The first is the jackpot HCP: a single protein that survives purification and dominates the residual mass. If the antibody covers it well, the ELISA reads high relative to LC-MS. If not, the ELISA reads low while LC-MS reports it accurately. The second is immunization bias: “low molecular weight (Mw) proteins often are less immunogenic than higher Mw proteins, resulting in ELISAs with poor detection of low Mw HCPs” [6]. Small proteases and lipases sit in that blind zone, and they matter most for product quality.
This is also why switching kits changes your number with nothing changing in the process. The standard changed.
3. Dynamic range and the matrix
Both assays detect something present at 1 part in 10^5 to 10^6 against one overwhelming background protein, and both distort under that pressure, in opposite directions. For LC-MS, the product protein floods the analysis. Kiyonami and colleagues describe the requirement as “the wide dynamic range (five-six orders of magnitude) needed to detect HCPs at <10 ppm levels in the presence of the dominant therapeutic proteins” [7]. Peptides from an antibody at 10 mg/mL suppress ionization of peptides from an HCP at 10 ng/mL, and consume duty cycle that would otherwise be spent on the HCP. This biases LC-MS totals downward, which is why enrichment exists. Native digestion, in which the product antibody stays largely intact while HCPs are digested, is what allowed the 0.007 ppm figure above [7]. If your LC-MS number is the higher of the two, dynamic range is not your explanation. If it is unexpectedly low, it often is.
For ELISA, the matrix distorts through dilutional linearity, the requirement that a sample, diluted and corrected back, gives the same concentration at every dilution. Acceptable linearity is corrected values varying no more than ±20% between doubling dilutions [8]. When that fails, the reported number depends on the dilution you happened to run. One cause is the high-dose hook effect: very high HCP concentration saturates the capture and detection antibodies, so more analyte gives less signal. “It is only under conditions of antibody excess that the dose response curve is positively sloped and the assay quantitation accurate” [8]. A hooked ELISA under-reports, sometimes severely. The other causes are HCPs that associate with the product, and buffer components that interfere directly [8].
4. Quantitation basis
Each assay converts signal to mass on a different assumption. An ELISA interpolates optical density on a curve made from a total-protein-quantified lysate, assuming a nanogram of your sample’s HCP gives the same signal as a nanogram of the standard’s. It does not, for all the coverage and composition reasons above.
An LC-MS assay converts peptide ion response to mass by one of the approaches in USP <1132.1>, all three qualified by Chrone and colleagues: Method A quantifies against peptides from the product protein, Method B against spiked intact protein standards added before digestion, Method C against spiked stable isotope labeled peptides [1]. Each assumes ion response per unit mass is comparable between standard and analyte. It is not, and the effect is measurable: different peptides from one protein (clusterin) gave 171, 930 and 422 ng/mL, a 5.4-fold spread, while staying highly reproducible run to run [1]. Reproducible is not accurate. Hi3, which estimates concentration from the three most intense peptides per protein, carries the same assumption.
Method A, Method B and Method C
So your LC-MS total has an accuracy envelope of its own, wider than its precision suggests. Before concluding the two assays disagree, confirm which USP <1132.1> method produced the LC-MS number. Switching methods can move it severalfold.
5. What is summed
This is the least intuitive mechanism and the one that most often closes the gap on paper. An ELISA total includes signal from proteins it never identifies. That is the design. Every immunoreactive species in the well contributes, including species you would never see in a protein list: degraded fragments, aggregates, proteins below any LC-MS detection limit. Pilely and colleagues call it “a semi-quantitative measure of the total HCP content” dependent on “the reactivity and coverage of the polyclonal HCP antibody” [6].
An LC-MS total is a sum over a finite, named list, bounded by two filters. The first is the identification threshold: many HCP workflows require at least two quantifiable peptides per protein, excluding single-peptide identifications [6]. The second is the FASTA search space. A FASTA file is the protein sequence database the search engine matches spectra against. If a protein is not in it, its peptides go unassigned and its mass never enters your total, however much is present. Incomplete host proteomes, missing propeptide forms, unannotated variants and media-derived proteins all fall out this way. A 2026 method paper is framed around illuminating the “dark” host cell proteome that current LC-MS assays do not reach [11].
