Justice Marshall
delivered the opinion of the Court.
This case arose from an animal welfare dispute. At issue is the fate of certain monkeys used for medical experiments funded by the Federal Government. The case comes before us, however, on a narrow jurisdictional question: whether a suit filed in state court challenging the treatment of these monkeys was properly removed to the federal court by respondent National Institutes of Health (NIH), one of the defendants. We hold that removal was improper and that the case should be remanded to state court.
1 — I
Petitioners, who are organizations and individuals seeking the humane treatment of animals, filed this suit in Louisiana civil district court; the monkeys are housed at a primate research center in that State. Three defendants were named
and are respondents here. Respondent Institutes for Behavior Resources (IBR) is a private entity that owns the monkeys.
Respondent NIH now maintains custody of the monkeys, with IBR’s consent. Respondent Administrators of the Tulane Educational Fund (Tulane) is the governing body for the primate research center that, in 1986, entered into an agreement with NIH to care for the monkeys. The suit sought to enjoin further experimentation on the monkeys and to obtain custody over them. Petitioners based their claim for this relief upon Louisiana law, including provisions that (1) impose criminal sanctions for cruelty to animals, La. Rev. Stat. Ann. §14:102.1 (1986 and Supp. 1991); (2) permit officers of humane societies to remove, to a “stable,” animals being subjected to cruelty or that are “bruised, wounded, crippled, abrased, sick, or diseased,” La. Rev. Stat. Ann. §3:2431 (1987); (3) authorize tort damages for “[ejvery act whatever of man that causes damage to another,” La. Civ. Code Ann., Art. 2315 (1979 and Supp. 1991); and (4) direct courts to “proceed according to equity” in situations not covered by “legislation or custom,” La. Civ. Code Ann., Art. 4 (Supp. 1991). See App. to Pet. for Cert. A-35 to A-37.
Shortly after the suit was filed, NIH removed the case to federal court pursuant to 28 U. S. C. § 1442(a)(1), which authorizes removal of state suits by certain federal defendants. The federal District Court then granted a temporary re
straining order barring NIH from carrying out its announced plan to euthanize three of the remaining monkeys and, in the process, to complete some of the medical research by performing surgical procedures. The court extended this order beyond its 10-day limit, see Fed. Rule Civ. Proc. 65(b), and NIH accordingly appealed the court's action under 28 U. S. C. § 1292(a)(1), which permits appellate review of preliminary injunctions.
On appeal, NIH argued, inter alia, that petitioners were not entitled to the injunction because they lacked standing to seek protection of the monkeys. Petitioners, in turn, argued that the District Court had no juris~1iction over the case because 28 U. S. C. § 1442(a)(1) permits only federal officials-not federal agencies such as NIH-to remove cases in which they are named as defendants. The Court of Appeals for the Fifth Circuit agreed with NIH that petitioners could not satisfy the requirements under Article III of the United States Constitution for standing. It also held that federal agencies have the power to remove cases under § 1442(a)(1). Accordingly, the Court of Appeals vacated the injunction and dismissed the case. See 895 F. 2d 1056 (CA5 1990). We granted certiorari to resolve a conflict between the Courts of Appeals for the Fifth and Third Circuits on the question whether § 1442(a)(1) permits removal by fedei'al agencies.
498 U. 5. 980 (1990). We conclude that it does not.
II
We confront at the outset an objection raised by NIH to our jurisdiction over the removal question. NIH argues that, because the Court of Appeals found that petitioners lack Article III standing to seek protection of the monkeys, petitioners also lack standing even to contest the removal of
their suit. We beli,eve NIH misconceives both standing doctrine and the scope of the lower court’s standing ruling.
Standing does not refer simply to a party’s capacity to appear in court. Rather, standing is gauged by the specific common-law, statutory or constitutional claims that a party presents. “Typically, . . . the standing inquiry requires careful judicial examination of a complaint’s allegations to ascertain whether the particular plaintiff is entitled to an adjudication
of the particular claims asserted.
”
Allen
v.