The ELISA sums more kinds of thing and knows less about them; the LC-MS sums fewer and knows exactly what they are. Neither is the true total HCP mass, and the two error terms point in opposite directions.
A worked reconciliation
The numbers below are illustrative, not measured data, but the arithmetic is what you should do with your own.
Suppose LC-MS reports 180 ppm across 24 proteins, with three accounting for 130 ppm. You check the coverage dataset for your ELISA kit: two of those three are not captured by the anti-HCP reagent. They contribute 95 ppm by mass and zero to the ELISA signal. Of the remaining 85 ppm the antibody can see, the calibration standard over-responds roughly twofold, because it is dominated by highly immunogenic, high molecular weight cytosolic proteins and your residual population is not. Read against that curve, 85 ppm of poorly-responding HCP reports near 40 ppm.
Both instruments did their job. The reconciliation is the output worth having: 95 ppm of your residual burden is invisible to your release assay, and you know which proteins.
If your numbers disagree, check this
Before running the table, fix the two numbers that tell you whether either assay is behaving. USP <1132> states that “Acceptable spike recovery is typically between 70% and 130%”, widening to 50% to 200% for a spike at or near the quantitation limit. Recovery outside that window on a Method B spike means the LC-MS side is not ready to be compared to anything. The chapter also describes action limits set at a fraction of the reject limit, on the order of 50% to 75% of it, which is the level at which a drift in either assay should trigger investigation rather than a failed lot.
What to do when they disagree: a protocol
Run these in order. Most discrepancies resolve by step 5.
- Confirm you are comparing like with like. Same lot, same retain, same units. Both results must be ng HCP per mg product protein, with the product protein concentration from the same assay.
- Identify the LC-MS quantitation basis. Ask which USP <1132.1> method produced the value: A, B or C. A value without a stated method is not comparable to anything.
- Re-run the LC-MS with a spiked intact protein standard. Method B recovery tells you whether digestion and dynamic range are behaving. Check missed cleavage rates too: poor digestion under-reports HCPs while leaving the product signal intact.
- Check ELISA dilutional linearity on the disputed sample. Run a full doubling series. If corrected values vary by more than ±20% across dilutions [8], the ELISA number is dilution-dependent, and the comparison is void until the minimum required dilution is re-established.
- Check antibody coverage against the actual LC-MS hit list. Take the proteins contributing most of the LC-MS mass and determine whether your anti-HCP reagent captures them, by immunocapture-MS or immunoaffinity chromatography coupled to MS. A headline percentage is not enough; you need the per-protein answer.
- Quantify the gap. Sum the mass from hits that are ELISA-blind and compare it to the difference between the two totals. When those roughly agree, you have your explanation.
- Check the FASTA search space. Confirm the database held the full host proteome at the right version, plus media- and process-derived proteins. Re-search the raw data against an expanded database and see whether the total moves.
- Decide whether this is an assay problem or a product problem. If LC-MS surfaced a named high-risk HCP at a level that matters, reconciliation is secondary. Handle the protein.
When the ELISA really is failing and LC-MS is telling you something real
Not every discrepancy is a measurement-basis artifact. In one documented scenario a low ELISA number is false reassurance: a high-risk HCP the antibody cannot see, at a level low enough to vanish in a total and high enough to damage the product.
Polysorbate-degrading enzymes are the clearest case. Li and Richardson report that “some lipases, such as LPLA2, can degrade polysorbate significantly when present at sub-ppm levels, thus going undetected by routine proteomics platforms” [12]. They list eight Chinese hamster ovary (CHO) enzymes implicated in polysorbate 20 and polysorbate 80 degradation, including lipoprotein lipase (LPL) and phospholipase A2 group XV (PLA2G15) [12]. Chiu and colleagues found LPL difficult to remove by conventional purification and knocked it out of the CHO host line to improve polysorbate stability in antibody formulations [3]. Sub-ppm is far below any level that would move a total. A shift from 40 ppm to 40.3 ppm is invisible. The polysorbate degradation is not.
The second case is immunogenicity. CHO phospholipase B-like 2 (PLBL2) co-purified with lebrikizumab, and Fischer and colleagues reported approximately 90% of subjects developed a specific and measurable immune response to PLBL2, though no correlation with safety events could be made and no adjuvant effect on anti-lebrikizumab antibodies was seen [2]. The lesson is not that low-level HCPs are necessarily dangerous. It is that identity matters independently of total, and a total HCP assay cannot supply identity.
In neither case is the right reading “the ELISA is wrong”. It answered the question it was built for, total immunoreactive HCP. That was not the question that mattered.
Which number goes in the regulatory filing?
Both, as orthogonal methods, with the ELISA as the release assay and the LC-MS as supporting characterization, unless you have built a specific case for something else.
An orthogonal method is a second analytical procedure that measures the same attribute by a physically independent principle, so the two are unlikely to fail the same way. USP General Chapter <1132> sets out the framework: “Orthogonal assay methods used to provide additional assurance of purity and the lack of HCPs may be used in any of three modes: ‘detection’, ‘identification’, or ‘quantitation’.” It requires coverage to be established “using 2-D gel Western blots or immunoaffinity fractionation of the total HCP population by immobilizing the anti-HCP antibodies on a column”. Immunocapture-MS is a modern implementation of the second.
The chapter also tells you when to reach for each orthogonal technique, and it puts LC-MS/MS in the late-stage and post-approval column rather than the exploratory one: “The control strategy should also provide for the use of orthogonal methods to provide added assurance of overall purity and the lack of HCPs that may have been missed by the primary immunoassay method. The selection of orthogonal methods depends on the particular product and the stage of product development. For example, during early development, 1-D or 2-D SDS-PAGE with fluorescent or silver staining may be used. At later stages of development and after approval, Western blot analysis or LC-MS/MS may be used selectively on important lots.” >> CONFIRM this passage against the chapter PDF before publication. It is quoted from section 6.2 and I was able to retrieve it once; a second retrieval returned a truncated copy of the PDF that stopped before section 6. << That sentence is worth quoting in a filing, because it is the pharmacopeia rather than a vendor saying that a total immunoassay result on its own is not the whole control strategy. It also frames LC-MS as selective confirmation on important lots, not as a replacement for release testing, which is the position most sponsors are actually in.
ICH Q6B, adopted 10 March 1999, describes the primary assay: “For host cell proteins, a sensitive assay e.g., immunoassay, capable of detecting a wide range of protein impurities is generally utilized.” Note the wording. Immunoassay is offered as an example, and the guideline states purity “is assessed by a combination of analytical procedures”. ICH Q6B neither mandates ELISA nor excludes LC-MS.
USP General Chapter <1132.1>, published in USP-NF 2025 Issue 1 on 1 November 2024 with an official date of 1 May 2025, gives LC-MS HCP quantitation a formal, citable procedural framework. That strengthens an LC-MS package in a filing, but does not by itself make LC-MS a release method, and the chapter’s own conclusions say as much.
Section 8, Discussion and Conclusions, is the closest thing to an official verdict anyone has published, and it is even handed in a way that is useful to quote in a development report. It concludes that a properly designed ELISA remains the workhorse assay for process development and often for drug substance release testing; that LC-MS/MS provides an orthogonal approach which enables a more intelligent risk assessment of residual HCPs; and that LC-MS/MS is technically amenable to quality control release testing and may become the preferred method if the obstacles of cost and data analysis are overcome [25]. Read those three clauses in order and you have the current state of the field and the direction of travel, sourced. The practical filing position follows from them:
- Report the ELISA value as the specification-setting and lot release result, with the kit, coverage data and dilutional linearity qualification.
- Report the LC-MS result as orthogonal characterization, naming the USP <1132.1> method used and the significant contributors.
- Explain the difference rather than hiding it. A reviewer who finds two unexplained numbers will ask; one who finds a mechanistic reconciliation sees a sponsor who understands their impurity profile.
- Where LC-MS found a high-risk HCP, address that protein with a control strategy, not by pointing at the total.