Wright,
468 U. S. 737, 752 (1984) (emphasis added). See also Fletcher, The Structure of Standing, 98 Yale L. J. 221, 229 (1988) (standing “should be seen as a question of substantive law, answerable by reference to the statutory and constitutional provision whose protection is invoked”).
It is well established that a party may challenge a violation of federal statute in federal court if it has suffered “injury that fairly can be traced to the challenged action of the defendant,”
Simon
v.
Eastern Kentucky Welfare Rights Org.,
426 U. S. 26, 41 (1976), and that is “likely to be redressed by the requested relief.”
Allen
v.
Wright, supra,
at 751. In the case now before us, petitioners challenge NIH’s conduct as a violation of § 1442(a)(1). Petitioners’ injury is clear, for they have lost the right to sue in Louisiana court — the forum of their choice. This injury “fairly can be traced to the challenged action of defendants,” since it directly results from NIH’s removal of the case. And the injury is “likely to be redressed” if petitioners prevail on their claim because, if removal is found to have been improper under § 1442(a)(1), the federal courts will lose subject matter jurisdiction and the “case shall be remanded.” 28 U. S. C. § 1447(c); see
infra,
at 87-89. Therefore, petitioners clearly have standing to challenge the removal.
Nothing in the Court of Appeals’ decision undermines this conclusion. The court below found that petitioners did not have standing to protest “disruption of their personal relationships with the monkeys,” 895 F. 2d, at 1059, to claim
“harm to their ‘aesthetic, conservational and environmental interests,’”
id.,
at 1060, or to act as advocates for the monkeys’ interests,
id.,
at 1061. But at no point did the Court of Appeals suggest that petitioners’ lack of standing to bring these claims interfered with their right to challenge removal. Indeed, it was only
after
the court rejected petitioners’ standing to protect the monkeys
that it considered the question whether NIH’s removal was proper.
Id.,
at 1061-1062. NIH argues that, were we also to consider the propriety of removal, “the Court would be resolving the removal question in a context in which the court below specifically found the injury in fact necessary to [the concrete] adverseness [required for standing] to be lacking.” Brief for Respondent NIH 7, n. 4. We disagree. The “adverseness” necessary to resolving the
removal
question is supplied not by petitioners’ claims for the monkeys’ protection but rather by petitioners’ desire to prosecute their claims in state court.
III
A
Section 1442(a)(1) permits a defendant in a civil suit filed in state court to remove the action to a federal district court if the defendant is “[a]ny officer of the United States or any agency thereof, or person acting under him, [in a suit challenging] any act under color of such office . . . 28 U. S. C. § 1442(a)(1).
The question before us is whether this provision permits agencies to remove. “‘[T]he starting point in every case involving construction of a statute is the language itself.’”
Watt
v.
Alaska,
451 U. S. 259, 265 (1981) (citation omitted). We have little trouble concluding that the statutory language excludes agencies from the removal power. To be sure, the first clause in § 1442(a)(1) contains the words “or anj agency thereof.” IBR argues that those words designate one of two grammatical subjects in § 1442(a)(l)’s opening clause (namely, agencies) and that the clause’s other subject is “[a]ny officer of the United States.” But such a reading is plausible only if this first clause is examined in isolation from the rest of § 1442(a)(1). “We continue to recognize that context is important in the quest for [a] word’s meaning,”
United States
v.
Bishop,
412 U. S. 346, 356 (1973), and that “[statutory construction ... is a holistic endeavor.”
United Savings Assn, of Texas
v.
Timbers of Inwood Forest Associates, Ltd.,
484 U. S. 365, 371 (1988). We find that, when construed in the relevant context, the first clause of
§ 1442(a)(1) grants removal power to only one grammatical subject, “[a]ny officer,” which is then modified by a compound prepositional phrase: “of the United States or [of] any agency thereof.”
Several features of § 1442(a)(1)’s grammar and language support this reading. The first is the statute’s punctuation. Cf.
United States
v.