Sponsors do use LC-MS at release in specific circumstances, most often for a named individual HCP with a defined limit rather than a total. Replacing a total HCP ELISA at release remains uncommon, and requires a validation package showing the method detects the HCPs relevant to your process at the levels that matter, plus bridging against historical ELISA data. >> CONFIRM: Steven may want a regulatory read on this paragraph <<
The reconciliation is the deliverable
The common mistake is collapsing the discrepancy into a single number. Decomposed properly, it tells you what neither assay reports alone: how much of your residual HCP burden your release assay cannot see, and which proteins make it up. That requires an LC-MS side reproducible enough to trust batch to batch: a fixed FASTA, a fixed quantitation method, and a record of how every number was produced. TotalLab’s SpotMap MS is LC-MS HCP analysis software that automates that work using data-independent acquisition (DIA), which acquires fragment data for all precursors in each mass window rather than selecting precursors in real time, and is designed to support alignment with USP <1132.1>. We set out the USP <1132.1> software requirements separately, and the wider control strategy this sits inside is covered in HCP analysis in biologics.
Data-independent acquisition (DIA)
USP <1132.1> software requirements
Frequently asked questions
Q: Can LC-MS replace HCP ELISA?
A: Not routinely, not yet, and USP says so. Section 8 of USP General Chapter <1132.1> concludes that a properly designed ELISA remains the workhorse for process development and often for drug substance release, with LC-MS/MS as the orthogonal approach that enables better risk assessment. It adds that LC-MS/MS is technically amenable to QC release testing and may become preferred if the cost and data analysis obstacles are overcome. Most programs run both today.
Q: Why is my LC-MS total lower than my ELISA?
A: Usually one of three things. The product protein is suppressing ionization of HCP peptides, so low-abundance proteins never reach the identification threshold. Digestion is incomplete, so HCPs are under-recovered. Or your FASTA database is missing proteins that are present in the sample, in which case their peptides are never assigned and their mass never enters the total. Spike an intact protein standard before digestion and check recovery first.
Q: Do regulators accept LC-MS for HCP release testing?
A: LC-MS is widely accepted for characterization, identification and orthogonal confirmation, and USP <1132.1> has strengthened its standing considerably. Acceptance for release testing is case by case and depends on the validation package, the product and the specific claim. Targeted LC-MS for a single named high-risk HCP is more readily accepted than a total HCP LC-MS value replacing an ELISA.
Q: Should my ELISA and LC-MS HCP numbers agree?
A: No, and USP <1132.1> states that directly in Section 7.3: the two are not expected to be the same because they have different mechanisms of measurement. The chapter specifically cautions against summing individual LC-MS amounts and comparing that total to a total HCP ELISA value, because the ELISA weights HCPs by antibody affinity while the LC-MS list depends on sample preparation, instrument sensitivity and the quantitation approach. The comparison it does endorse is a single HCP ELISA against a targeted MRM or PRM assay for the same protein.
Q: Does a bigger gap between ELISA and LC-MS mean a worse process?
A: No. The size of the gap reflects how well your anti-HCP reagent matches your residual HCP population, not how clean the process is. What matters is whether the gap is stable. Track the ratio of the two values across batches. A consistent ratio indicates a controlled process; a ratio that suddenly shifts is worth investigating even if both absolute numbers stay in specification.
Q: My ELISA coverage is 65%. Can I just divide my ELISA number by 0.65?
A: No. Coverage is a count of proteins recognized, not a mass-weighted response factor, and USP <1132> warns that numerical coverage comparisons should be used with caution because of method variability. A reagent can cover 65% of proteins by count while missing the single protein that makes up most of the residual mass. Use coverage to identify which proteins are invisible, not to scale the total.
Q: Which USP <1132.1> method gives the most accurate number?
A: Method C, using spiked stable isotope labeled peptides, gives the tightest linearity and the lowest working range, but it only quantifies the proteins you chose peptides for. Method B, using spiked intact protein standards added before digestion, controls for digestion efficiency and covers the whole profile. Method A, relative to the product protein, is simplest but has the highest working range. Choose by purpose and state the method with every value.
Q: How different can two LC-MS numbers for the same sample be?
A: More than most people expect. In a published qualification of the USP <1132.1> methods, three different peptides from a single protein gave concentrations of 171, 930 and 422 ng/mL, a 5.4-fold spread, while each individual measurement was highly reproducible. Precision and accuracy are separate properties. Fix your peptide selection rules and quantitation method, then compare like with like over time.
References
1. Chrone VG, Blaszczyk AJ, Zhang DH, Nielsen SB, Crawford J, Schwämmle V, Højrup P, Kofoed T, Peckham N, Mørtz E. Host cell protein quantitation by LC-MS. Experimental demonstration, qualification, and comparison of methods in USP 1132.1. Journal of Pharmaceutical and Biomedical Analysis, 2025, 265:117051. DOI 10.1016/j.jpba.2025.117051. https://doi.org/10.1016/j.jpba.2025.117051
(Source for: the qualification of Method A, Method B and Method C on the same samples in the quantitation basis mechanism; and the three peptides from clusterin returning 171, 930 and 422 ng/mL, a 5.4-fold spread with high run-to-run reproducibility, used there and in the final FAQ. The Elsevier page returns 403 to automated fetching; the DOI and bibliographic detail were confirmed through Semantic Scholar and the University of Southern Denmark research portal.)
2. Fischer SK, Cheu M, Peng K, Lowe J, Araujo J, Murray E, McClintock D, Matthews J, Siguenza P, Song A. Specific Immune Response to Phospholipase B-Like 2 Protein, a Host Cell Impurity in Lebrikizumab Clinical Material. The AAPS Journal, 2017, 19(1):254-263. https://link.springer.com/article/10.1208/s12248-016-9998-7
(Source for: PLBL2 co-purifying with lebrikizumab, approximately 90% of subjects developing a specific and measurable immune response, and the absence of a demonstrated correlation with safety events or an adjuvant effect on anti-lebrikizumab antibodies.)
3. Chiu J, Valente KN, Levy NE, Min L, Lenhoff AM, Lee KH. Knockout of a difficult-to-remove CHO host cell protein, lipoprotein lipase, for improved polysorbate stability in monoclonal antibody formulations. Biotechnology and Bioengineering, 2017, 114(5):1006-1015. DOI 10.1002/bit.26237. https://doi.org/10.1002/bit.26237
(Source for: lipoprotein lipase (LPL) being difficult to remove by conventional purification, and the knockout of LPL from the CHO host line to improve polysorbate stability.)
4. Pilely K, Nielsen SB, Draborg A, Henriksen ML, Hansen S, Skriver L, Mørtz E, Lund R. A novel approach to evaluate ELISA antibody coverage of host cell proteins, combining ELISA-based immunocapture and mass spectrometry. Biotechnology Progress, 2020, 36(4):e2983. DOI 10.1002/btpr.2983. https://doi.org/10.1002/btpr.2983
(Source for: the ELISA-MS immunocapture coverage method, and the quoted output being a list of individual HCPs covered by each HCP antibody rather than a percentage.)
5. Waldera-Lupa D, Jasper Y, Köhne P, Schwichtenhövel R, Falkenberg H, Flad T, Happersberger P, Reisinger B, Dehghani A, Moussa R, Waerner T. Host cell protein detection gap risk mitigation: quantitative IAC-MS for ELISA antibody reagent coverage determination. mAbs, 2021, 13(1):1955432. DOI 10.1080/19420862.2021.1955432. https://doi.org/10.1080/19420862.2021.1955432
(Source for: the qIAC-MS immunoaffinity coverage approach that names both detected and missed HCPs, cited in the antibody coverage mechanism and in the diagnostic table.)
6. Pilely K, Johansen MR, Lund RR, Kofoed T, Jørgensen TK, Skriver L, Mørtz E. Monitoring process-related impurities in biologics: host cell protein analysis. Analytical and Bioanalytical Chemistry, 2022, 414:747-758. https://link.springer.com/article/10.1007/s00216-021-03648-2
(Source for: the quoted description of the ELISA as a semi-quantitative measure of total HCP content dependent on the reactivity and coverage of the polyclonal antibody; the quoted statement that low molecular weight proteins are often less immunogenic, giving ELISAs poor detection of low Mw HCPs; and the requirement in many workflows for at least two quantifiable peptides per protein.)