Ron Pair Enterprises, Inc.,
489 U. S. 235, 241 (1989) (statute’s meaning is “mandated” by its “grammatical structure”). If the drafters of § 1442(a)(1) had intended the phrase “or any agency thereof” to describe a separate category of entities endowed with removal power, they would likely have employed the comma consistently. That is, they would have separated “or any agency thereof” from the language preceding it, in the same way that a comma sets apart the subsequent clause, which grants additional removal power to persons “acting under” federal officers. Absent the comma, the natural reading of the clause is that it permits removal by anyone who is an “officer” either “of the United States” or of one of its agencies.
Secondly, the language that follows “[a]ny officer of the United States or any agency thereof” confirms our reading of that clause. The subsequent grant of removal authority to any “person acting under him” makes little sense if the immediately preceding words — which ought to contain the antecedent for “him” — refer to an agency rather than to an individual. Finally, the phrase in § 1442(a)(1) that limits exercise of the removal power to suits in which the federal defendant is challenged for “any act under color of such office” reads very awkwardly if the prior clauses refer not only to persons but to agencies. An agency would not normally be described as exercising authority “under color” of an “office.” In sum, IBR’s interpretation of § 1442(a)(1) simply does not accord with the statute’s language and structure.
IBR tries to rescue its argument by invoking the well-established principle that each word in a statute should be given effect. See 2A N. Singer, Sutherland on Statutory
Construction §46.06 (C. Sands 4th rev. ed. 1984). IBR contends that any officer of an agency is also an officer of the United States and therefore that the reference to “agency thereof” in § 1442(a)(1) is redundant unless it signifies the agency itself. IBR notes, in support of this contention, that when Congress enacted § 1442(a)(1) it also defined “agency” as “any department, independent establishment, commission, administration, authority, board or bureau of the United States or any corporation in which the United States has a proprietary interest.” 28 U. S. C. §451. Since the words “of the United States” modify all of the entities listed in § 451, IBR concludes that an officer of an agency is necessarily an “officer of the United States.” Brief for Respondent IBR 16-17.
We find this argument unpersuasive. IBR’s broad definition of “officer of the United States” may well be favored today. Cf.
Buckley
v.
Valeo,
424 U. S. 1, 126 (1976) (“ ‘[O]ffi-cer of the United States,’ ” as used in Art. II, § 2, cl. 2, refers to any “appointee exercising significant authority pursuant to the laws of the United States”). But there is no evidence that this was the definition Congress had in mind in 1948, when it enacted § 1442(a)(1) and the companion provision defining “agency.” Indeed, in 1948 and for some time thereafter, the relationship between certain independent agencies and the “Government of the United States” was often disputed. See,
e. g., Pierce
v.
United States,
314 U. S. 306 (1941) (holding that an officer or employee of the Tennessee Valley Authority was not “‘an officer or employee acting under the authority of the United States, or any Department, or any officer of the Government thereof’” within the meaning of a criminal statute first enacted in 1884); see also
Rainwater
v.
United States,
356 U. S. 590, 591 (1958) (resolving a conflict among the courts of appeals and finding that a claim against the Commodity Credit Corporation was a claim “‘against the Government of the United States, or any department or officer thereof,’” within the meaning of the
False Claims Act);
United States
v.
McNinch,
356 U. S. 595 (1958) (overturning the Fourth Circuit’s decision that the Federal Housing Administration was not covered by the same provisions of the False Claims Act). Given the uncertain status of these independent federal entities, Congress may well have believed that federal courts would not treat every “officer of. . . a[n] agency” as an “officer of the United States.” Thus, the most likely explanation for Congress’ insertion of the “any officer of . . . any agency thereof” language is that Congress sought to eliminate any doubt that officers of the Tennessee Valley Authority and like entities possessed the same removal authority as other “officer[s] of the United States.” See
Cannon
v.
University of Chicago,
441 U. S. 677, 698-699 (1979) (“evaluation of congressional action . . . must take into account its contemporary legal context”). In any event, this reading of the “any agency thereof” language gives full effect to all of § 1442(a)(1)’s terms while avoiding the grammatical and linguistic anomalies produced by IBR’s interpretation.