7. Kiyonami R, Melani R, Chen Y, De Leon A, Du M. Applying UHPLC-HRAM MS/MS Method to Assess Host Cell Protein Clearance during the Purification Process Development of Therapeutic mAbs. International Journal of Molecular Sciences, 2024, 25(17):9687. https://pmc.ncbi.nlm.nih.gov/articles/PMC11396427/
(Source for: the 0.007 ppm detection figure quoted in Table 1 and in the dynamic range mechanism, and for the quoted requirement of five to six orders of dynamic range to detect HCPs below 10 ppm against dominant therapeutic proteins.)
8. Cygnus Technologies. Establishing Dilution Linearity for Your Samples in an ELISA. Accessed 2026. https://www.cygnustechnologies.com/establishing-dilution-linearity-for-your-samples-in-an-elisa
(Source for: acceptable dilutional linearity as corrected values varying no more than plus or minus 20% between doubling dilutions, used in the dynamic range mechanism, the diagnostic table and step 4 of the protocol; the quoted statement that only under antibody excess is the dose response curve positively sloped and quantitation accurate; and the high-dose hook effect, product-associated HCPs and buffer interference as the causes of non-linearity.)
9. Hu X, Lyu P, Wang D, Li Y, Xu K, Ling F, Dou M, Liang C, Li J. A quality evaluation strategy for residual host cell proteins based on orthogonal analysis. Frontiers in Bioengineering and Biotechnology, 2026. https://www.frontiersin.org/journals/bioengineering-and-biotechnology/articles/10.3389/fbioe.2026.1886078/full
(Source for: the orthogonal-analysis framing of residual HCP evaluation, and for treating a coverage percentage as a qualitative indicator inside that strategy rather than as a correction factor.)
10. Zilberman A. HCP ELISA and HCP Antibody Coverage Analysis Methods. BioPharm International, 5 April 2021. https://www.biopharminternational.com/view/hcp-elisa-and-hcp-antibody-coverage-analysis-methods
(Source for: the commonly cited expectation that more than 50% of total HCP should be reactive and spread across the gel; and the quoted statement that coverage percentage depends on the assessment method and can significantly differ between methods, used in the coverage subsection and in Table 2.)
11. Illuminating the Dark Host Cell Proteome: A host cell protein coverage method for LC-MS impurity assays. Journal of Pharmaceutical and Biomedical Analysis, 2026. https://pubmed.ncbi.nlm.nih.gov/42128475/
(Source for: the “dark” host cell proteome that current LC-MS assays do not reach, cited in the fifth mechanism on what is summed.)
12. Li X, Richardson DD. Analysis of Trace-Level High-Risk HCPs: Proteomics Advances for Preventing Degradation of Polysorbates in Biotherapeutic Formulations. BioProcess International, 30 September 2021. https://www.bioprocessintl.com/product-characterization/analysis-of-trace-level-high-risk-hcps-proteomics-advances-for-preventing-degradation-of-polysorbates-in-biotherapeutic-formulations
(Source for: the quoted statement that some lipases such as LPLA2 degrade polysorbate significantly at sub-ppm levels and go undetected by routine proteomics platforms, and for the list of eight CHO enzymes implicated in polysorbate 20 and polysorbate 80 degradation, including LPL and PLA2G15.)
13. United States Pharmacopeia. General Chapter <1132> Residual Host Cell Protein Measurement in Biopharmaceuticals. USP 39 and NF 34, official 1 May 2016. Free full-text PDF posted by USP. Note that this free copy is the 2016 version and that no Notice of Intent to Revise has been published for <1132> since it became official. https://www.usp.org/sites/default/files/usp/document/our-work/biologics/USPNF810G-GC-1132-2017-01.pdf
(Source for: the quoted statement that the value can give greater weight to HCPs for which high-affinity antibodies are present; the definition of coverage and the absence of any numeric target; the caution on numerical coverage comparisons; the three modes of orthogonal assay use; the requirement to establish coverage by 2-D gel western blot or immunoaffinity fractionation; the section 6.2 passage on selecting orthogonal methods by development stage, which carries a confirmation marker in the body; and spike recovery of 70% to 130%, widening to 50% to 200% near the quantitation limit, plus the action limit fraction, which carries its own marker.)