B
Respondent NIH finds an alternative basis for agency removal power in the subsequent clause of § 1442(a)(1) that grants removal authority to any “person acting under him.” In NIH’s view, since the word “him” refers to an officer of the United States, an agency would be a “person acting under him” because each agency is administered or directed by such an officer. This is a rather tortured reading of the language. We doubt that, if Congress intended to give removal authority to agencies, it would have expressed this intent so obliquely, referring to agencies merely as entities “acting under” the agency heads.
NIH faces an additional hurdle, moreover, in arguing that the word “person” in the phrase “person under him” should refer to an agency. As we have often noted, “in common usage, the term ‘person’ does not include the sovereign, [and]
statutes employing the [word] are ordinarily construed to exclude it.”
Will
v.
Michigan Dept. of State Police,
491 U. S. 58, 64 (1989) (citation omitted; internal quotes omitted; brackets in original); see also
id.,
at 73. (Brennan, J., dissenting). This Court has been especially reluctant to read “person” to mean the sovereign where, as here, such a reading is “decidedly awkward.”
Id.,
at 64.
Nevertheless, “there is no hard and fast rule of exclusion” of the sovereign,
United States
v.
Cooper Corp.,
312 U. S. 600, 604-605 (1941), and our conventional reading of “person” may therefore be disregarded if “[t]he purpose, the subject matter, the context, the legislative history, [or] the executive interpretation of the statute . . . indicate an intent, by the use of the term, to bring state or nation within the scope of the law.”
Id.,
at 605 (footnote omitted). In the present case, NIH argues that Congress’ intent to include federal agencies within the term “person” in § 1442(a)(1) can be inferred from contemporary changes that Congress made in the federal administrative structure.
During the 15 years prior to enactment of § 1442(a)(1) in 1948, Congress created several independent agencies that it authorized to “sue and be sued” in their own names in both state and federal courts. In NIH’s view, these selective waivers of sovereign immunity gave Congress a reason to extend the removal authority to include agencies. Thus, NIH argues, the word “person” in the removal statute should be read as referring to such agencies. Although none of these early “sue and be sued” statutes involved major departments of the Federal Government,
we agree that those laws
could have
prompted Congress to change its removal policy. However, we find no persuasive evidence that Congress actually made such a change when it revised the removal statute in
1948. NIH concedes that each of the nine preceding versions of the removal statute, extending as far back as 1815, limited the removal authority to some subset of federal
officers.
See Brief for Respondent NIH 21-23, and n. 18; see also
Willingham
v.
Morgan,
395 U. S. 402, 405-406 (1969). In revising this removal provision to its present text, the House Committee Report offered only this comment to explain the change: “The revised subsection ... is extended to apply to all officers and employees of the United States or any agency thereof. [The predecessor provision] was limited to revenue officers engaged in the enforcement of the criminal or revenue laws.” H. R. Rep. No. 308, 80th Cong., 1st Sess., A134 (1947). This is the only legislative history on the 1948 revision and, as even NIH admits, it does not express a clear purpose to extend the removal power to agencies. See Brief for Respondent NIH 21. At best, the report language could be described as ambiguous on this point. Thus, the evidence that Congress intended to give agencies removal power is insufficient to overcome both the presumption against designating the sovereign with the word “person” and the awkwardness of referring to an agency as a “person acting under him.” Accord,
Mesa
v.
California,
489 U. S. 121, 136 (1989) (“[section 1442(a) . . . seek[s] to do nothing more than grant district court jurisdiction over cases in which a federal officer is a defendant”).
C
NIH argues, finally, that even if a literal reading of § 1442(a)(1) would exclude agencies from the removal power, we should reject that construction because it produces absurd results. See,
e. g., Public Citizen
v.
Department of Justice,
491 U. S. 440, 454 (1989) (court can look beyond statutory language when plain meaning would “compel an odd result”). NIH points out that if agencies are denied removal power the removability of the present lawsuit would turn on the mere
technicality of whether petitioners named NIH or only individual officers of NIH as defendants.