14. United States Pharmacopeia. General Chapter <1132.1> Residual Host Cell Protein Measurement in Biopharmaceuticals by Liquid Chromatography-Mass Spectrometry. USP-NF Notice of Intent to Revise, 1 November 2024. https://www.uspnf.com/notices/gc-1132-1-nitr-20241101
(Source for: the 1 November 2024 publication date of the chapter in USP-NF 2025 Issue 1.)
15. ECA Academy. USP Announces Corrections to General Chapter <1132.1> on Residual Host Cell Protein Measurement. 2025. https://www.gmp-compliance.org/gmp-news/usp-announces-corrections-to-general-chapter-1132-1-on-residual-host-cell-protein-measurement
(Source for: the errata published 31 January 2025 and effective 1 May 2025, stated in the key facts and in the regulatory section.)
16. International Council for Harmonisation. ICH Harmonised Tripartite Guideline Q6B: Specifications: Test Procedures and Acceptance Criteria for Biotechnological/Biological Products. Adopted 10 March 1999. https://database.ich.org/sites/default/files/Q6B%20Guideline.pdf
(Source for: the adoption date; the quoted section 6.2.1 wording that a sensitive assay, for example an immunoassay, is generally utilized for host cell proteins; and the quoted statement that purity is assessed by a combination of analytical procedures.)
17. Seisenberger C, Graf T, Haindl M, Wegele H, Wiedmann M, Wohlrab S. Questioning coverage values determined by 2D western blots: A critical study on the characterization of anti-HCP ELISA reagents. Biotechnology and Bioengineering, 2021, 118(3):1116-1126. DOI 10.1002/bit.27635, PMID 33241851. https://doi.org/10.1002/bit.27635
(Source for: the two quoted causes of apparent detection gaps in 2D western blots, detection limit and loss of conformational epitopes; the quoted finding that absent antibodies play only a minor role; and the quoted conclusion that CHO-HCP ELISA antibodies are better than 2D western blot qualification indicates.)
18. Gillespie PF, Wang Y, Yin K, Groegler E, Cunningham N, Stiving AQ, Raffaele J, Marusa N, Tubbs CM, Loughney JW, Winters MA, Rustandi RR. Automated, Quantitative Capillary Western Blots to Analyze Host Cell Proteins in COVID-19 Vaccine Produced in Vero Cell Line. Vaccines, 2024, 12(12):1373. DOI 10.3390/vaccines12121373, PMID 39772035. https://doi.org/10.3390/vaccines12121373
(Source for: the capillary western counterargument, including the quoted description of the denaturing 2D western blot to native ELISA pairing as a rather weak link that is currently accepted, and the quoted claim that reagent coverage can be directly linked between the 2D methodology and capillary western because both run under denatured and reduced conditions.)
19. Pearson C, et al. Capillary western analysis of host cell proteins. Journal of Pharmaceutical and Biomedical Analysis, 2023, 233:115420. DOI 10.1016/j.jpba.2023.115420. https://doi.org/10.1016/j.jpba.2023.115420
(Source for: related capillary western HCP work cited alongside Gillespie and colleagues in the antibody coverage mechanism.)
20. Cygnus Technologies. Frequently Asked Questions. Accessed 2026. https://www.cygnustechnologies.com/resources/faqs/
(Source for: the quoted in-house experience that a well generated and affinity purified antibody will react to more than 70% of individual HCPs by traditional 2D western blot correlated to silver stain, used in the coverage subsection and in Table 2.)
21. Cygnus Technologies. Antibody Affinity Extraction (AAE). Accessed 2026. https://www.cygnustechnologies.com/antibody-affinity-extraction-aaetm See also Why You Should Not Rely on Western Blotting for HCP Antibody Coverage, https://www.cygnustechnologies.com/why-you-should-not-rely-on-western-blotting-for-hcp-antibody-coverage
(Source for: the description of antibody affinity extraction as immobilizing the anti-HCP antibody on a column to compare bound and unbound fractions; the quoted claim that AAE is over 100 times higher in sensitivity than 2D western blot; and the conventional more-than-50% criterion row in Table 2.)
22. Cygnus Technologies. Antibody Affinity Extraction (AAE) white paper. https://www.cygnustechnologies.com/media/productattach/c/y/cygnus-aae-whitepaper_f_v3.pdf
(Source for: the three coverage figures for one goat anti-CHO reagent in Table 2, 55% by 2D western blot with 717 of 1293 spots, 73% by AAE with silver stain with 827 proteins in common, and 92% by AAE with 2D-DIGE with 896 of 976 spots, each quoted with its spot counts.)
23. Rockland Immunochemicals. AccuSignal E. coli HCP ELISA Kit Datafile. Accessed 2026. https://www.rockland.com/globalassets/documents/literature/AccuSignal-E.coli-HCP-ELISA-Kit-Datafile.pdf
(Source for: DIBE coverage of 94% for DH5-alpha, 90% for Origami2 and 94% for Rosetta, and the quoted description of DIBE as an enhanced version of 2D gel electrophoresis that demonstrates how well a polyclonal antibody can interact with antigen targets in a lysate.)
24. CASSS. Host Cell Proteins: Reagent Coverage, Identification and Risk Assessment. WCBP 2020 roundtable notes. https://www.casss.org/docs/default-source/wcbp/2020-roundtable-notes/host-cell-proteins-reagent-coverage-identification-and-risk-assessment.pdf
(Source for: the three quoted industry statements on accepted coverage, that generally more than 50% is needed, that an in-house assay achieving 65% was accepted by the agency, and that some reviewers like to see 60% or higher with justification between 50 and 60%.)
25. United States Pharmacopeia. General Chapter <1132.1> Residual Host Cell Protein Measurement in Biopharmaceuticals by Liquid Chromatography-Mass Spectrometry. USP-NF, official 1 May 2025. DOI 10.31003/USPNF_M17756_03_01. Subscription access. https://doi.usp.org/USPNF/USPNF_M17756_02_01.html
(Source for: Section 4.1, the immunoaffinity sample preparation caveats including the quoted phrase that it is not orthogonal to ELISA; Section 7.3, the quoted different mechanisms of measurement, the caution against summing individual LC-MS amounts for comparison with a total ELISA and the two reasons given; the single endorsed comparison of a single HCP ELISA against targeted MRM or PRM with heavy labeled surrogate peptides; Section 7.4 Complementarity; and Section 8, that a properly designed ELISA remains the workhorse, that LC-MS/MS provides an orthogonal approach enabling better risk assessment, and that LC-MS/MS is technically amenable to QC release testing and may become preferred if cost and data analysis obstacles are overcome.)
26. Jones M, Palackal N, Wang F, Gaza-Bulseco G, Hurkmans K, Zhao Y, Chitikila C, Clavier S, Liu S, Menesale E, Schonenbach NS, Sharma S, Valax P, Waerner T, Zhang L, Connolly T. “High-risk” host cell proteins (HCPs): A multi-company collaborative view. Biotechnology and Bioengineering, 2021, 118(8):2870-2885. DOI 10.1002/bit.27808, PMID 33930190. https://doi.org/10.1002/bit.27808
(Source for: the BioPhorum Development Group HCP Workstream survey of 26 member companies with 18 responses; the individual HCP split of 13 of 18 using mass spectrometry, 11 of 18 an HCP-specific ELISA, 4 of 18 enzyme or activity assays and 1 of 18 gel excision with LC-MS; and the total HCP release picture of 13 of 18 commercial or generic kits, 13 of 18 process specific assays and 10 of 18 platform ELISAs.)
Run the comparison in your own lab
If you send samples out for the LC-MS half of this comparison, you are waiting weeks for the data that explains your ELISA. SpotMap MS is licensed software your lab runs itself, so the reconciliation happens on your timeline, against your own FASTA and your own coverage data. The case to bring HCP analysis in-house is set out separately.