We think Congress could rationally have made such a distinction. As we have already noted, for more than 100 years prior to 1948, Congress expressly limited whatever removal power it conferred upon federal defendants to individual officers. NIH does not suggest that any of these earlier statutes produced absurd results; indeed, it acknowledges that, “[i]n drafting these removal provisions, Congress referred to federal officers because they, and not federal agencies, were the ones being sued in state courts.” Brief for Respondent NIH 23. The reason agencies were not being sued, of course, was that Congress had not consented to such suits and the agencies were therefore shielded by sovereign immunity. See,
e. g., Larson
v.
Domestic & Foreign Commerce Corp.,
337 U. S. 682, 693 (1949) (“suit to enjoin [federal action] may not be brought unless the sovereign has consented”); S. Breyer & R. Stewart, Administrative Law and Regulatory Policy 1018 (2d ed. 1985) (same). That fact, however, would not have prevented a plaintiff from erroneously naming — as NIH argues that petitioners have erroneously named — an agency as a defendant in state court. The first nine incarnations of the federal officer removal statute clearly reflect Congress’ belief that state courts could be trusted to dismiss the agency as defendant. The determination of an agency’s immunity, in other words, was sufficiently straightforward that a state court, even if hostile to the federal interest, would be unlikely to disregard the law. Thus, agencies would not need the protection of federal removal.
By contrast, the question of the immunity of federal officers who were named as defendants was much more complicated. Such immunity hinged on “the crucial question . . . whether the relief sought in a suit nominally addressed to the officer [was] relief against the sovereign.”
Larson
v.
Domestic & Foreign Commerce Corp.,
337 U. S., at 687 (footnote omitted). Often this question was resolved by examin
ing whether an officer’s challenged actions exceeded the powers the sovereign had delegated to him. See
id.,
at 689-690. Determining whether a federal officer had acted
ultra vires
was fraught with difficulty and subject to considerable manipulation. See Jaffe, Suits Against Governments and Officers: Sovereign Immunity, 77 Harv. L. Rev. 1, 20 (1963) (“The question always has been which suits against officers will be allowed and which will not be”);
id.,
at 29-39 (discussing seeming inconsistencies in this Court’s resolution of the question); see also Davis, Suing the Government By Falsely Pretending to Sue an Officer, 29 U. Chi. L. Rev. 435 (1962). Given these complexities, we think Congress could rationally decide that individual officers, but not agencies, needed the protection of a federal forum in which to raise their federal defenses. See
Willingham
v.
Morgan,
395 U. S., at 405 (“Obviously, the removal provision was an attempt to protect federal officers from interference by hostile state courts”).
The situation in the present case is no different from what would have obtained under the pre-1948 statutes. NIH’s defense in this case is precisely that it is not amenable to suit in state court by reason of sovereign immunity.
As noted, there is nothing irrational in Congress’ determination that adjudication of that defense may be safely entrusted to a state judge. The only question remaining, then, is whether the distinction Congress initially drew between agencies and officers continued to be rational in 1948, when Congress revised the removal statute. Although by then Congress had waived the immunity to suit of several independent agencies,
see
supra,
at 83, and n. 6, we find no fatal inconsis
tency in Congress’ determination that these few agencies’ other federal defenses (i. e., those aside from immunity) could be adjudicated in state courts. A crucial reason for treating federal officers differently remained: because of the manipulable complexities involved in determining their immunity, federal officers needed the protection of a federal forum. See
Willingham
v.
Morgan, supra,
at 407 (“[0]ne of the most important reasons for removal is to have the validity of the defense of official immunity tried in a federal court”); see also
Arizona
v.
Manypenny,
451 U. S. 232, 242 (1981). Accordingly, we see no reason to discard our reading of the current removal statute, which excludes agencies from this power.
IV
Having concluded that NIH lacked authority to remove petitioners’ suit to federal court, we must determine whether the case should be remanded to state court. Section 1447(c) of Title 28 provides that, “[i]f at any time before final judgment it appears that the district court lacks subject matter jurisdiction [over a case removed from state court], the case shall be remanded.” Since the district court had no original jurisdiction over this case, see n. 4,
supra,
a finding that removal was improper deprives that court of subject matter jurisdiction and obliges a remand under the terms of § 1447(c). See, e. g.,
Brewer
v.
Department of Housing and Urban Development,
508 F. Supp. 72, 74 (SD Ohio 1980).
Notwithstanding the clear requirements of § 1447(c), NIH asks us to affirm the Court of Appeals’ dismissal of this suit on the ground that a remand of petitioners’ claims to Louisiana court would be futile. NIH reasons that it is an indispensable party to the suit and thus that petitioners will be required, on remand, to retain NIH as a defendant (in which case the suit will have to be dismissed, since NIH cannot be
sued in state court) or to substitute an NIH official as defendant (who presumably will then remove the case pursuant to § 1442(a)(1)). Alternatively, NIH argues that even if the suit can proceed without an NIH defendant, Tulane will be able to remove the case under § 1442(a)(1) since, in caring for the monkeys, Tulane is a “person acting under” an NIH officer. See Tr. of Oral Arg. 30, 33. Obviously, if any of these events is certain to occur, a remand would be futile.
NIH finds authority for a futility exception to the rule of remand in
Maine Assn. of Interdependent Neighborhoods
v.
Commissioner, Maine Dept. of Human Services,
876 F. 2d 1051 (CA1 1989) (hereinafter
M. A. I. N.).
See Tr. of Oral Arg. 39. We believe NIH’s reliance on
M. A. I. N.
is misplaced. In that case, the plaintiff in a suit that had been removed under § 1441(b) was found to lack Article III standing.
The District Court invoked futility to justify dismissing rather than remanding the case, but the court was overruled by the First Circuit, which did remand the case to state court. Given the factual similarities between
M. A. I. N.
and the case now before us, we find that the result in
M. A. I. N.
supports our view that a remand is required here.
The purported grounds for the futility
of
a remand in
M. A. I. N.
were (1) the plaintiff’s lack of standing, (2) the state Commissioner’s declared intent to remove the case (following remand) in his capacity as a “person acting under” the Secretary of Health and Human Services (HHS), and (3) the ability of the Secretary of HHS (a third-party defendant) also to effect removal, as an “officer of the United States.” The First Circuit concluded that none of these anticipated barriers to suit in state court was sufficiently certain to render a remand futile. To begin with, plaintiff’s lack of Article III
standing would not necessarily defeat its standing in state court. Secondly, plaintiff’s suit challenged an action by the state Commissioner that was not necessarily an “act under color of [federal] office,” a prerequisite to the exercise of removal power under § 1442(a)(1). Finally, the First Circuit doubted whether the Secretary of HHS would be an indispensable party in state court. 876 F. 2d, at 1054-1055.
Similar uncertainties in the case before us preclude a finding that a remand would be futile. Whether NIH is correct in arguing that either it or one of its officers will be deemed an indispensable party in state court turns on a question of Louisiana law, and we decline to speculate on the proper result. Similarly, whether Tulane will be able to remove the remanded case requires a determination whether it is a “person acting under” the Director of NIH within the meaning of § 1442(a)(1). This mixed question of law and fact should not be resolved in the first instance by this Court, least of all without an appropriate record. We also take note, as did the First Circuit, of “the literal words of § 1447(c), which, on their face, give ... no discretion to dismiss rather than remand an action.”
Id.,
at 1054. The statute declares that, where subject matter jurisdiction is lacking, the removed case
“shall
be remanded.” 28 U. S. C. § 1447(c) (emphasis added). We therefore reverse the decision of the Court of Appeals and remand the case to the District Court with instructions that the case be remanded to the Civil District Court for the Parish of Orleans, Louisiana.
It is so ordered.
Justice Scalia took no part in the decision of this case.